DeepVu

AI Powered Autonomous Supply Chains

[deepvuinc]

VuDecide: AI Agent for Shock Resilient Demand Planning Now Available on Microsoft AppSource

For Immediate Release, July 10th, 2024 14:17 PT

San Ramon, CA--(PR WIRE)-- DeepVu, an innovator in AI Agents for resilient supply chain planning for manufacturers, today announced the availability of its flagship product, VuDecide: AI Agent for Shock Resilient Demand Planning, on Microsoft AppSource, an online cloud marketplace providing tailored line-of-business solutions.

Once creating a baseline demand forecast either from Dynamics 365 Sales or an internal forecast, AI Agent for Shock Resilient Demand Planning’s AI Agents integrate shock scenario planning, offering a compelling shock-resilient demand planning solution that gives human planners the power to choose from multiple shock-resilient AI powered forecasts. The beta release includes macroeconomic shocks such as consumer spend shocks (personal consumption expenditures) which impact all consumer goods verticals, and more shocks are being added based on customer demand. AI Agents are further enriched with world/industry context by taking hundreds of external signals from DeepVu’s rich Supply Chain Knowledge Graph “VuGraph.” Such signals include extensive macroeconomic signals such as interest rates, unemployment rates, average wages, commodity prices, export/import volume, and forex rates among others.

Giovanni Mezgec, Vice President, Modern Work + Business Applications Field & Partner Marketing, Microsoft Corp. said, “We welcome AI Agent for Shock Resilient Demand Planning to AppSource, where global customers can find thousands of line-of-business partner solutions that work with the Microsoft products they already use. Thanks to trusted partners like DeepVu, AppSource is part of a cloud marketplace landscape predicted to grow revenue 500% from 2022 to 2025.”

“Through DeepVu’s AI solutions, we strive to provide our manufacturing customers with best-in- class shock resilient planning in order to future-proof their supply chains against external shocks,” said Moataz Rashad, Founder and CEO of DeepVu. “The availability of our flagship VuDecide AI Planning Agents product in Microsoft AppSource enables us to offer these resilient planning and margin optimization benefits to a wider range of enterprise customers across the globe.”

The Dynamics 365 ecosystem is a compelling end-to-end suite of business operations and supply chain management solutions that offers an attractive value proposition to enterprises of all sizes globally. DeepVu’s suite of VuDecide Resilient Planning AI Agents and Knowledge Graph enriches this ecosystem with a set of compelling use-cases that optimize resilience, profit margins and sustainability for every enterprise that aims to future-proof its operations against external risks and disruptions.

About DeepVu

DeepVu (Vufind, Inc.) is an emerging AI innovator in resilient supply chain planning for manufacturing enterprises. DeepVu is pioneering a new category called autonomous resilient planning, in which generative AI planning agents are trained on top of digital twins that simulate multiple shock scenarios such as commodity spikes, consumer spend shocks, Mississippi river drought, labor and component shortages, trade restrictions, and more. The agent then recommends decisions along with their impact on the business KPIs. Human planners choose the actions to deploy fully informed by their KPI impact. DeepVu has R&D centers in San Ramon, CA, Belfort, France, Montreal, Canada, and Dublin, Ireland. For more information, visit https://deepvu.co

Contacts

Moataz Rashad, CEO

[email protected]

650.862.5113

[deepvuinc]

Demand Planning: Moving Beyond Basic ML Models with Shock Resilient AI Decisioning Agents

In the rapidly evolving landscape of business analytics, demand forecasting stands as a critical pillar for manufacturers  across FMCG  industries. Accurate predictions of customer demand enable businesses to optimize inventory management, production schedules, and resource allocation, ultimately driving efficiency and profitability. However, as many organizations have experienced, maintaining the accuracy of demand forecasting models can be a persistent challenge, especially with basic machine learning (ML) approaches that lack resilience to shocks and unforeseen events.

In recent years, the shortcomings of conventional ML models in handling sudden disruptions and maintaining accuracy over time have become increasingly apparent. Factors such as commodity spikes, new trade constraints and complexities, market fluctuations, consumer spend (Personal Consumption Expenditures), and unexpected events like supply chain disruptions or global geopolitical crises can quickly render traditional forecasting models obsolete, leading to costly errors and missed opportunities.

DeepVu is at the forefront of addressing these challenges with our VuDecide product, with its innovative Shock Resilient Decisioning Agents. These AI Agents represent a paradigm shift in demand planning, leveraging advanced AI techniques to adapt dynamically to changing conditions, recommending actions that optimize directly for your KPIs maintaining# accuracy in the face of uncertainty.

So, what can businesses do when their demand forecasting models start fading in accuracy? Here are some key considerations:

Evaluate Current Model Performance: Before making any changes, it's essential to assess the performance of existing forecasting models. Identify where and why inaccuracies are occurring, whether due to data quality issues, model limitations, or external factors.

Explore Advanced AI Solutions: Basic ML models may struggle to cope with the complexity and volatility of real-world demand dynamics. Exploring advanced AI solutions like DeepVu's Shock Resilient Decisioning Agents can provide a more robust framework for demand forecasting, capable of adapting in real-time to changing conditions and outlier events.

Integrate External Data Sources: Enhance the accuracy and robustness of forecasting models by incorporating a diverse range of external data sources. We call this VuGraph an expansive and continuously growing supply chain knowledge graph. This  includes macroeconomic indicators (interest rates, treasury yields, unemployment rates, wages etc),  commodity prices, production volumes, PPI (producer price index), weather data, or industry reports, providing valuable context and insights for more informed decision making.

Implement Continuous Learning: Static models quickly become outdated in today's dynamic business environment. Implementing mechanisms for continuous learning and model refinement ensures that forecasting algorithms remain adaptive and responsive to evolving patterns and trends.

Invest in Resilience and Flexibility: Recognize the importance of resilience and flexibility in demand forecasting. By investing in technologies like DeepVu's Shock Resilient Decisioning Agents, businesses can build a more agile and responsive forecasting infrastructure capable of withstanding shocks and disruptions.

Monitor and Iterate: Continuous monitoring and iterative refinement are essential for maintaining the relevance and effectiveness of demand forecasting models over time. Regularly evaluate model performance, solicit feedback from end-users, and iterate on improvements to ensure ongoing alignment with business objectives.

In conclusion, the challenge of maintaining accuracy in demand forecasting/planning models is a pervasive issue faced by many enterprises today. While traditional ML approaches may struggle to adapt to changing conditions and unforeseen events, innovative solutions like DeepVu's Shock Resilient AI Decisioning Agents offer a promising path forward. By embracing advanced AI techniques, integrating diverse data sources from a rich industry specific knowledge graph, and fostering a culture of continuous learning and collaboration, businesses can enhance the resilience and effectiveness of their demand planning efforts, driving better decision-making and sustainable growth in an increasingly dynamic marketplace.

[deepvuinc]

Embracing the Evolution: AI Planning Agents Assisting Human Planners

In the landscape of supply chain resilience and planning, a fundamental shift is underway—one that challenges traditional forecasting-centric approaches and embraces the transformative power of AI Decisioning Agents. Building upon the insights shared in my previous article, "Resilient AI Planning vs Human Planning," let's dive deeper into this unique relationship between AI decisioning agents and human planners in shaping the future of shock-resilient supply chain optimization.

Reimagining Planning Paradigms

In the realm of supply chain planning, the mantra "demand planning is the backbone of our supply chain efficacy" echoes throughout boardrooms and planning sessions. Yet, as the complexities of global markets and dynamic environments continue to evolve, the efficacy of traditional forecasting models, that have inconsistent accuracy and are unresponsive to shocks, is being called into question.

Enter AI decisioning agents—a paradigm-shifting process that augments the capabilities of human planners by leveraging the power of AI, reinforcement learning with human feedback, digital twin simulations and knowledge graphs.

Unlike conventional forecasting models limited by human cognitive constraints and lacking scenario planning capabilities, AI decisioning agents possess the capacity to process vast datasets, simulate many diverse scenarios, and optimize decisions for business KPIs in real-time. No more focusing on model performance in terms of MAPE metric, but rather on business outcomes (OTIF, freight costs, inventory holding costs, labor costs etc.

Autonomous Resilient Planning in Action

At the forefront of this evolution is the concept of autonomous resilient planning—a framework characterized by the orchestration of multiple AI decisioning agents trained on digital twins that simulate normal and shock scenarios. These agents, informed by a rich tapestry of external signals and a comprehensive knowledge graph, empower human planners with AI-driven insights and recommendations.

Within this framework, human planners serve as orchestrators, wielding AI superpowers to navigate the intricacies of supply chain dynamics and risks. They define the shocks scenarios, current or contemplated, that matter to the business, and let the AI system map them fully to the dynamics of the value chain. Armed with real-time KPI impact assessments and scenario analyses, human planners can make informed decisions to approve/deploy, selecting the most applicable AI agent to address specific contexts and challenges.

Embracing Complexity, Embracing Innovation

The integration of AI decisioning agents into supply chain planning heralds a new era of resilience and adaptability—one where complexity is embraced as an opportunity for innovation. From demand forecasting and production planning to inventory management and order fulfillment, every facet of the supply chain ecosystem stands to benefit from the agility and foresight afforded by AI-driven decisioning.

Moreover, the continuous learning loop inherent in autonomous resilient planning ensures that the system evolves and adapts in tandem with dynamic market conditions and emerging risk events. By enriching datasets with a diverse array of external signals (the supply chain Knowledge Graph), organizations can capture the pulse of the world and proactively mitigate risks before they escalate into disruptions.

Join the Conversation

As we chart the course towards AI-driven autonomous resilient planning, it's essential to foster dialogue and collaboration within the supply chain community. Have you encountered challenges or success stories in deploying AI decision models within your organization? Are there specific use cases or scenarios you're curious about exploring?

I invite you to share your insights, questions, and experiences in the comments below or on our page DeepVu. Together, let's harness the transformative potential of AI decisioning agents and human expertise to forge a more resilient and agile future for supply chain management.

The convergence of AI-driven innovation and human diligence and ingenuity holds the key to unlocking unprecedented levels of resilience, efficiency, and sustainability across global supply chains. Let's embrace the evolution and reimagine the possibilities of supply chain planning in this exciting digital age.

In that light, and given the timeliness of this topic, we’re launching a new “Resilient Planning using AI” podcast in March, and we invite all senior planning executives to comment or  DM me directly on Linkedin or email ceo (at) deepvu (dot)co with their interest to participate.

It’ll be only 19min, with 9 questions that will be shared with everyone selected to interview well ahead of your scheduled date.

Thank you for joining us on this journey of exploration,  learning,  and growth.

[deepvuinc]

Expanding the Spectrum of Supply Chain Shocks and Resilient Planning

The landscape of supply chain vulnerabilities extends beyond the macro-scale shocks that capture global attention. While major events like pandemics and trade wars seize headlines, minor yet impactful local shocks often go unnoticed. Consider the effects of a Mississippi River drought or a local worker strike – seemingly minor incidents, yet capable of disrupting regional supply chains, and costing manufacturers and distributors millions of dollars.

Moreover, shock scenarios need not be restricted to sweeping global occurrences. They can manifest within a company's realm, stemming from last minute unforeseen alterations in customer orders or the bullwhip effect in demand fluctuations. These micro-shocks might seem insignificant on a broader scale but hold the potential to wreak havoc within a specific supply chain, and often cause material impact on bottom line margins.

Furthermore, shocks aren't confined geographically. A disruption in a remote country like Thailand or Vietnam can reverberate across borders, indirectly affecting your supply chain. The interconnectedness of global trade means that shocks in distant regions can transmit ripple effects, impacting businesses far removed from the epicenter.

Navigating the Complex Web of Supply Chain Shocks

Contemplating and simulating these diverse shocks become paramount in resilient supply chain planning. While macroeconomic trends and geopolitical conflicts remain pivotal, the agility to simulate and have AI models that help planners respond to minor, localized disruptions holds equal significance.

Consider a scenario where a tier-2 supplier in a remote location faces unforeseen production challenges due to a regional issue. Even though this seems distant from your operations, it can lead to delays or shortages down the line for your tier-1 suppliers, and your Bill of Materials could change dramatically. Such instances emphasize the need for a comprehensive approach to shock simulation and preparedness.

Fostering Adaptive AI Decisioning Strategies

Supply chains built solely on forecasts are vulnerable to these myriad shocks. The essence of resilience lies in adaptability and readiness. Hence, simulating diverse shock scenarios isn't just about prediction; it's about cultivating adaptable strategies that can withstand shocks, big or small.

By envisioning scenarios that encompass local, regional, and even company-specific disruptions, supply chain planners gain agility in decision-making. This adaptability ensures continuity and minimizes the impact of unforeseen events, irrespective of their scale or origin.

Conclusion

At DeepVu, our commitment to resilient AI planning extends beyond predicting disruptions. We empower supply chain planners to navigate a spectrum of shock scenarios, enabling them to fortify their strategies against the unpredictable. Together, let’s transcend the conventional boundaries of supply chain planning and create resilient AI planning systems that thrive amidst all uncertainties.

Join us in redefining supply chain resilience. #AdaptiveSupplyChainPlanning #resilientAutonomousPlanning

[deepvuinc]

Navigating Uncharted Waters: Why Simulating Shock Scenarios Is Critical in Resilient Supply Chain Planning

In an ever-evolving global landscape, the art of supply chain planning has transcended mere forecasting. It's about foreseeing the unexpected, anticipating disruptions, and steering through the uncharted waters of economic uncertainties and geopolitical turbulences.

Consider this: as we navigate the complex web of global commerce, shock scenarios such as an expected recession/slowdown or geopolitical conflicts can cause ripples felt throughout supply chains. Recent geopolitical upheavals, record inflation in some countries, and an expected global economic downturns have starkly underscored the significance of preparing for the unforeseen.

Simulating shock scenarios, including economic downturns or geopolitical unrest, isn't merely a hypothetical exercise. It's a proactive strategy that can fortify supply chains against unforeseen disruptions. Here's why it matters:

1. Anticipating and Simulating Economic Fluctuations: With the global economy in a state of flux, simulating scenarios like a looming recession or growth slowdown in the USA isn't just prudent; it's essential. So, it's true the US economy is still growing strong as of this month, and the recession camp has been shifting it forward for a while now, so what's a planner to do? Can your company really afford to wait until macro economists declare it a recession! How about no recession just a drop form 4.9% YoY growth to 2%-- that's not technically a recession, but it surely will dent your revenue!

Predictive analytics can offer insights, but simulating these scenarios goes beyond prediction, empowering supply chain planning executives to proactively strategize for resilience.

2. Mitigating Geopolitical Risks: Recent geopolitical tensions and conflicts can significantly impact commodities, transportation routes, and international trade. By simulating the impact of such events, supply chain planners can identify vulnerabilities and devise contingency plans to mitigate risks.

Image: Ukraine's wheat being loaded onto cargo vessel. Src: Global Trade Review.

3. Importance of Resilience: Supply chains built solely on forecasts are fragile. Resilient supply chains, on the other hand, thrive amidst uncertainties. Simulating shock scenarios allows for the cultivation of robust, adaptable supply chain strategies that can weather the storm, ensuring continuity even in turbulent times.

4. Agility in Decision-Making: By envisioning and preparing for shock scenarios, supply chain planners gain agility in decision-making. They're equipped to swiftly adapt and implement decisioning that optimize directly for their business KPIs when faced with unforeseen disruptions, thereby minimizing the impact on operations.

At DeepVu, we recognize the transformative power of simulating shock scenarios. Our AI Decisioning Agents empowers supply chain planners with autonomous, resilient decision-making tools that thrive on adaptability, not just predictions. Our aim is not just to predict disruptions but to prepare your supply chain to navigate and emerge stronger from them.

In the dynamic realm of supply chain management, it's not merely about surviving disruptions; it's about thriving despite them. Let's not just predict the future; let's shape it, together.

Join us in reimagining supply chain resilient planning. #autonomousResilientPlanning

[deepvuinc]

Resilient Demand Planning with AI Decisioning Agents

Navigating today's supply chain planning landscape is more challenging than ever, riddled with unforeseen external factors like climate changes, geopolitical tensions, and labor-related disruptions.

Despite these shocks, manufacturers must maintain accurate demand planning to optimize production capacity, efficiently allocate inventory to DCs, and meet customer service expectations and OTIF delivery targets.

So what's a planner to do?

Enter autonomous resilient planning!

We can take a customer's demand forecast as an input, which maybe manual i.e. simply aggregated from human sales reps estimating which orders are coming by when, or it maybe a simple traditional ML based forecast. Then we add our own external signals data from our VuGraph knowledge graph and perform the shock simulations in our digital twins along with the shock's impact on your own supply chain dynamics. Consequently, the AI Agents are trained to recommend the production quantity levels that optimize for the customers business KPIs for both the nornal scenarion, as well as the shock scenarios.

It's important to note: while AI agents offer recommendations, the human planner retains the ultimate decision-making authority. Informed by the agent's insights and the recommended actions, the planner can make informed choices, carefully weighing their impact on business KPIs.

In summary, the integration of AI-driven decisioning agents into the demand planning process empowers companies to proactively anticipate and respond to disruptions, ensuring a more resilient and adaptive supply chain.

For further insights into how AI can optimize your demand planning, explore DeepVu's website and discover how these innovative strategies can boost your enterprise's resilience, margins, and sustainability. Feel free to email us directly at [email protected] or ceo<at>deepvu<dot>co

[deepvuinc]

Generative AI decisioning for Order Fulfillment : Three KPIs -- Delivery Date, Freight Cost and Emissions

Typically there's a tradeoff between Key Performance Indicators (KPIs) where, for example, if you use a greener more sustainable packaging material, you’d pay more for it.

However let’s take a supply chain use case in order fulfillment where two KPIs are very synergistic and if you maximize efficiency of the order fulfillment decision, you'd hit both KPIs, and that is avoiding splitting the order.

Using AI decision models for order fulfillment can be very efficient and powerful. By optimizing for multiple KPIs simultaneously, businesses can achieve significant cost savings and environmental benefits.

For instance, consider the use case of minimizing both freight costs and emissions in order fulfillment. Traditionally, companies face a trade-off between these two KPIs when choosing packaging materials for example. However, in inventory optimization and order fulfillment, these two KPIs are synergistic, and by maximizing the efficiency of order fulfillment decisions, businesses can satisfy both KPIs.

The key to achieving this is to avoid splitting the order and instead choosing a single distribution center (DC) to fulfill the order from. By doing so, the entire quantity in the purchase order can be filled from one DC and shipped on the same day. This means that only one truck is required to deliver the order, regardless of whether it is being delivered to a B2B customer, a retailer, or a consumer's home.

The tension, of course, is to do that while honoring the MABD (delivery date) which is a customer experience metric.

To accomplish this, an AI decision model is used to optimize the fulfillment process. The model is trained using historical order fulfillment data, including the customer order date, promised date (MABD), actual delivery date(s), and the DC(s) that filled the order for each SKU in the order.

The AI decision model is designed to optimize for a reward that includes both freight cost and emissions. The model learns from historical patterns how to fill the order to satisfy these constraints. It fills the order from a single DC that has sufficient stock levels to cover the ordered quantity for each SKU in the order without causing stockouts. Additionally, the model chooses a DC that is not too far from the customer's address, as excessive freight costs and trucking delays could cause the order to deliver late. Finally, the model avoids splitting the order into multiple DCs to optimize freight costs and minimize emissions.

Overall, the use of AI decision models in order fulfillment can help businesses achieve significant cost savings and environmental benefits by optimizing for a Reward function composed of multiple KPIs simultaneously, i.e meeting MABD, freight cost, and emissions.

#order_fulfillment, #supplychainOptimization, #autonomous_planning #resilient_planning #inventoryOptimization #supplychain_planning

[deepvuinc]

Perpetual shocks: can anyone predict the next supply chain shock?

Perhaps you have Yoda on your planning team – he’s got centuries under his belt, he’s seen every conceivable shock in the multiverse and he’s basically a 99.99% accurate crystal ball so perhaps he can predict the next shock. But the vast majority of us mere mortal planners cannot, and kidding aside, Yoda isn’t recruitable!

The C-suite execs acknowledge it’s not realistic to predict a pandemic, an EU war or a continuously morphing trade war, however, they still expect their planners to be ready to mitigate these shocks and have a “resilient” plan that optimizes the business KPIs. The supply chain planners are stressed, the C-suite execs are stressed and the suppliers are stressed!

So what’s the solution?

Clearly, the solution is to actually envision and map out all potential shocks, rank them in order of likelihood, and simulate each one of the top priority ones in your supply chain digital twin. An AI Decision Model would subsequently recommend actions for each one of these scenarios, along with their impact on the business KPIs, and your planners can grade such actions and select which ones are most appropriate per geography, market segment, set of suppliers, set of SKUs etc.

Scenarios, Scenarios, Scenarios

If this brings to mind what military leaders do in their war game simulations, you’d be right on point. Even though, luckily, this is operational planning not an armed conflict,  it’s very much analogous to fighting against an invisible enemy such as a pandemic - causing factory shutdowns and container pile up at seaports, or climate-catastrophe-causing droughts impacting semiconductor fabs, or a crippling components/materials shortage.

Okay so you get it, the solution is a decision model trained on a multi-environment digital twin that can simulate numerous shock scenarios. That seems logical, we can get behind that.

But it’s still missing a key component.

So let's say the overall shock is, to take a very timely one, the war in eastern Europe. It’s clearly imposing global shortages of wheat.  As you can see in the price chart below, Wheat was a $618 $/Bu back in July 2021! On May 31st, it was $1103 $/Bu, and today it's at $774 $/Bu, which is understandable since commodities markets tend to price-in near term forecasted shortages.

Src: Trading Economics

That key component is the supply chain Knowledge Graph those external signals that capture the shocks in near-real-time as they manifest. Such signals such as the primary, secondary and tertiary commodities that impact a given industry vertical. They also include all the macro-economic signals that influence and interact with the industry specific signals such as interests rates, unemployment, CPI and PPI etc.

The more expansive and cross-industry a knowledge graph is, the more impactful its signals and predictive weights.

Role of Governments and Resilient Planning

The most obvious example of this is the latest baby formula shortages due to the closure of one of Abbott’s Similac factories in the US due to contamination. Clearly no one in the FDA simulated that scenario, and everyone was scrambling to find production capacity with other manufacturers to cover the shortage to feed American babies. Ultimately, it took the POTUS invoking the Defense Production Act to instruct the other manufacturers to increase production capacity to meet the formula demand. Luckily this intervention was possible this time, but clearly one can envision scenarios where it may not be, and it could unfortunately lead to loss of life!

That’s the importance of putting scenario-based resilient planning and risk mitigation decisioning at planners’ fingertips so they can see all recommended mitigation actions and their impact on the KPIs, and choose the best decision for the scenario at hand!

Conclusions

Supply chains are truly a matter of life or death and national security at this point, whether they affect infant formula, food commodity shortages, vaccines, silicon components, battery materials, or otherwise. 

It will take public private partnerships and the best resilient planning solutions to address this pressing problem effectively and at scale. It is a top priority for the nation this decade and the decades to come. It's our focus at DeepVu and we're available and eager to partner.

[deepvuinc]

Sustainability must be a primary business KPI. Three practical approaches to make it a reality.

California wild fires 2020.   Source: The Desert Sun.

By Moataz Rashad

Feb 17, 2021

         Last year’s California wildfires and air quality issues piled on top of a pandemic lockdown have made life comparable to a hub on an inhospitable remote star.

Earlier this year in January, NASA’s Goddard Institute of Space Studies released a study  back-testing older climate projection models and confirmed their accuracy. More recent models are even more sophisticated and higher accuracy. We’ve turned a beautiful planet into an ailing mess. With President Biden’s decisions on his Day 1 to rejoin the Paris Agreement on climate change, and appointing Mr Kerry as Climate Envoy, we’re finally on a good path.

It is high time that the manufacturing and construction sectors take sustainability seriously and heed that policy guidance. AI can facilitate that in an automated seamless way.

Many manufacturers produce annual sustainability reports. They often read like everything is great, and you’d think that if the top 50 companies in every manufacturing sector were as mindful of sustainability as these reports would make it seem, then we shouldn’t really have any environmental problems.

The truth, of course, is that the manufacturing supply chain is amongst the top 5 polluters in every country across the globe.  Every major manufacturer has a lot of room to improve.

The most conscientious manufacturers have recently been addressing the issue seriously and exploring how AI/ML can play a pivotal role in effectively optimizing their sustainability in every stage of their supply chain.

Some have even reached out to us specifically for advice on AI for architecting approaches for optimizing sustainability in procurement or logistics, which is certainly a very healthy and encouraging sign.

It’s tempting to reduce sustainability to recycling, using greener materials, diminishing  your energy footprint, and so on. And while these individual steps certainly help, the problem is much broader than that.

In this article, we’ll touch upon three top level proposals that every Chief Supply Chain Officer or COO ought to consider in their supply chain sustainability plans for 2021 and beyond.

Sustainability in Procurement

It goes without saying that you may have perfect sustainability practices in your own manufacturing operations and internal processes, but if the components you buy are from dubious suppliers, your overall product sustainability score will suffer and your product’s environmental footprint will be negative. Think of it as the multiplication of these individual component scores: all it takes is for one key component to have a sustainability score of 0. Therefore, it is important that you investigate every supplier you onboard for their own sustainability practices. It simply can’t be based on the stale, static questionnaire forms that many companies still use today.

Some industries, such as automotive, have strict legislative sustainability compliance requirements, which makes supplier review and qualification a bit more manageable since every supplier has to have their compliance reports. But many industry sectors don’t and it is certainly true that compliance reports can be gamed!

There are three AI/ML modules that can help address sustainability in procurement. You can start with any one of these, and then phase in the rest. However, ultimately, you will need all three for an end to end procurement sustainability optimization.

Build your own sustainability scoring model for your suppliers. The model can use those supplier questionnaires as an input, but they are just one low-trust input.

Almost all procurement departments we work with rank suppliers in quality tiers (A, B, C)  that are largely based on yield, OTIF fill rate, and capacity. Sustainability should be a separate quality tier; the model will treat it as such and it will have its own feature weight.

Build an AI decision model that would allocate POs to suppliers to optimize the Bill of Materials for an overall KPI, where overall sustainability is one of those highly weighted KPIs. Of course, Costand Delivery Date are typically KPIs as well. Thus, you have a Reward that is a weighted formula such as  R = w1* SustainabilityScore + w2* Cost + w3*DeliveryDate

Build a Supplier Risk decision model that can recommend actions as follows:

Approve a supplier for a given country/region

Upgrade/downgrade the tier-rank of a supplier 

Increase/decrease the risk score of a given supplier

Initiate an eDiscovery process pertaining to a specific risk item (material-id)

Initiate an investigation into a supplier’s claim/statement in questionnaire

Flag a supplier for an onsite investigation

Disqualify a supplier

Warren Buffet has famously said “It takes 20 years to build a reputation, and 5 minutes to ruin it. If you think about that, you’ll do things differently. “ It could even be 5 minutes by one of your suppliers that cut corners and ended up ruining your product’s sustainability score, not your own team or process.

Sustainability in Manufacturing

The vast majority of manufacturers have their own plants but also augment them with contract manufacturers. Therein lies a major problem for your product’s environmental impact. You do have full control over your own plant operations, the eco-friendly materials you procure, their treatment, storage and disposal of toxic waste, and the energy usage in the plant, to name a few.

For your own factories, you can build a decision model that recommends actions that directly impact your Triple Net Zero goal (Energy, Water, and Waste). However, it’s very hard to have full visibility into contract manufacturers’ practices and compliance, and where they stand on each one of these sub-goals.

Some of the largest manufacturers currently do surprise inspection visits, which are a definite part of the solution, but are not sufficient on their own. In fact, in healthcare, the FDA is known for frequently making these surprise visits to international drug manufacturing operations, which has been quite effective in improving generic drug quality and compliance.

For a coveted tier1 customer, CMs can also be required to provide video feeds of operations on a recurring basis, and those feeds can be analyzed for credence and potential violations.

Once this data is available, it can be fed into an AI decision model that can then recommend the build-order-quantity and target dates per CM in order to optimize for safety stock, time-to-market, and the sustainability score of the final product.

Sustainability in Logistics

Supply chains have two logistics links -- the first is procuring raw materials and components and getting them shipped to the factories, and the second is shipping the finished product to distributors and customers. Therefore, optimizing your logistics network is  crucial to your overall sustainability score.

It’s a given that shipping and freight forwarding has a high environmental impact, so you’ve got to look at your carrier and distribution networks closely and build a model that makes these carrier, route/lane, and shipping date decisions intelligently to optimize for MABD (Must Arrive By Date) along with the sustainability score. Making these decisions with high efficacy in a COVID shocked environment such as 2020 has proved challenging even for the most successful manufacturers and 3PLs. AI can play a key role in building upon and improving these human decisions. 

Conclusions

Despite  good intentions, a patch work of sustainability report here, and a line item in a questionnaire there haven’t and can’t be expected to deliver quantifiable sustainability improvement results. If we’re serious about optimizing sustainability in the  manufacturing sector, we think the viable solution is to make it a highly weighted business KPI for the supply chain AI models. These decision models will take care of the actions and the KPI Reward outcomes, and will continue to improve and optimize that KPI with usage. That’s the path to a green future!

Reference:

Sustainable Global Value Chains https://link.springer.com/book/10.1007%2F978-3-319-14877-9

How to measure a company’s real impact https://hbr.org/2020/09/how-to-measure-a-companys-real-impact

A comprehensive framework for automotive sustainability assessment

https://www.sciencedirect.com/science/article/pii/S0959652616309155

[deepvuinc]

Supply chain resilience and risk intelligence in a post-pandemic world

By Moataz Rashad

     Supply chain teams at most large manufacturers are currently in crisis management mode. For some companies, demand has shrunk by 40% or more, whereas for others it has narrowed to the essential SKUs category. The majority of such teams aren’t yet at a stage where they’re allocating intellectual bandwidth to the post-pandemic environment and the wide-ranging implications of new norms like social distancing, higher hygiene requirements, tighter trade constraints, slower logistics, and consumer unease.

So we at DeepVu thought we’d spark the discussion by posting our thoughts and observations based on our own customer engagements and recent conversations.

Crisis-Adaptive Demand Planning

Recently, it’s become clear to everyone across the supply chain ecosystem that in a crisis of this magnitude, people’s psychology is significantly impacted. Buying patterns, for both consumers and businesses, shift accordingly.

In fact, many experts predict that these demand patterns will continue to shift during the partial economic re-opening, the full re-opening, and may even linger for a while past that. Nielsen compared YoY % change for certain essential goods during the Feb/Mar period in the graph below.

There are two types of inventory optimization approaches: a) forecasting and b) decisioning models. A forecasting model would forecast a stockout of SKU N at Warehouse W 5-weeks out. A decisioning model would actually recommend an action such as "change the replenish rate for the specified SKU at the specified warehouse". A decisioning model trained on a dataset spanning pre-pandemic, during the pandemic and into the partial re-opening, has the added benefit of continuously learning to recommend dynamic decisions that improve over time. In fact, the model will prove even more effective for any crisis that causes comparable demand pattern shifts down the road.

Forecasting stockouts accurately and preventing them could literally save lives, whether the item is a drug, a device, a mask, or any PPE. Even a disinfecting wipe that helps an elderly person ensure whatever items they just bought aren’t contaminated can be critical.

Procurement KPIs Are Changing

Most folks in procurement consider BoM optimization to be a major priority and surely it’ll continue to be a key focus. However, supply assurance at times of crisis becomes an equal priority. A reputable tier 1 supplier that would not grant you the allocation you need or commit to the delivery date required for your material flow dependencies should certainly get a lower reliability score. On the other hand, it’s equally important to ascertain whether a supplier is indeed honoring the requested quantity, but its delivery dates are impacted by logistics network delays outside of its control, in which case that supplier’s score shouldn’t be penalized.

For manufacturers with over 10,000 suppliers, there’s no human procurement team that can possibly keep track of these factors and the corresponding scores manually.

An ML model that generates a Supplier Credence Score becomes of critical importance at times of supply chain disruption like these. Such a model needs to be trained on historical data spanning “normal” pre-crisis times, as well as the procurement and delivery records over the past 8 weeks of crisis activity.

Our world is full of uncertainties, but one certainty is that there will be other shocks to the supply chain ecosystem. Clearly, such an ML model would be invaluable to use for supply risk mitigation when the next ecosystem shock arises, whether it’s health-related, climate-related, or otherwise.

Locally Sourced Supply Chains

The trade war chatter has been getting louder, and we already see many customers starting to think about reducing dependence on foreign suppliers, and sourcing locally as much as realistically feasible. If parts that used to take one week to arrive can now take 21-30 days to do so, and with logistics surge-pricing and tariffs added on top, it may certainly be more cost-effective to source locally in many cases.

Furthermore, with respect to actual workplace and work flow changes, our view is that the changes will vary widely by sector and by company size, so we’ll only provide examples from a few sectors that we understand well enough to think through and make predictions for.

Workplace and Work Flow Impact

1. Construction

We foresee significant changes in the construction industry. Construction workers already have very tight regulations for protective gear and safety and these are likely to get even tighter, requiring smaller crew sizes and physical separation between workers. Such constraints will affect work flow and schedules in many ways. Imagine a high-rise building where the electrical and plumbing crews have to now work on alternating floors, and each crew is limited to only 3 workers, while also maintaining social distancing guidelines.

The construction supply chain includes both digital (blueprints) and physical supply chain such as building materials, excavators and cranes, and so on. Scheduling efficacy for material, equipment, and labor flow is a key priority for construction projects, and it does have a substantial impact on ROI as delays often cost millions of dollars.

An AI model that is trained on all these datasets and historical schedules, and recommends decisions representing integrated schedules that cover labor, equipment and materials, along with their cost and overall project KPI implications. Such a model would be an invaluable AI assistant. This model would eliminate the error-prone process of coordinating multiple disparate schedules, separate cost implications, budget overruns, and more. Additionally, if the model is trained on pre-crisis (“normal”) times project data, along with some mid-crisis projects, it will be crisis-hardened and will perform intelligently for scheduling during future ecosystem disruptions.

  2. Manufacturing

Factories are unique environments as they have well-thought-out material flow plans, and with robotic arms and autonomous forklifts, most have been ranging from 30-80% automation depending on the vertical and the product. A natural decision for a manufacturing GM in a pandemic environment is to acquire more robots and increase the level of automation of their production line. In this scenario, ideally a good percentage of the manufacturing personnel whose duties become automated get transferred to warehousing or re-trained for other roles in the organization. Opting to go for a higher level of automation by deploying more robots and autonomous forklifts inherently requires re-layout and scheduling changes to deliveries and shipments. If production scheduling and production capacity planning are done old-school, the above changes will induce chaos.

However, if one has a decision model that produces recommendations on production schedule, labor allocation, robot allocation, and material flow, then all the changes above would be reflected in the simulation environment, and the AI decision model would adapt its recommended actions to optimize the Reward accordingly.

  3. Warehousing and Logistics

Although they work great in normal times, static warehouse networks are just too constraining at times of crisis. On-demand warehouse capacity allocation is invaluable at times of disruption. This crisis has highlighted to fulfillment managers that they really need access to both on-demand storage as well as static in order to meet delivery dates and the dynamic fulfillment needs in an increasingly fluid ecosystem.

Furthermore, track and trace will be more important than ever, for essentially all items. It's been very successful for big ticket items and in pharmaceuticals for quite a few years. However, increasingly, it'll just be expected as a core feature, and the risk management and insurance companies will demand it for all B2B commerce in most sectors.

Track and trace has primarily been implemented used IoT, and it is actually one of the highest drivers of IoT spend-- see the Forrester chart below. However, an AI model that's been trained on historical IoT trace data can be used to forecast whether a current shipment will get damaged, tampered with, or even the exact minute it'll get delivered and so on. A forecast that carries high confidence may even directly trigger an insurance claim.

Conclusions

I have firm conviction that we'll get through this crisis together and we'll emerge stronger, smarter and more diligent as a critically important industry that civilized society relies upon. However, it is imperative that we first spend the time and intellectual cycles on evaluating existing inefficiencies, organizational readiness, and lax measurement of underperforming models. Second, identify the highest priority use cases that need to be addressed first, and embark on a diligent evaluation of credible AI/ML proposals that address it. Third, establish the current baseline accuracy for your existing model or rule-based software against your current and desired forecast horizons. Finally, commit to AI/ML as a strategic long term approach, and deploy piloted models gradually in production, and keep scaling them one use case at a time. The companies that do this methodically will win in both normal and abnormal times.

Stay safe, and I look forward to your thoughts and comments.

[deepvuinc]

Federalized supply chains at times of national emergencies

By Moataz Rashad & JC Zhou

DeepVu is an AI company focused on optimizing supply chain efficiencies with smarter forecasting and decisioning models. Our engagements with tier1 manufacturing leaders in multiple verticals such as electronics, construction and consumer goods gives us a unique vantage point.

In terms of the current pandemic, the highest priority components needed for caring for COVID19 patients seem to be ventilators, hospital beds and masks. Let’s take ventilators, which are essentially a medium complexity electronic device that’s composed of a pump, hose, display, and a simple microcontroller for regulating the flow. It essentially breathes for the patient, pushing oxygen into the lungs and also takes co2 out.

What would a Federalized supply chain entail for this? 

For context and clarity, a supply chain for a medical electronic product like this has the following steps:

Procurement of raw materials and manufactured components  (ex plastic/PVC for casing, and microcontroller chip/board) from multiple suppliers

Storage at a distribution center

Shipment of these components to a plant/assembly

Manufacture, Assemble and Test

Storage at a distribution center

Receive orders from various hospitals and clinics

Shipment of the product to hospitals and clinics

Note that the design and specifications of the device itself are already FDA approved and since we’re following that spec, there are no additional approvals required.

How would the Federal Gov help expedite and boost this process? The government can provide the following:

Air freight capacity -- The military has a massive cargo fleet and this can help tremendously in getting the microcontroller chips which would typically come from TSMC in Taiwan, as well as other components that would come from US companies, and need to be delivered to the factories

Distributed Storage capacity -- both the Military and USPS have massive warehouses all over the nation

Shipping capacity -- similarity both the Military and USPS have the logistics network to distribute these ventilators and components from the DCs to the hospitals across the nation without relying on a 3PL carrier!

So you can see that other than step (4) above, the Federal Gov can significantly contribute to boosting the efficiency of this supplychain.

Note that this doesn’t invoke the Defense Production Act in the sense that we’re not assuming the president will force commercial companies to do any parts of this. If the Defense Production Act is invoked, then the manufacturing step itself would benefit b/c more companies that Medtronic and Philips will be asked to manufacture the ventilators 

We’re just using the Federal Gov resources to help address current gaps and materially optimize the speed of getting products to the point of care where the patients need them. 

In conclusion, we think that the Federal government can indeed help in saving our fellow citizens lives by boosting the efficacy of distribution of ventilators, masks and test kits. May God protect the elderly and the vulnerable! 

[deepvuinc]

Nintendo’s scarcity, and ensuring on-shelf availability

By Moataz Rashad

Recently Nintendo has been facing a problem many consumer electronics manufacturers envy-- consumers love their game consoles! But now the iconic company has a brand image problem, ie. why can’t it accurately forecast the demand, and ensure its new products are available at its retail partners.  See this and this.

In a nutshell, there are two problems in play here: a) Demand forecasting, and b) On Shelf Availability. In this case, Nintendo is struggling with both. Firstly, they under-forecasted initial demand for the new product at the launch date that they selected. Secondly, subsequent to the actual launch, they didn’t coordinate their supply chain with the retailers’s data (point of sale and distribution network) to ensure On Shelf Availability for the product so its eager customers can buy it.

On Shelf Availability is a sub problem of demand forecasting that deals with optimizing the likelihood that a customer who walks into a store (physical or online) searching for a product X, will find that product both in-stock and, in case of physical stores, actually on the shelf to complete the purchase.

It is clearly the one metric that ensures no sales are lost. If a customer intends to buy, they will find the product available at the time and location of their choice.  

This problem is a key supply chain problem that’s been around for decades. It requires close real-time collaboration between the manufacturers and the retail channels with daily updates of on-hand inventory and point-of-sale transactions. But most importantly, it requires deep-learning to solve it at scale. 

Unfortunately, traditional machine learning techniques deliver only 55--65% accuracy which costs millions in lost sales. They are unable to ingest the numerous external signals that impact demand, and they are unable to handle the massive volume of data the tier1 manufacturers and tier1 retailers deal with. So they’re left with building numerous disparate models instead of one large monolithic deep learning engine that can learn from all the patterns and transfer learnings from one store’s patterns to a sister store etc.

A deep learning solution is able to handle all kinds of data feeds that impact the customers’ demand patterns, including economic and demographic data which vary by vertical and product category. It also ingests the PoS and inventory data along with daily feedback from the manufacturer and retailer in order to adapt and self-tune its predictions in real-time. That’s what Nintendo and comparable major enterprise manufacturers need in order to address these key supply chain challenges that could potentially make or break the company.

[deepvuinc]

Future of commerce: AI replaces BI

For the past couple of years, many well-respected folks have opined on the future of retail, claiming that all brick and mortar retail is dead, dying, or in a vegetative state, and that the future of shopping therefore lies entirely online! See "future of shopping" and "retail stores will die"

I certainly wouldn’t mind that notion materializing at all, since for obvious reasons, the more commerce moves online--which is an irreversible trend--the better for us at Vufind, a startup building the AI layer for eCommerce.

However, I think that the hypothesis is flawed. The future of retail is AI-commerce, i.e. smarter shopping experiences in every sense of the word, whether it happens online, offline, or hybrid. “Smart” as opposed to “not so smart” commerce, as in what you often see now i.e : a) having to wait in line to buy a staple product for which there is little choice, minimal price variation, and there’s no emotional investment in the purchase b) being chased after you've left a store (offline or online) and abandoned an item to be shown the same exact item over and over (aka retargeting) or c) the seller keeps telling you those who bought this also bought that, as if all shoppers are automatons (aka collaborative filtering)

See this emarketer presentation for more detailed proof points on consumer data regarding how they view online vs. offline shopping.

It is indisputable that online retailers have much more attractive cost structure, and are advantaged by the digital consumer and product data that can drive decisions in real-time at nano-second speeds. However, offline retail has some advantages as well, including immediate delivery, try before you buy, impulsive buys, and lower return rates.

Some online-offline hybrid experiences are already happening. For an example of online companies expanding offline, just monitor the investment moves of Amazon, one of the most forward-thinking online retailers globally, and you’ll observe that they are serious about online-offline interplay. Similarly, Apple retail has a hybrid show-room/store model which is being copied by other consumer electronics companies. On the other hand, a forward-thinking brick and mortar retailer such as Nordstrom is experimenting with exciting online-offline hybrid experiences with the smart mirror for example. It will be intriguing to see the A/B test results of these efforts as they will pave the way for even more fascinating experiences.

Courtesy: Nordstrom-ebay fitting room

So what is smarter commerce, or AI commerce? AI commerce has one broad goal: matching the shopper with their desired product by factoring in their style, brand preference, price range, size, fit, and delivery time, in order to effect a frictionless speedy transaction.

If a brick and mortar retailer builds a sophisticated site/app but they are primarily interested in driving traffic to their physical stores, then they aren’t fostering an AI commerce experience. By the same token, if an online-only retailer gets user feedback in favor of same-day delivery and ignores this input, then they aren’t developing an AI commerce experience. So what would an AI commerce experience entail?

Let’s start with some definitions since AI is such a land-mine these days, what with everyone from famous loved entrepreneurs such as Elon Musk and Bill Gates to the occasional scientific journalist speaking on the topic.

AI in this context refers to the field of Artificial Intelligence; software algorithms that use pattern recognition, computer vision, deep-learning neural networks, or whatever other techniques we practitioners come up with, to produce intelligent decisions that impact the shopper’s user experience. Such decisions include recommendations, merchandizing, dynamic pricing, promotions, placement of products on sites/apps/store-window, or whatever decision the marketing/product management executive needs the algorithm to assist with that day. In other words, we’re not talking about an artificially intelligent entity such as a robot, vehicle, drone, or anything else that can do any harm to people in the real world. And we’re certainly not talking about “Strong AI” which is where most of the fears of the critics revolve, where the algorithms would decided to learn from data outside their domain, and in fact will consume all accessible online knowledge, and for some reason they'll take special interest in learning about destroying humanity!

In AI-commerce, we're only discussing deep-learning software running in the cloud with no tentacles in the physical world, and where it’s trained on where to look, what to look at, and what type of output we'd want it to produce, and it's only domain is commerce.

This begs the question-- how is AI different form Business Intelligence (BI) which has been the traditional software layer of eCommerce? At the highest level, the fundamental difference is that BI is static or doesn't learn from the data, while AI's missions is to continuously improve and learn from the data to get "smarter". In other words, Business Inteligence is a misnomer, it's really business reporting, or business stats-- dry static stale reports on what's already happened a while ago.

So what do these AI algorithms work off of? the online shopping experiences have the upper hand here, because every action is already digital, cloud stored, and hence trackable and learnable in real-time. Every action including product impression, image click, recommendation click, purchase, like, save, share, cart-abandonment,email open, email link click, etc, is fed to the algorithms with the goal of making the shopper’s experience smarter and more pleasing.

We believe a true AI-commerce experience should accommodate most of the following:

1. Smarter visual recommendations: A smarter experience means that you get to see a recommendation that visually matches your style preference so much that it persuades you to purchase the recommended item rather than the item you originally searched for. We get super stoked when we count those transactions.

2. Smarter personalized recommendations: If someone is a loyal repeat user of an etailers site, then they do in fact expect a personalized experience that accounts for more than gender and age group, regardless of whether or not they use a social login. If they have purchased before, browsed, liked, and abandoned carts before, they expect the site to be smart enough to know their taste preferences and which products they would like to see. How is this different than (1) ? A user might be looking at a heather grey lambswool sweater, and you could show her ski goggles and cycling shoes, because you’ve learned that she’s into the sports associated with these items and haven’t purchased them in over 9 months. You might even show her girl’s soccer shoes because she bought a pair for her daughter last February. Or you might surprise her with an item that she’s never bought from your site before because it matches her style.

3. Smarter merchandizing: the more AI-powered decisions the retailer employs, the smarter the product mix in their catalog. As such, their buyers would know exactly what to stock each season based on transaction trend analytics, thus eliminating waste across the supply chain therefore saving time, and resources.

4. Psychic Personal Shopper Apps (PSA): A truly intelligent PSA wouldn't just have a warm voice and uses Google and Amazon to run searches for you and send you a list, it would know what you'd want and would just get it delivered to your door step. And if you dislike the product, you tap a button and it'll get it returned for you as well. It'll know that it's time to re-order socks, or that winter is coming 2 weeks earlier, so it'll order those thermal under-shirts. And it'll buy a different tasteful gift for your best friends' birthdays each year, and it'll ask you what the friend said about it so it gets better at gifting. Most importantly, it'll do all that algorithmically without any *crowd-sourcing* or asking you to fill out forms.

5. Smarter "show-rooms": Recall when show-rooming was considered a threat, well it isn't, it's part of the shopping experience now, and every online and offline retailer designs for it. However we think in the very near future, the physical shops will evolve into AI powered show-rooms, amazingly chic physical spaces that allow shoppers to touch, feel, try on, learn about the origin of the product, or quickly browse similar products to explore. But not necessarily wait in line or carry it home, unless in fact they do need it that exact moment.

6. Smarter payment: Some of the amazing international etailers we work with offer cash payment upon delivery. Some have added BitCoin. Some experimenting with pick up at partner physical stores. Retailers should certainly make it as frictionless as possible to pay for the item you’ve already invested the time and emotional energy to shop for and decided to purchase. So smarter payments means choice, speed, convenience, and quite frankly none of those hideous 2.9% fees, which I personally think should be eliminated.

7. Social: Many interesting developments there, but that’s for another article.

The history of commerce goes back thousands of years. It’s catering to a human need that’s existed since the oldest civilizations used the barter system. The future of commerce, however, is truly awesome; it's intelligent commerce powered by AI, easier, faster, cheaper, more relevant and more delightful. It’s already happening.

Are you excited about smarter commerce?--would love to see your thoughts below.

[deepvuinc]

Will AI “save” manufacturing?

By Scott Lyon

“Manufacturing Loses Most Jobs Since Election” reads the latest headline.

According to the latest job report, the U.S. manufacturing sector once again reported job declines (4,000 jobs lost).  This further extends an unfortunate erosion in our country’s manufacturing base and in fact Institute of Supply Management (ISM) surveys suggest job losses will continue.

Although efforts by the current administration to reverse the trend toward “off-shoring” as well as aggressive incentives prepared by the states (e.g. check out Wisconsin’s just-announced Foxconn package) may prove helpful in the short-term, however, clearly the manufacturing sector needs to reinvent itself and develop new market opportunities.

One increasingly popular theme for doing so involves Artificial Intelligence (AI).  Namely, could deeper machine learning enable manufacturing companies to either reduce costs and/or drive incremental revenues without launching new products or securing additional customers? There are numerous conceptual opportunities suggesting the answer is yes, here are three recent compelling use cases:

Two ex-Apple engineers have created hardware which takes photos at critical junctures of a manufacturing line and can pinpoint in real-time when devices are being assembled incorrectly (all without requiring on-site visits by managers or quality control inspectors);

GE launched an internal IT initiative which is now providing advanced analytics to external customers for example advanced sensors for their oil / gas clients can better predict blade health and determine optimal replacement times;

One of the leading semiconductor companies is leveraging DeepVu’s AI engine  to determine the “optimal production allocation” for maximizing chip revenue and optimizing yield amongst multiple contract manufacturers;

In each of these three examples, manufacturers are leveraging their existing data and service offerings to drive either reduced equipment failures or greater revenue-per-channel.  

[deepvuinc]

Supplier Pricing Breakdown Optimization: A Complex Data Problem

Every supply chain executive is concerned about optimizing margins in this competitive environment.  Supply chain costs are typically 60-70% of operating costs and hence a real focus in executive discussions.  Supply chain leaders are constantly asking themselves questions about how to do things faster, cheaper and better.

Are you paying the lowest possible price to source parts for this product to optimize the BoM while maintaining the same quality?  Is that a question you have the tools to answer?  

Due to the following common complexities, it’s very likely that manufacturers are overlooking inefficiencies in their sourcing and overpaying large sums over time:

Multiple suppliers

Price fluctuations

Reliability and quality factors

Stock availability

Business guidelines

Eco and government compliance requirements

For supply chain professionals that are close to this process, they understand that it’s not a simple spreadsheet exercise to look for the lowest possible price for any given component.  The lowest price could have been an anomaly due to a supplier wanting to offload stock at that particular time.  8 months later that supplier could have increased the price, no longer hold the quantity needed, decreased in quality rating, or another supplier could now have a higher likelihood of a lower price.  Multiply this problem by the number of suppliers and components that you have and it becomes a large, complex data problem.  At the same time, the amount of savings and contribution to margin lift is also extremely significant when optimized across this scale.

Here are some different approaches on how you can optimize your supplier pricing breakdown:

Mathematical models

Mathematical models have been used to optimize supply chains for decades.  They are an improvement from eyeballing spreadsheets but they are also limited in that they can’t account for all of the variables very well and are static.  The models also need to be updated regularly and you may find the technique you were using is no longer relevant to your current business operations and you have to change models completely.

Machine Learning

With increased access to compute power, many organizations can now turn to machine learning to process large amounts of data and produce insights.  This approach is more adaptable and eliminates the need to manually select a specific type of model for your supplier sourcing.  However, machine learning engineers are a rare and expensive resource. Traditional machine learning techniques require tedious manual feature engineering before developing the models. Also, traditional techniques work best on small datasets and cannot handle all datasets from different product lines or different geographies together.  Each dataset has to be analyzed and modeled separately which doesn’t scale.

Deep Learning

The most advanced form of artificial intelligence, deep-reinforcement learning also takes advantage of increased access to compute power but takes it to another level by solving problems as humans would by reacting in real-time to dynamic changes in the environment.  The major benefits of this approach are:

constantly learning and continuously improving

does not require manual feature engineering so it scales gracefully

outperforms the most with massive datasets

proven to deliver the highest accuracy

This is the approach that DeepVu applies to solve for the supplier pricing breakdown problem in the most accurate way possible.  Because we have built the product to train from historical supply chain data and provide the optimal output, our customers do not have to invest in a data science or engineering team to achieve the end goal.  The reduction in costs from doing this optimally is significant for all manufacturers that have numerous suppliers and components.  See our website for more information.

[deepvuinc]

Why every brand will capitalize on deep-learning to optimize revenue & margins

By Moataz Rashad

Every brand has become an online retailer, some managing their own e-stores, and some relying on hosted solutions such as DemandWare. Similarly, every brand has invested in social presence, social listening, and social campaigns. In mobile, we’re definitely lagging behind the emerging markets, as many still have 2-star apps where a web experience has been shoe-horned onto a mobile screen, while an e-tailer like Flipkart in India has gone mobile-only.

Brands realize that they increasingly need to own the customer journey from intent to purchase, as they know the user experience (UX) is what inspires and retains users. A recent HBR article emphasizes how crucial it is for every brand to eliminate the guesswork in understanding customer’s emotional motivators. For online retailers, the study showed customers who are “Fully connected and satisfied and able to perceive brand differentiation” were 52% more valuable than the “highly satisfied” baseline users. The retailer study on the bottom line impact of the “Fashion Flourishers” category and its strategy impact on online-to-offline and store locations is well worth a careful look. However, most brands still face several key problems in understanding their customers’ motivators and optimizing their UX to get them to the fully connected stage.

Source: “The new science of customer emotions” by Magids, Zorfas, Leeman, Nov 2015         Harvard Business Review

Problems and decisions

For the vast majority of brands there isn’t much intelligence aggregation across these channels and touch points, nor is there a feedback loop to leverage the relevant analytics to inform the decisions various stake-holders are making every day. It’s decisions like these that need every byte and every pixel of intelligence you can gather: 

- Is this the right collection for Thanksgiving week sales? 

- Why is this recommendation converting against this SKU for San Francisco users, but not for Boston users?

- Is 20% lift in CTR good, or should we continue to refine to aim for 30% lift?

- Should we discount the handbags by 21% or 29% to meet X revenue goal in Q4? 

- Does that Twitter or Tumblr photo featuring our product carry positive sentiment or negative? 

Viewing excel sheets--or even business intelligence dashboards--that lack the continuous deep-learning behind the scenes on all these activities almost always puts you at a significant disadvantage. It gives you, the decision-maker, a partial, spotty view of your world. Also, because it’s already stale, you lack the most relevant real-time ammunition to make the smartest decision at any given moment. Compound this over time and across functions and you’re essentially flying almost blind, kind of like Wattny in The Martian during the scene where he punctures his astronaut suit and hopes to make it to the rope by free-flying through space. His endeavor was fun to watch, but it could have cost him his life and billions of dollars!

What is 360 degree AI-merchandising?

We call this deep-learning across all touch points to affect a smarter user experience, AI-merchandising. We’ll cover only 2 aspects of this in this article: product design and catalog design. Other elements will follow in subsequent articles.

Product design

A deep-learning engine that aggregates all clicks, purchases, save-to-wish-list, cart-abandonment, and returns for every product in your catalog gains internal insights (in the neural network) into what’s appealing and what’s not about your products. The information isn’t captured in some metadata attributes in a row or column in a database, it’s embedded in its “mind” and can be recalled. This intelligence can then be queried for suggestions on what to design. The engine can act as an AI-assistant to your designers, suggesting cuts, styles, patterns, colors, and fabrics for next seasons’s collection, which can even be refined by geography and demographic to meet revenue and margin goals.  

Catalog design

Cross-category recommendations, often called “Style it with”, or “Complete the look” recommendations, are a unique discovery mechanism used to inspire shoppers to create ensembles from your catalog. It is one way to encourage shoppers to maximize basket size, and to love your brand even more as they dress head-to-toe from your catalog. A deep-learning aiCommerce engine would garner the intelligence to advise you what’s working well with what for your shoppers (for ex: what handbags and shoes with what shirts, etc) so your catalog collections can indeed deliver that head-to-toe ensemble.

We’re very excited to collaborate with brands of all sizes on deep-learning and AI-merchandising to inspire shoppers with a futuristic and joyful user experience, while maximizing revenue.

[deepvuinc]

Maximizing eCommerce revenue with AI

For the past couple of years, many well-respected folks have opined on the future of retail, claiming that all brick and mortar retail is dead, dying, or in a vegetative state,and that the future of shopping therefore lies entirely online! See "future of shopping" and "retail stores will die"  We certainly wouldn’t mind that notion materializing at all, since for obvious reasons, the more commerce moves online--which is an irreversible trend--the better for us at DeepVu, as we’re building the AI layer for eCommerce.

However, I think that the hypothesis is flawed. The future of retail is AI-commerce, i.e. smarter shopping experiences in every sense of the word, whether it happens online, offline, or hybrid. “Smart” as opposed to “not so smart” commerce, as in what you often see now i.e : a) having to wait in line to buy a staple product for which there is little choice, minimal price variation, and there’s no emotional investment in the purchase b) being chased after you've left a store (offline or online) and abandoned an item to be shown the same exact item over and over (aka retargeting) or c) the seller keeps telling you those who bought this also bought that, as if all shoppers are automatons (aka collaborative filtering)

See this emarketer presentation for more detailed proof points on consumer data regarding how they view online vs. offline shopping.

It is indisputable that online retailers have much more attractive cost structure, and are advantaged by the digital consumer and product data that can drive decisions in real-time at nano-second speeds. However, offline retail has some advantages as well, including immediate delivery, try before you buy, impulsive buys, and lower return rates.

Some online-offline hybrid experiences are already happening. For an example of online companies expanding offline, just monitor the investment moves of Amazon, one of the most forward-thinking online retailers globally, and you’ll observe that they are serious about online-offline interplay. Similarly, Apple retail has a hybrid show-room/store model which is being copied by other consumer electronics companies. On the other hand, a forward-thinking brick and mortar retailer such as Nordstrom is experimenting with exciting online-offline hybrid experiences with the smart mirror for example. It will be intriguing to see the A/B test results of these efforts as they will pave the way for even more fascinating experiences.

Courtesy: Nordstrom-ebay fitting room

So what is smarter commerce, or AI commerce? AI commerce has one broad goal:matching the shopper with their desired product by factoring in their style, brand preference, price range, size, fit, and delivery time, in order to effect a frictionless speedy transaction.

If a brick and mortar retailer builds a sophisticated site/app but they are primarily interested in driving traffic to their physical stores, then they aren’t fostering an AI commerce experience. By the same token, if an online-only retailer gets user feedback in favor of same-day delivery and ignores this input, then they aren’t developing an AI commerce experience. So what would an AI commerce experience entail?

Let’s start with some definitions since AI is such a land-mine these days, what with everyone from famous loved entrepreneurs such as Elon Musk and Bill Gates to theoccasional scientific journalist speaking on the topic.

AI in this context refers to the field of Artificial Intelligence; software algorithms that use pattern recognition, computer vision, deep-learning neural networks, or whatever other techniques we practitioners come up with, to produce intelligent decisions that impact the shopper’s user experience. Such decisions include recommendations, merchandizing, dynamic pricing, promotions, placement of products on sites/apps/store-window, or whatever decision the marketing/product management executive needs the algorithm to assist with that day. In other words,we’re not talking about an artificially intelligent entity such as a robot, vehicle, drone, or anything else that can do any harm to people in the real world. And we’re certainly not talking about “Strong AI” which is where most of the fears of the critics revolve, where the algorithms would decided to learn from data outside their domain, and in fact will consume all accessible online knowledge, and for some reason they'll take special interest in learning about destroying humanity!

In AI-commerce, we're only discussing deep-learning software running in the cloud with no tentacles in the physical world, and where it’s trained on where to look, what to look at, and what type of output we'd want it to produce, and it's only domain is commerce.

This begs the question-- how is AI different form Business Intelligence (BI) which has been the traditional software layer of eCommerce? At the highest level, the fundamental difference is that BI is static or doesn't learn from the data, while AI's missions is to continuously improve and learn from the data to get "smarter". In other words, Business Intelligence is a misnomer, it's really business reporting, or business stats-- dry static stale reports on what's already happened a while ago.

So what do these AI algorithms work off of? the online shopping experiences have the upper hand here, because every action is already digital, cloud stored, and hence trackable and learnable in real-time. Every action including product impression, image click, recommendation click, purchase, like, save, share, cart-abandonment,email open, email link click, etc, is fed to the algorithms with the goal of making the shopper’s experience smarter and more pleasing.

We believe a true AI-commerce experience should accommodate most of the following:

1. Smarter visual recommendations: A smarter experience means that you get to see a recommendation that visually matches your style preference so much that it persuades you to purchase the recommended item rather than the item you originally searched for. We get super stoked when we count those transactions.

2. Smarter personalized recommendations: If someone is a loyal repeat user of an etailers site, then they do in fact expect a personalized experience that accounts for more than gender and age group, regardless of whether or not they use a social login. If they have purchased before, browsed, liked, and abandoned carts before, they expect the site to be smart enough to know their taste preferences and which products they would like to see. How is this different than (1) ? A user might be looking at a heather grey lambswool sweater, and you could show her ski goggles and cycling shoes, because you’ve learned that she’s into the sports associated with these items and haven’t purchased them in over 9 months. You might even show her girl’s soccer shoes because she bought a pair for her daughter last February. Or you might surprise her with an item that she’s never bought from your site before because it matches her style.

3. Smarter merchandizing: the more AI-powered decisions the retailer employs, the smarter the product mix in their catalog. As such, their buyers would know exactly what to stock each season based on transaction trend analytics, thus eliminating waste across the supply chain therefore saving time, and resources.

4. Psychic Personal Shopper Apps (PSA): A truly intelligent PSA wouldn't just have a warm voice and uses Google and Amazon to run searches for you and send you a list, it would know what you'd want and would just get it delivered to your door step. And if you dislike the product, you tap a button and it'll get it returned for you as well. It'll know that it's time to re-order socks, or that winter is coming 2 weeks earlier, so it'll order those thermal under-shirts. And it'll buy a different tasteful gift for your best friends' birthdays each year, and it'll ask you what the friend said about it so it gets better at gifting. Most importantly, it'll do all that algorithmically without any *crowd-sourcing* or asking you to fill out forms.

5. Smarter "show-rooms": Recall when show-rooming was considered a threat, well it isn't, it's part of the shopping experience now, and every online and offline retailer designs for it. However we think in the very near future, the physical shops will evolve into AI powered show-rooms, amazingly chic physical spaces that allow shoppers to touch, feel, try on, learn about the origin of the product, or quickly browse similar products to explore. But not necessarily wait in line or carry it home, unless in fact they do need it that exact moment.

6. Smarter payment: Some of the amazing international etailers we work with offer cash payment upon delivery. Some have added BitCoin. Some experimenting with pick up at partner physical stores. Retailers should certainly make it as frictionless as possible to pay for the item you’ve already invested the time and emotional energy to shop for and decided to purchase. So smarter payments means choice, speed, and convenience.

7. Social Spontaneous commerce: Many interesting developments there, but that’s for another article.

The history of commerce goes back thousands of years. It’s catering to a human need that’s existed since the oldest civilizations used the barter system. The future of commerce, however, is truly awesome; it's intelligent commerce powered by AI, easier, faster, cheaper, more relevant and more delightful. It’s already happening.

Are you excited about smarter commerce?--would love to see your thoughts below.

Moataz Rashad is founder & CEO of DeepVu. Follow him on twitter at: @moatazr

Credits: Thanks to Amit Kumar and Aya Rashad for reviewing an earlier draft of this article.