Bits & Bokstaver

Lettere tilfeldige innfall fra digitalt forlagsmenneske

[maritro]

7 reasons why software estimates don't work

First let’s start of by defining estimates. Estimates can mean a lot of things. In the widest sense, as a synonyme for forecasting or guessing about the (near) future or at all thinking and imagining what’s ahead, it refers to a practise we indulge in all the time, probably hundreds of times a day. We estimate when we have to leave home in order to make it to work in time. We estimate when we have to start a task in order to get done by a deadline. We estimate in a matter of seconds whether a task is large or small. This is not what I’m talking about here.

Estimates in this context are expert estimates of software development effort. It’s the detailed breaking down of tasks in order to reveal complexity and limit uncertainty. Oftentimes estimates are made by an entire group of experts, an entire team. In Scrum methodology this is considered important in order to include more perspectives on the task at hand, to reach consensus and to have the whole team on board.

Ok, now we’ve got the definition in place. So what’s wrong with this practise of expert estimation? I argue there are (at least) 7 things:

1. 

Estimates cost

The hours it takes to gather a whole Scrum team, dig deep into the tasks, uncover hidden complexity, check with subject matter experts/product owners and do at least a minimum amount of documentation - it amounts to a great deal of hours. In a team I was working with, i found they spent 80 hours estimating during a period of 3 months. That’s two weeks worth of potentially valuable time. And it’s a lot of money to spend on a practise that neither yields value nor is in any way accurate in predicting actual software development effort.

2. 

Estimates have short shelf life

In the worst of cases, teams makes estimates months ahead of the time they are actually going to put in the effort they are estimating. And in order to make “good” estimates, developers need to dig into the subject matter, uncover dependencies, unveil complexity. And make assumptions about solutions. A lot of good thinking and good conversations goes into the process. And, as the Agile Manifesto teaches us (because we are of course good, agile practitioners), we value runnings software over comprehensive documentation. So we try to keep the documentation of these good conversations as light as possible.

And the, weeks or even months later, we pick up the tasks again - and we have (almost) no clue as to what we were thinking back then. All the valuable insights are lost to us.

3. 

Estimates become irrelevant

Cause who haven’t experiences that half their initial backlog has been deemed irrelevant by the time you get halfway into the project. So all that time estimating is indeed worthless

4. 

Estimates inhibit creativity

Since you have to make choices about solution, those choices seem to stick with us whether we like it or not. This limits what creativity we get to use during the project, either because the choices we made actually the solutions we are committed to or because they are the solution we have imagined and we are anchored to them.

5.

Estimates are counter productive

“What gets measured, gets done”

...and we start doing what the estimates say against our better judgement, against what we learn as the project moves forward.

6.

We are unable to estimate

There are numerous papers and research on how lousy we are at estimating software effort. There are a number of cognitive biases that contribute to this, the most prominent being optimism bias, anchoring and confirmation bias.

7.

Estimating are killing motivation

No comment needed, right?

  So where does this leave us? How then do we plan, budget and control our software projects? The next blog post will provide the Answer.

Stay tuned!

[maritro]

New Deal on Data

I et intervju med HBR snakker Sandy Pentland om hvordan vi må finne en måte å balansere eierskap til (våre) data. Vi må ha mulighet til innsyn og til å dele eller trekke tilbake data selv. Noen av disse kravene ser ut til å bli bygget inn i EUs nye personvernforordning.

Så hvordan vil det i praksis se ut? Det vil ihvertfall kreve gode, brukervennlige løsninger for at alle som deler personopplysninger på Internett i dag kan få noen reell grad av kontroll, innsikt og eierskap til sine data.

"Alex “Sandy” Pentland, the Toshiba Professor of Media Arts and Sciences at MIT, suggests that companies don’t own the data, and that without rules defining who does, consumers will revolt, regulators will swoop down, and the internet of things will fail to reach its potential. To avoid this, Pentland has proposed a set of principles and practices to define the ownership of data and control its flow. He calls it the New Deal on Data."

Harvard Business Review

[maritro]

Vi ble lovet flyvende biler, men alt vi fikk var Internett!

...eller begynner vi å nærme oss tiden for flyvende biler - takket være Internett? Vi nærmer oss ihvertfall selvkjørende biler, kunstig intelligens og en hel rekke andre sci-fi-aktige artefakter. 

Hvorfor skjer dette nå? Hva er det som gjør det mulig? Dette er noen av spørsmålene McAfee og Brynjolsson søker å besvare i The Second Machine Age. 

Det er tre krefter som muliggjør og former den andre maskinalderen, ifølge McAfee og Brynjolsson: 

Moores lov; eksponensiell utvikling i lagringskapasitet, prosesseringskraft og overføringshastighet de siste 20 åra

Digitalisering: alt som kan bli digitalisert, vil bli det.

Kombinatorisk innovasjon; åpen innovasjon og crowdsourcing forsterker og aksellererer denne innovasjonen. 

[maritro]

Spotify Engineering Culture - Henrik Kniberg

Keynote at Smidig 2014

Henrik Knibergs foredrag om bedriftskultur og utviklingspraktiser i Spotify er vel verdt en titt. 

[maritro]

Utekino på Tøyen

[maritro]

The key to or the end of growth? Race with the machines

Er digitaliseringen slutten på Arbeid? Slutten på vekst og innovasjon? Eller er det starten på en annen måte å jobbe med Maskinen på? MIT Sloans Erik Brynjolfsson snakker entusiastisk på TED.

As machines take on more jobs, many find themselves out of work or with raises indefinitely postponed. Is this the end of growth? No, says Erik Brynjolfsson -- it’s simply the growing pains of a radically reorganized economy. A riveting case for why big innovations are ahead of us … if we think of computers as our teammates. Be sure to watch the opposing viewpoint from Robert Gordon.

[maritro]

Ta regi på egne data?

MIT MediaLab PhD har forsket på en løsning som lar deg ta kontroll på egne smarttelefon-data - OpenPDS:

"Researchers at the MIT Media Lab have created a personal data store system, called openPDS, that protects your personal data while still allowing applications to access the information they need."

http://www.fastcoexist.com/3033414/mit-wants-you-to-own-your-own-data-not-give-it-away