I’ve had this debate several times with other agile developers and project managers. I find myself staunchly on the side of not tracking actual effort toward tasks on an agile project.

I am assuming a few things when making this point.

1. The team in question is operating as a self-organized team.
2. As a self-organized team, the team itself is responsible for making sure its members are motivated and working efficiently.
3. The team’s members are trustworthy and there is no reason to believe that someone must be micromanaged to ensure they are not wasting their time or working toward something other than the team’s goal.

If the purpose of tracking actuals is to resolve mistrust of the team or a team member’s ability to manage his own work time, the problems with the team are fundamental and the discussion of whether or not to track actuals is a moot one. A self-organized team must be trusted until there is reason not to do so and as such, micromanagement by scrutinizing the effort an individual spends on a particular task is not a valid reason for tracking actual effort.

As such, one of the reasons I believe tracking actual effort on a project is detrimental to a team’s velocity is that it makes its members feel micromanaged. It’s a perfectly natural and well-founded reaction, because tracking actuals often turns into micromanagement. On a project where actual effort is being tracked there are often questions like, “why did x take 4 hours to complete, we only estimated it as a 1 hour task?”. A valid-enough question, but most people will become a little defensive when asked a question like this. That type of negative energy is hurtful to a team’s effectiveness. I always wanted to say “because it did, back off!”. Perhaps I’m just defensive but the implication is that you did something poorly, you either estimated improperly, were inefficient, didn’t ask enough questions, etc. Any one of these things could be true, but self-motivated people recognize when they make mistakes and pointing them out a second time almost always does more harm than good.

The most substantial reason that I think tracking actuals should be avoided is that it really doesn’t do any good. What are we trying to determine by tracking tasks at all? I would contend that the ultimate goal of tracking is to determine whether or not we are on target to meet our goals (so that we can adjust resources or expectations accordingly). To determine whether or not we are on track, all we really need to know is how much work is left relative to the time available. If we know how much work we thought we had when we started (task-level estimates) and how much work we have remaining (updated to-do estimates), we have all of the necessary information to determine our status. If we finish early, we can pick up more items from the backlog and adjust our velocity predictions for the next sprint. If we don’t finish, we can estimate the story points remaining and adjust our velocity expectations for the next sprint.

As far as the argument that tracking actuals helps us track the accuracy of our estimates, I would say:

1. No it doesn’t. Estimating is a kind of a voodoo science anyway. The more numbers we invent, the more we introduce opportunity for error. How do you determine how much time you actually spent on a task? Do you try to subtract all time spent going to the bathroom or checking email or having design discussions, etc. I’ve heard people suggest you spend roughly the same amount of time each day doing that miscellaneous stuff, so you should track which tasks you worked, the relative time that each took and divide your 8 hours accordingly. I know that’s not true for me, some days I am 90% efficient while others I am 40% efficient. I end up bloating actuals for some tasks and cheating time on others. Then what about the question of how actuals against tasks correspond to time on a timesheet? We end up losing accuracy because too many concerns are brought into play. Tracking actuals is far from a science. In fact, who pays that much attention to how much time they spend doing what each day? And if they do pay attention, how do they report the time they spend thinking about what they’ve been doing?
2. Who cares how accurate our estimates are, as long as they are consistent and accurate relative to each other. As long as we can accurately predict how much time something will take based on relative units, we have already reached our goal, who cares what the time value of each unit is? The fact that our “ideal days” end up being 3.2 days long is irrelevant, as long as they keep averaging to roughly 3.2 days each iteration. I’m in favor of looking at everything as points. Points haven’t let me down yet. It helps to have story points and task points, and we can think of task points as days or hours so we have a common reference point, but who cares how closely our task points resemble actual hours in the end?

There are circumstances where tracking actuals is unavoidable, if it’s mandated by a client or from above for accounting reasons, but I think it hurts more than it helps and should be avoided if possible.