Show all posts
3 years ago

Why Agile Estimates Don’t Work – Part 2

In Why Agile Estimates Don't Work - Part 1 I've explained why estimates don't work if someone sees them primarily as a commitment to timing. And, just as I expected, some aficionados rushed to educate me on the subject of estimates in agile, that they are not a commitment but, in short, a discussion of chances and odds of how the development will go, considering the challenges of this particular production environment. Probably, some of those aficionados have accused me of the gravest sin ever, and namely, not reading Mike Cohn's "Agile Estimating and Planning". Relax, guys. I studied Cohn's book long ago, and time after time I would flip its pages to refresh things in my memories, not to mention other books, articles and from-the-trenches stories. My most reliable source for making conclusions, however, is my work. If someone stays out-of-the trenches and theoretizes about estimates, this is just theory. My view on estimates lies in the practical, pragmatic context: if they don't work as commitment to timing, but as a discussion of chances and odds, why most companies continue to play this game? What makes them go on with it? Why spending lots of time on discussing chances is valued more than action itself?

What Is an Estimate? (take 2)

I've cited two options to answer this question in Part 1. Some people, who are, likely, not educated in agile theory, look at agile as a next best silver bullet to complete projects on time and they might wrongly view estimates as a promise of that. They genuinely believe that agile estimates will give them so much sought after reliable reference point about the time of completion. The second group of believers consciously accepts that estimating is a discussion of chances, a probability forecast. The burndown chart provides such forecast based on velocity. Let's refresh the classical definition of velocity in our memory, quoting from here: "The main idea behind velocity is to help teams estimate how much work they can complete in a given time period based on how quickly similar work was previously completed". Does it ring any bells now? If we never build the same feature twice, just as you can't step twice into the same river, then why velocity-based forecast should be relied on? In general, this stands true for all the forecast techniques based on past performance, including forecast models. Yes, there are cases when a team's work is monotonous, iteration in, iteration out, but from what I've been able to observe, it happens very rarely. Mostly, in any company and team, the tasks to be done and challenges to be resolved are unique, for each iteration, and for each release. You never know when something pops up and kicks this neat forecast in the butt.

The Devil Is In...

.. not only in the details. The second most common habitat of the said devil, which goes after the details, is human nature itself.  Nothing else explains this better than the good old Parkinson's Law:


Yes, indeed. Having all the time in the world is loose. It's either you have time, or you don't have it. It's either you have the guts and sixth sense to define what should be included to the minimal viable product, for instance, or not. Let's not forget that no one cares about software development for its own sake, except the software developers who view their work as craft. We do things for the market. For the customers, and they don't care about the development kitchen constraints, challenges and brilliant solutions. Same stands true for UX.

Now, how this reasoning fits into the subject of estimates, someone might ask? Here's the astounding truth. Teams and companies start playing around and messing with estimate rituals when they have some extra fat to burn. There's no room for activities that are waste in a bootstrapped, mission-oriented, do-or-die start-up squad of several people. If you are in such a team, and tempted to start a planning poker session, don't do this. Rather than waste your time on playing with probabilities, get some real work done. Write code, do a UI sketch,  instill clarity to the work of your team. Some mathematical forecast model surely has it that a brick will fell on your head one day. But you'd hardly be wasting your time to estimate how many more bottles of champagne are likely to slip out of a torn plastic bag, when one of those bottles has already hit the concrete, and there are 3 more in the bag. You'd rush to catch the rest of the bottles, not to let them slip, right? Or will you freeze and estimate the probability of all of the bottles being shattered? This reminds me of the fact, that some business people who are skeptical about shamanism, astrology and other such things, devotedly indulge into what is, in essence, shaman rituals with estimates. Come on, the estimate of completion based on burndown or a planning poker session, is as valid as an astrological forecast. There's no big difference. It's either you're "fat" enough as an organization to afford wasteful rituals or not. In fact, even in large companies that seem to be so safe and secure there's always the bottomline point of "do or die". That's what a recent story with massive job cut by Microsoft proves. Ritual is a waste. If there's time for rituals left, this is a sign of unhealthy fat. Burn it. If a workgroup discusses development, there's no need to wrap it in the ritual of estimating, because when a discussion turns into a draining debate of "how probable" this timeframe is, the work suffers. Someone said, there's a limited number of brains to do the job, and they should be used efficiently. One can suffice with a draft estimated timeframe, there's no use trying to gauge on the likelihood of this happening, when there's real work to be done.

Worship the Idol: How Do I Tell My Higher-Ups ..?

As life has it, however, most of us have to cope with the fallacy around estimates being employees in fat organizations, and, hard as you might, a mere human being can not move a mountain. There's no way to persuade a higher-up non-developer manager, or a client, or a stakeholder in the vanity of estimates. That's why people go on playing games, as they attend to those who expect a feature or a project to be done on time, as derived from estimate-related shamanic rituals. And, that's where another interesting booster for evolution is hiding. Luckily — and, yes, I mean it, luckily — there are more non-developers in the position of authority than developers. There's always a point of litmus test, when someone with a developer background (a project manager, team leader, or someone in middle management) meets the non-developer stakeholder. Why I call it a booster for evolution? If every stakeholder were a developer, they would have probably ended up whining on each other's shoulder about how difficult life is, and how impossible it is to commit to any timeframe. Having to deal with a non-developer stakeholder about a deadline is stimulating. If you've been thinking that something has changed from the hunter-gatherer times, I have bad news for you. The seeming "comfort" guises the basic instinct to act. You either act, or you rot. There's no other option. No one cares for reactive rants. It's your actions that define you. It's your choice to agree to play the estimate game by the rules and accept this as a given, or to quit and find a job where they will not f...k your brain with estimates. If you choose to deal with ruthless stakeholders that are oh-so-not-understanding of how hard a life of a true software craftsman is,  move the conversation from the level of rant to the level of action. Use every opportunity to spread the awareness of the challenges that software development portends, and why this domain is un-deadline-ifiable by nature. Amazingly, there are so many people in this world who sincerely believe that an estimate is a credible measure for completion date. Write articles, speak on conferences, join the "no estimates" movement. Fix the gap between what you know, and what they know. If everyone has their say, this world will become a better place, with less projects and software screwed. And, even if you'd still have to deal with the waste of estimates, you'd feel better inside, because you'd be doing your all to change things, instead of ranting.

Enough of thought boosters (or busters?). In Part 3 of the series I will give an outline of some techniques, commonly regarded as techniques for estimates, that might work as a tool for workgroup discussions in some teams. Keep in mind the waste-value balance, though.

Related articles:

Why Agile Estimates Don't Work - Part 1

  • Geoff Schardein

    I like the way this series is shaping up. Any estimating is a challenge in the least and is more art than science. Without knowing the risks and purely speculating on how effective you are going to be at an unknown task is a slippery slope to failure. One thing I do to attempt at more realistic estimates is ask the question, “Who is doing the work?” and the follow up question, “How much experience does the developer have in that area?” If the estimator is not the one doing the work and the expertise is over estimated then all bets are off.

  • indigo777

    Thanks for the feedback, Geoff. Unfortunately, I was not able to complete Part 3 of the series, because I had to resign from Targetprocess for a number of personal reasons. You can read this story here if you want:

  • David Holl

    Can you please make available some of the techniques you mentioned? On another site/blog or even row on email?

Try for free

Account url: *
How many people would be using Targetprocess?
  • Myself
  • 2–20
  • 21–100
  • 101–1000
  • 1000+
By clicking Continue you agree to our Terms of service and Privacy policy