Predictions
A wrote blog post over on Medium musing on Predictions.
Cosimo Galluzzi

Kaledo Art
styofa doing anything
h
art blog(derogatory)
Show & Tell
Game of Thrones Daily
KIROKAZE
"I'm Dorothy Gale from Kansas"
we're not kids anymore.
Aqua Utopia|海の底で記憶を紡ぐ

JVL

No title available

shark vs the universe

❣ Chile in a Photography ❣
Three Goblin Art

@theartofmadeline
Jules of Nature

No title available

JBB: An Artblog!

seen from United Kingdom
seen from Jordan

seen from Italy
seen from United States
seen from United States
seen from United States

seen from France

seen from Iraq
seen from United States

seen from Japan
seen from Iraq

seen from Türkiye

seen from Russia
seen from Venezuela
seen from Colombia

seen from Türkiye

seen from United States

seen from United States

seen from Jordan

seen from Colombia
@thegongshow
Predictions
A wrote blog post over on Medium musing on Predictions.
Quantopian and Steve Cohen
Quantopian has a big announcement today. The company will manage up to $250 million of investment capital, provided by Steve Cohen.
The investment capital will be allocated to members of Quantopian who create successful trading algorithms on the Quantopian platform. The algorithm authors own all their own IP and are paid a royalty if they decide they want to accept investment capital to power their algorithm.
The WSJ has more details. My favorite pull quote from the article is where they describe the backgrounds of successful algorithm authors:
[T]he creators of winning algorithms include a mechanical engineer with a Ph.D. in computational fluid dynamics in Sydney, a data scientist at an internet mapping company in Denver and a consultant in Malta with a master’s degree in mineral and energy economics
If you see yourself echoed in the description of these folks and have been curious about algorithmic trading, explore the community at Quantopian. You’ll feel right at home.
This is a huge milestone for Quantopian, and I feel privileged to have the opportunity to work with Fawce, Jean, and the rest of their incredible team on their journey.
Privacy Model for Notifications
We live in a weird new era where I nearly always have full control of what information I share and whom I share it with (assuming I have an indefatigable interest in navigating permissions settings for my various social services), but I have no control over my information once it leaves me. The consumption of my social content is entirely controlled by my followers, not me. This control model is simultaneously intuitive, correct, and disconcerting.
A classic example that comes up frequently for me is location. I’m perfectly fine with sharing my location with my friends through foursquare/swarm. I update Swarm multiple times per day and derive a lot of value from doing so. But I think it’s odd that, for people who have updates from me set to always notify them, some folks are constantly being reminded of my location, buzzing away in their pocket. This problem is not isolated to Swarm, it’s true for Facebook, Instagram, or any other means of sharing current location. I have no problem sharing this information, but I have a little problem with how “in your face” it can be under the most aggressive notification settings. It’s a vanity issue... I think my location is unimportant to the point that I’d rather it not be in my friends faces multiple times per day.
One could argue a few counterpoints:
(1) Why should I care how other people consume the content I create? Their mode of consumption is their choice.
(2) Just check-in “off the grid” more and share less.
(3) Just unfriend (most) people.
For me, (2) doesn’t feel like an option. I’m happy to share and when I get helpful comments about my location, it’s terrific.
If (3) is the answer, then the product is broken, not my usage. So, lets give well-intentioned designers the benefit of the doubt and toss out (3).
So, I think the whole meat of the issue is (1). I don’t want my friends to be able to control my notification settings, and yet, I wish I could control their settings when it comes to my content. My desire is obviously inherently contradictory and why I find it interesting enough to blog about. There is a difference between (A) sharing information such that others have opt-in access to it and (B) broadcasting information aggressively. Designing these settings (and their defaults) is really tricky, and I wish I could be a fly on the wall in these design meetings.
Bonus Corollary: Sometimes you send an email and instantly regret it. You want to edit or delete it, but you can’t. Your content is your own until you share it, and then suddenly it’s the recipient’s content too, as a received message and intuitive privacy controls implies that a recipient should control their content. Yet everyone has felt this exact email pain point, and so when Slack decided to design for this use case, they allow the author to both edit (with an “edited” sign next to the message) and also delete any message the author writes. Slack decided that tie goes to the runner and email decided that tie goes to the fielder. The email design decision is less intentional and more an artifact of the information architecture of how email works in a stateless distributed early-internet design (how can you take an email off a remote server you don’t own?), but it’s still an interesting design choice nonetheless.
This week Gizmodo wrote an in-depth story about the bias in Facebook’s trending news product. The first paragraph from G…
I published another longer read over on Medium. Syndicating here for Tumblr followers and email subscribers.
Related lazy-web request: does anyone know a good way to mix together the RSS feeds of Tumblr and Medium so that I don’t have to do these cross posts for my Feedburner email subscribers?
I wrote a post over on Medium. It will be my new full time place for long form writing. I'll probably continue to post more Tumblr-ish content here. Go follow me on Medium.
The Darwinism of Encryption (Or... Why It Doesn’t Matter If You Side With Apple or The FBI)
I’ve restrained my commentary on the Apple/FBI encryption debate to tweets so far, but I couldn’t find a way to say this in 140 characters, so blog post it is.
Digital communication is running a multi-decade inevitable march towards end-to-end encryption. In the beginning when the first ever TCP/IP packets were scooting around the ARPANET, all communication happened in the clear, unencrypted. There were no bad actors on the network; it was just a bunch of altruistic geeks freely routing forward each others’ packets (both academics and military, but all geeks nonetheless) cooperating together. After initial attacks against the network, it quickly became clear that one could not assume the man in the middle of your network path was your friend, and encryption started to emerge as a second layer of abstraction on top of TCP/IP (the military started first).
Fast forward to today, and much of the basic communication of average internet citizens are fully encrypted, without any knowledge or effort on their part. The Google search at the start of an average internet user’s session is encrypted by default, and Google now uses HTTPS encryption as a ranking signal; meaning they’re deliberately trying to send users to encrypted destinations, rewarding encrypted websites with more traffic than before.
This is not a pendulum that will swing backwards towards less encryption in the future. This is a forward march, in lock step with technical progress itself. And it’s a forward march taken at each step out of necessity. Continued innovation by hackers with increasingly sophisticated attacks is the fuel that feeds the progress towards more reliably secure communication.
In this light, the Apple/FBI debate is mostly moot. You could side with Apple and support their fight for privacy. Or you could believe the FBI’s intentions are noble and Apple is being obstinate for the sake of impractical ideals. Either way, we all will move towards more secure communication and the FBI’s demands of Apple are the latest spitfire of fuel into the combustion chamber of progress. Whether Apple does or does not comply, the public nature of this case has shown that the privacy of data we presume to be locked down properly in our devices is at risk. It’s secure enough to create a public debate between Apple and the FBI, but it’s clearly not secure against Apple itself. Apple is concerned about setting a “dangerous precedent” by complying with the FBI’s court order, but by virtue of Apple’s response alone, it’s now public knowledge that Apple can defeat its own security (regardless of whether they comply), and this knowledge will generate open source encryption innovation that will (on the longest time horizon) make all our digital communication more secure.
There is an analogy here in the history of online music piracy. Umair Haque has the perfect quote:
Every time the music industry kills an underground distribution channel, a more efficient one arises in its place. Goodbye mixtapes, hello www. Bye www, hello Napster. Bye Napster, hi BitTorrent. Bye BitTorrent, hi anonymous, ciphered, totally decentralized p2p nets.
Like the darwinism in file sharing, there is a darwinism in encryption technology. The FBI’s request could successfully generate the master access to Apple’s encryption they seek, but it will ultimately just generate an even more secure form of encryption that will be open, distributed, and not depend of the benovalence of a corporation. The government could take counter-actions to ensure that America doesn’t led the security tech march forward (by mandating government access to backdoors) or deem encryption of private citizens’ data illegal. But that won’t stop bad actors from using encryption; bad actors are already breaking the law, so what do they care about breaking the law twice? Encryption is here. It’s not a policy decision or a stylistic choice that will shift back over time; it’s math. We cannot un-invent it, and it’s only getting stronger.
Podcast with Nick Moran
Nick Moran keeps a great podcast about VC and startups called Full Ratchet (the name is in reference to a particularly thorny term in venture deals... I hope you never need to face it). He interviewed me a few weeks back, and Part I just went live today. Check it out. Part II coming tomorrow.
This chart from fivethirtyeight shows a histogram of movie reviews from 5 different sources. The reason they made this chart is to show that aggregated reviews on Fandango are skewed too high and thus untrustworthy (which is an appropriate conclusion).
But I find this chart interesting for a different reason. The disproportionate 3.5 star reviews from IMDB and Metacritic caught my attention.
My partner Mo once asked my opinion of something by saying, "What's your rating from 1 - 10 in a world of no sevens." I didn't quite follow at first, but then he explained that a rating of "seven" in general is like a non-answer. It's safely positively neutral in a way that contains very little signal. If you can't say "seven," but you think seven is what you would say, you're forced off the fence into either a "six" or an "eight." It's a clever trick make ratings contain more conviction.
If I ran these movie review sites, I'd apply Mo's system and ban 3.5 stars as a possible response, which would put a massive divot in place of the peaks in these two histograms.
Why Mobile-Optimized Works (pt 2)
Yesterday I wrote a post showing how companies’ mobile-optimized websites are generally better than their desktop websites when viewed from a desktop browser. It’s a somewhat dramatic conclusion to make given that companies usually have a comparatively rag-tag team focused on mobile-optimized design (all the mobile efforts typically get aimed towards App design instead of mobile-optimized design) in comparison to the richer and more established desktop design efforts. Why would the output of an afterthought team be better for a desktop than the hard-earned, constantly-A-B-tested, decade-long-work-of-love from the desktop web design team? I think there’s a few reasons:
Constraints are good. When you’re designing for less screen real-estate, you are forced to make tough choices. For example, you can’t decorate the right rail of a webpage with 20 junk-drawer menu links on mobile because there isn’t enough screen real estate for a right rail at all. When you’re forced to design for 25% of the space, you are forced to decide what *really* matters.
The mobile mindset is get in and get out. It’s not assumed that a mobile web session is going to be the start of a 30-minute rabbit hole of discovery. Instead, people landing on a mobile-optimized site are likely coming from Google and just need what’s on the permalink being called. A good mobile-optimized site serves this relevant information and very little else.
Less ads. You can’t put 15 different banners interlaced in a poorly executed game of tetris above the fold when you’re designing for mobile because no one will ever trust your domain when they’re browsing from a mobile phone again. Less screen real estate corresponds to less intrusive ads.
Whitespace. When you take a mobile-optimized site and you view it on a desktop, the various elements of the page scale up proportionally, including the whitespace. Whitespace is an absolutely essential design tool; the more of it you use, the more you can focus your user’s attention on what matters. If you want a button to grab your user’s attention, surrounding it with ample whitespace is your best tool. So, not only are their less competing design elements on the page in a mobile-optimized site, but the ones that did make the cut get a healthy dose of whitespace to amplify them when rendered on a desktop.
That’s my hit list. Add what I missed in the comments.
Mobile Optimized is the New Ideal Desktop Browsing Experience
When cruising through my Twitter feed on my desktop, I click on links that sometimes drop me on mobile-optimized pages. They are so much better designed than their desktop counterparts. These clicks inspired me to spend 5 minutes exploring the design contrasts at some of the most common sites I use.
Wikipedia: Here’s the normal Wikipedia experience in my browser...
... and here’s the equivalent mobile-optimized Wikipedia entry as viewed from my desktop browser.
NYTimes: Here’s the normal NYTimes experience in my browser...
... and here’s the equivalent mobile-optimized NYTimes story as viewed from my desktop browser.
Amazon: Here’s the normal Amazon experience in my browser...
... and here’s the equivalent mobile-optimized Amazon product page as viewed from my desktop browser.
For this last amazon example, I couldn’t simply rewrite my url to visit m.amazon.com (which was my methodology for the first two examples). Amazon’s server-side logic was reading my user-agent and force-feeding me their crappy desktop experience. I had to send them a false iPhone 6 user-agent via Chrome’s Developer Tools in order to get a screenshot of what the mobile optimized experience looks like on a desktop browser.
The takeaway: All of these sites would be better served by simply throwing away their desktop optimized CSS and just serving the mobile-optimized experience to everyone. The mobile-optimized sites are simply better designed. Simpler, cleaner, better use of white space to focus attention. Unimportant options/links are all eliminated. Maybe I should just leave my user-agent as a mobile device permanently on my desktop.
Creativity Today
Steven Johnson has an excellent long read coming up this weekend in the NYT Magazine. It’s available online now. Its title says exactly what it’s about: The Creative Apocalypse That Wasn’t. In short, Napster was supposed to be a harbinger of doom for all creative talent as piracy and digital boogeymen were going to mean the end of any viable revenue streams in the creative industries. Steven explores digital implications for music, movies, books, and television in a pre-Internet context compared to today. It’s just great; I’ll refrain from stealing his thunder on conclusions (if the title didn’t already tell you everything).
Reading Steven’s piece, at the very end he had a tiny throwaway line about how two musicians separated by the Atlantic can collaborate on a work today, but he never elaborates on this half-sentence at all. I was instantly reminded of a recent episode of Song Exploder (a podcast that interviews the creators of popular songs about their creative process) that featured the musician behind the title score for Game of Thrones, Ramin Djawadi. I was really struck by one small section of the podcast (at time mark 5:45 in the stream, if you want to hear Ramin describe it himself). There are female voices singing in the score, a chorus of 20 women. These 20 women are in Prague. Ramin never travels to Prague to work with them. Instead, he directly them remotely from his apartment in LA. They collaborate live, where Ramin can hear the output of their singing in his home in excellent quality, and then he talks into a microphone that outputs into the sound booth in Prague to provide critical feedback, and then the chorus iterates on another take of the verse live. It’s just like the classic studio scene you saw in every episode of Behind the Music on VH1 in the 90s, except instead of 3 inches of glass separating the producer and the talent, it’s now 6000 miles of oceans and continents. I’m sure this is quite normal to people in the music industry today, but my mind was blown.
I love how music is nearly perfectly abstracted into information in such a way that geography is no longer a barrier for creative people to work with each other. It’s like we invented teleportation, just in a slightly lower fidelity than real life.
Steven’s one-liner alludes to the fact that this is possible, but it’s Ramin’s description that drove home what’s actually happening. Both works (Steven’s and Ramin’s) are a window on what it’s like to be a creative today. Creative technology is changing at an accelerating rate, so it will look even more different in 5 years than it did in the 15 year juxtaposition that Steven offers up in his piece. I can’t wait to see what happens.
Thoughts On Digital Healthcare
Jonathan Libov of USV and Angela Tran Kingyens of Version One Ventures are collaborating on a mini-series of long reads that they are calling On Digital Healthcare. They released Part 1 today (about Mobile Endpoints) and you can subscribe to updates via email as they are released on their site. It’s terrific, and I’m delighted they’re sharing their thoughts about this meaty and tangled subject with the world.
I just finished reading the first installment and found myself eager to leave a comment making a few points. But, as I started to type, my increasing disillusion with commenting took over, and I decided to migrate my thoughts to my own blog. So, If you’re interested in what I have to say here, I’d recommend pausing to go read On Digital Healthcare, and then coming back.
...
Welcome back :). I was struck by a few thoughts as I digested this post.
1. I approached the authors’ post with a core assumption: Doctors’ services are not a commodity. Some doctors are much better than others, and when I go to see a doctor, I care about the quality and reputation of that doctor. Quality in any service (not just healthcare) in general sits on a spectrum from complete commodity (all services are 100% fungible) to complete specialization (all services can be uniquely ranked relative to each other). When I take a taxi, my experience is much closer to the commodity end of the spectrum (it’s not exactly at the end of the spectrum, and that’s why Uber’s star rating system is an advantage over the status quo), but it’s still mostly a commodity to efficiently get me from point A to point B, literally, in the world. When I visit a doctor, I also want to get from point A (I have a problem) to point B (my problem is solved; I have a positive clinical outcome), but the quality of the experience and expertise needed to solve my problem is way on the opposite side of the spectrum from being a commodity. That’s why consumer news media, such as NY Mag, annually ranks best doctors across all possible specializations. Because we (US consumers) care about quality and don’t think doctors are commodities. Note that NY Mag does not run the same list ranking taxi drivers.
By contrast, most of the mobile endpoints referred to in the telemedicine section of the authors’ post are designed to connect you with *a* doctor. There is some differentiation around specialization, and many of the telemedicine services has a similar, Uber-like 5-star rating system. But that implementation, which is sufficient for differentiating the quality of recovering taxi drivers, feels woefully inadequate for measuring the various quality of doctors. Based on the implementation of most of these telemedicine web services, assumptions of quality by the patient are to be ascribed to the brand of the telemedicine service as opposed to the doctor. Meaning, a patient will learn to like Doctor on Demand better than First Opinion (or vice versa) because the quality of the experience is better or not.
I think great telemedicine is going to embrace US patients’ obsession with doctor quality and find some more suitable proxy to help patients connect with the best doctors. Anything short of this solution (specifically, any solution that treats doctors more like commodities, fungible to each other) will be perceived as a cheap, low quality solution that simply is just a calculated trade-off in the convenience of seeing a doctor now virtually VS waiting a week (or a month) for a PCP appointment IRL. This calculated trade-off doesn’t feel like *the future*... it feels like we would be aiming far too low.
2. Speaking of PCPs... assuming people value continuity of care and familiarity with their provider (which I feel is a safe assumption, but an assumption nonetheless), the most important shift in telemedicine will be getting existing PCPs to provide telemedicine to their existing patient base. This is not going to happen in one fell swoop. Instead, it will be a transition through increasingly higher fidelity experiences. The bare minimum in telemedicine, which many doctors do not provide today, is the ability to email you doctor. Why can’t you email your doctor? Because most payors won’t cover email consultations, so doctors have no compensation recourse for the advice and care they provide through that channel.
I don’t mean to imply that all doctors are ruthlessly financially motivated, in fact, I believe many doctors entering the profession today are mostly not financially motivated in their behavior. They could make a lot more money doing something that requires a lot less training and student loans. Doctors from older cohorts are a bit more financially motivated, but I still their their primary driver above all else is helping people, and the fact that being a doctor up through the 80s was financially lucrative was a nice secondary benefit.
But, if you follow the money, PCPs have active friction preventing them from providing telemedicine to their existing patients. UnitedHealthcare’s coverage of telemedicine visits is a big deal, but I’m not convinced jumping straight to a higher fidelity experience (a video call) is the right on-ramp for most PCPs. Instead, lets start with billing codes that cover emails and text chats... these low fidelity experiences will be a nice on-ramp that will make video visits more palatable for both the doctor and the patient.
3. The telemedicine survey of the market in the authors’ post covers experiences where a patient can have a video call with a physician. A video call is a level of fidelity that sits on a spectrum. On one end of the spectrum, the most primitive technical end, is face-to-face conversation IRL (ie zero tech). The next stop is lower fidelity digital interactions, first email, then text message, then voice-only, then video calling (all relatively low tech by 2015 standards). I think telemedicine only starts getting interesting when you think further out on this spectrum past video calls. For example, the Da Vinci robot, increasingly used for minimally invasive surgery in place of laparoscopy, is designed to be controlled remotely by a surgeon that could be halfway around the world. 99.9% of the time it is used, the surgeon controlling the robot is in the operating room with the patient, sitting at a console that handles the I/O. But, increasingly I think you will see surgery specialists that will become Da Vinci experts that will provide their services globally, and they will be assisted by a local surgeon on-site (lacking specialization) that can handle any crises that require aborting the robot surgery and potentially opening the patient as necessary.*
I feel most inspired by telemedicine when I think about it on this spectrum... because Skyping with my doctor feels perhaps more convenient, but mostly a poor substitute for reality. Whereas, the Da Vinci example I walked through provides better patient care than I would otherwise be able to receive, unless I happened to be lucky enough to live in a major academic medical center city or I’m willing to displace all else in my life and travel to such a center for specialized, expert medical care. It’s a case where technology doesn’t just reduce cost or increase convenience; it’s just better.
4. Lastly, a comment about publishing. The website hosting this mini-series looks really nice and clean. I checked out the page source code to see how it was built, and it’s not obvious to me there’s a CMS involved (no obvious WordPress telltales, but I’m not totally savvy on how to spot it). My gut says its hand-rolled, which, if so, is very cool.
In closing great job Jonathan and Angela! Can’t wait to read the rest of this series.
---
* Much like how Google trains self-driving cars with a training set of hundreds of thousands of hours of human driving recordings mapped onto digital sensory input (lidar, angle of the wheel, GPS, localization, etc...) of driving situations, I have to imagine that Intuitive Surgical (makers of the Da Vinci) see all the data they are collecting during robot operations as a similar training set that will someday (long time horizon) allow the robot to operate on humans autonomously. The FUTURE. (!?) ... I would link to the Prometheus c-section robot surgery scene here, but I’d hate for someone squeamish to wander in that direction... Google it if you’re curious.
Panorama Education
Nearly all successful modern companies employ some variation of a build-measure-learn feedback cycle. It’s a cycle that iterates as follows: you take an initial position on what your product should be and build it. Then, you measure your target audience’s response and interaction with the product. Then, you analyze your data measurements to figure out where your initial hypothesis was right or wrong and use the learnings from this analysis to inform that next iteration of the product. The cycle is now complete and begins anew.
Because this cycle is so fundamental to offering a great experience to customers, all companies purchase analytics of one flavor or another. It’s an essential component to building anything. Lacking great analytics, a company is flying blind.
Education is no different. In the market of providing education to students, schools need to engage in a build-measure-learn feedback cycle. This is not a new idea, and there has been a need for assessment of education quality for as long as we have been teaching students.
Panorama Education, a Boston-based company, addresses this market need with a modern web approach that brings the best software advantages to help schools measure and improve the education they offer. They offer a product to schools to survey students, parents, teachers, and staff and then analyze that data to help improve student outcomes.
I’m excited about Panorama Education for a few reasons:
1) Schools give Panorama glowing reviews. When you talk to the data departments of public and charter school districts, they see Panorama as a pinnacle example a great partner.
2) Existing incumbents in the education survey market take a sorely antiquated approach. Some incumbents make money by licensing proprietary survey designs to schools at absorbent prices. Because surveys are most effective when they are run longitudinally, these proprietary surveys create lock-in that needlessly wastes districts’ money as the licenses have to be renewed annually. By contrast, Panorama has collaborated with Harvard University’s Graduate School of Education to build a best-of-breed survey design optimized to capture the best possible data across a wide variety of themes and then open sourced the design. The survey is available for anyone to use for free, and then using Panorama’s survey platform with these surveys will offer a school the best possible product experience. It’s a web-native approach to this market that I love because it means more data will be collected using Panorama’s methodology (even if the schools are not Panorama customers today), and it changes the basis of competition to competing on the quality of the software platform for data collection and analysis (where Panorama shines).
3) The Panorama founders (Aaron Feuer and Xan Tanner) and the team they’ve built are a very impressive group with a mission-driven obsession to build the best possible platform for schools to measure and improve their education. They see analytics in combination with a library of educational best practices as the highest point of leverage to improve improve student outcomes, and they have the commercial and sales instincts to build a very strong business at the same time.
4) The customer traction is remarkable. Panorama has been used by 3 million students across 6500 schools to date. The sales momentum is a testament to the great value proposition they offer schools.
I’m excited to announce Spark Capital’s latest investment today, Panorama Education. Owl Ventures and Spark Capital co-led a $12MM round into the company. I’m delighted to have the opportunity to partner with Aaron and Xan as they continue their mission.
Permutation City
About 6 months ago I read Permutation City by Greg Egan. When I put it down, I was glad to be done with it. It wasn’t a gripping read, so despite being relatively moderate length (350 small trade paperback pages), it took me awhile to muscle through it.
I don’t usually write reviews on this blog about books I didn’t enjoy much (I don’t see the value in saying negative things publicly about others’ work), so I didn’t blog about it. But this book has really stuck in my mind, and I find myself continuing to think about it today.
Permutation City is of the “hard science” school of science fiction. It pulls in meaty ideas from a variety of different schools of science and philosophy... stuff like: the concept of emergence, synthetic biology, information theory, digital representations of consciousness, exploration of computational limits, and the ethical rights of digital entities.
Looking back after 6 months, I think that Greg Egan is a highly underrated sci-fi author if you consider the primary purpose of sci-fi to be speculation about the extrapolation of modern science trends and its limits. But, if you’re looking for a quick paced book with engaging characters, Greg Egan isn’t the first author I’d pick up. If another 6 month passes and I’m still thinking about Permutation City in the same light, I’ll definitely read another of Egan’s books.
Series A Investment Pace Across Select VC Firms
The blog post I wrote earlier this week talking about Sequoia’s investment pacing back around the nuclear winter got far more distribution than I initially expected. It made me curious to look at some more recent numbers about investment pacing. Here is my analysis and conclusions.
Investment pace is an interesting quality metric to use to evaluate a venture firm because it’s only valuable if you subscribe to the school of thought that you cannot time markets. If you believe you’re smart enough to time markets, then it would be very sensible to have highly uneven investment pacing. You should do a lot of deals when you think the timing is right, and you should do relatively few deals when you’re bearish about the timing. I personally think timing markets is impossible; no one has a crystal ball to know the future with any certainty. So, lacking a crystal ball, the correct corresponding behavior should be to invest consistently over time. That was what impressed me so much about Sequoia in my last blog post, their ability to invest so consistently across good times and bad times is admirable.
So, as a follow-up, I decided to analyze the investment pacing of a handful of VC firms in our current era of investing.
Methodology: I exported all Series A deals over the last 14 quarters from Mattermark. I then graphed the quarterly moving average number of Series A deals done for each VC firm (with a 4 quarter moving window) over time. I normalized each line to make zero on the y-axis equal to the firm’s average over all 14 quarters combined. I used a quarterly moving average instead of raw quarterly count of deals because quarterly numbers without the moving average were too spiky (the moving average smoothed the lines).
Results:
[bigger resolution version]
Here’s a helpful tip for reading this graph: If a hypothetical VC firm had *perfect* investment pacing (meaning all their deals were evenly distributed across all quarters in sample), then the line on this chart would run exactly along the X-axis at zero the whole way across. The further any line deviates away from the X-axis, the farther off of average pace they are.
So, based on this 14 quarter dataset, surprisingly, Sequoia has the least consistent investment pace. It’s easy to see the purple Sequoia line swinging both low and high around the mean. This is totally not what I expected when I started this exercise given how impressed I was with Sequoia’s investment pacing in my last post about “RIP Good Times.”
The way to use this chart to determine exactly which firm has the most consistent investment pace is to take each point on this line and measure it’s distance from the x-axis. Meaning, take the absolute value of each data point and add them up into firm totals. By this analysis the big winner is Benchmark (Benchmark’s absolute value total score is a low of 4.9, but that “4.9″ number is rather arbitrary, so just focus on that fact that they’re the lowest). Andreessen Horowitz comes in second place (with 5.8). And, as expected from just eyeballing the chart, Sequoia is in last place with a total absolute value distance from the mean of 22.3.
I would be happy to share my spreadsheet of analysis, but it contains data that I downloaded from Mattermark under their license terms, which means I’m probably not permitted to share it. If they give me the OK to share it, I’ll update this post with a link to my spreadsheet.
I hope you found this analysis interesting. There are so many corollary studies that could be run here, but this was just a fun friday exercise, so I don’t plan to do them myself. Follow up analysis (perhaps to be done by CB Insights, Pitchbook or Mattermark) could look at the following different deviations:
- Stages other than Series A
- Only look at rounds led (instead of both led and participated, which is what I did)
- Look at a bigger time horizon (I was limited to 5000 export rows by Mattermark unfortunately, hence I stopped after 14 quarters)
- Vary the moving average window size
- Use root mean square deviation as the distance error measurement instead of my crude raw distance from the mean approach. (I’m not enough of a statistician to know which is more sound for this use case.)
Or add your own variation in the comments.
Sequoia’s RIP Good Times and Their Subsequent Investments
Brian Chesky, CEO and Co-Founder at AirBnB, recently published 7 rejection letters he received from valley VCs in an effort to raise $150k for a 10% stake in AirBnB during mid-to-late 2008. My own view into that era was already published in an email thread between Paul Graham and Fred Wilson that shows how easy it is to miss deals at the earliest stages (I was in the USV pitch mtg with Fred for AirBnb that is the subject of that thread).
Brian’s rejection letters are entertaining alone, but they beg a more interesting question which is: who was smart enough to say yes to this attractive proposition? The answer is Sequoia (though not at the exact same time as these rejection letters... it took a few more months to get to “yes”). Sequoia was the first angel in AirBnB, leading a $600k angel round in early 2009, and this is noteworthy IMHO because of the timing.
On 10/10/2008, Sequoia presented to all their portfolio company management teams a deck titled “RIP Good Times”, to tell them all that, like the villain in a cheesy horror movie, the reaper of 2001 was back for vengeance. The sky was falling. It was time to make deep cuts to your team, conserve cash, and just try to ride out the storm with the assumption it would be virtually impossible to raise capital over the next few years. And then, a couple months later, Sequoia made their AirBnB angel investment.
I found this timing interesting enough that I took to CrunchBase to see what else they were doing at this time. I looked at all deals done by Sequoia between 10/1/2008 - 12/31/2009 that were either Seed or Series A (meaning other deals around the same time as AirBnB at a similar stage). It turns out there are 31 such deals. In addition to AirBnB there was GreenDot (which IPO’d), Dropbox, YCombinator itself (which was so clever both at the time and in hindsight), and OpenDNS (a recent impressive exit).
In order to set that investment pace and taste in context, I looked at Sequoia’s Seed and Series A behavior in another 15 month time period, 3/1/2007 - 5/31/2008. There, Sequoia made 33 such investments, and most notable of which was the Series A in MobileIron (a ‘14 IPO).
33 deals in good times VS 31 deals in RIP Good Times: remarkably consistent. And their investment taste (or luck?) was even better during “RIP Good Times”.
The consistency of investment pace is a wonderful lesson for all environments (good times or not). The taste piece I do not attribute to an artifact of the 2008 nuclear winter. Instead, I think it’s just a good lens applied consistently (and also benefitting from the proprietary relationship Sequoia built with YC during that time).
The other lesson to take away here is there is no such thing as a bad time to start a company. The companies founded in late 2008 (when it was incredibly difficult to raise capital) are some of today’s biggest successes. So, it doesn’t matter when you start; if you want to build a company, start today.
Seveneves Review
Yesterday I finished Neal Stephenson’s Seveneves. It is a great read, well suited for summer.
The plot is simple enough to explain spoiler-free: The moon blows up for an unexplained reason. What happens next?
It’s a long read, but that’s because it’s actually two books in one. It’s 870 pages. The first two thirds are one continuous story, and the last third is a separate story that takes the result of the first story and runs wild exploring the longer term implications speculatively.
The first story is The Martian on (a)steroids. Meaning: it’s a hard science fiction book where detail-oriented scientific solutions drive the plot forward, so if you’re uninterested in things like physics, orbital mechanics, material science, etc... then your attention span will be challenged. If these topics are your cup of tea, then this first story will hum along quickly.
The second story is Neal Stephenson’s first crack at what I would consider a more classic sci-fi theme... his take on a space opera. It took me longer to read this part because it’s so imaginative... you need to really absorb the details to get the same mental picture that the author has in his head.
Like all Stephenson books, this is a book of ideas, and the plot is there mainly to explore these different ideas. Some of them are political. Some of them are an exploration of social behavior. Some of them are physics thought experiments. You can feel an “idea section” coming on as you’re reading because the characters start to puppet the socratic method, batting the idea back and forth. If you indulge in these ideas and spin them further in your own mind, then you’ll have a lot of fun with this book. If you’re just here for the story, then you’ll probably walk away from this book thinking that it was roughly 300-400 pages too long.
Most science fiction written today is dystopian and makes technological progress into a bogeyman to be feared. Seveneves by contrast shows how science can be used for good to overcome catastrophic adversity. This is the single best aspect of the book.
I gave it a 4/5 stars on my Goodreads profile, but only because I was thinking about it in the context of all Neal Stephenson books. I think I would have given it a 5 if it was from a different author.