Not Too Much Measurement Detail

Overgathering Of Data: If a little status information is good, then a lot of status information must be wonderful, right? Please tell me how you spent each twelve minute interval of your work day over the past week. It sounds wonderful to some Professional Project Managers, but seems extremely onerous to staff. (I remember one meeting where someone said extra information would be needed for each bug report, but it would somehow reduce the effort needed to deal with bug reports. This came from someone who read bug reports, not someone who filled them out. My response went beyond impolite; I yelled and cursed and used phrases like, "Don't lie to me." It was unprofessional, but afterwards, people congratulated me. Did I mention I don't work on that project anymore?) --PaulChisholm (moved here from ProjectManagement)


I co-manage a very serious but not huge non-OO project with a non-IT person who has not done PM before. Our sponsors got (quite reasonably) very upset when they saw our schedule and work estimate, because it was nine times larger than the previous estimate (see MultiplyByNine). It was not that they didn't believe it, just sticker shock. One executive had the most beautiful sentence: "This document seems a little thin to me. I would expect more pages explaining just why the new estimate is so far from the old estimate." Kind of shut down the meeting.

So in the last month, while we kept working anyway, the managers have been really sticky about double checking every number in the plan. A real nightmare for the developers. We had a chance to enter the DilbertMoment. But after I checked with them twice each again, I stopped asking them and started asking my partner why her boss cared about this particular question. She was of course going crazy herself. It seemed every time she had added columns to the spreadsheet, her boss asked her a new question she couldn't answer.

She finally came in with yet another question, which turned out to be, How many hours each week to work on each module against each milestone (parallelism: some people are working against milestone 3 while some are working on milestone 1). So I had the pleasure of opening my book (SurvivingObjectOrientedProjects, sorry for the ad) to p.102 and reading out loud,

"Avoid the most obvious unit of measurement: number of hours spent on each domain class to reach each milestone. This one will drive both your developers and you mad, even though it is the obvious measurement to request if you do not know the actual question you want answered. ... Dilbert fans use a random number generator to fill in their timesheets."

She sighed a big sigh, "yes, it's driving me mad," she said. So she agreed to push back against her boss and tell him he didn't need to know how many work-days had been put in against milestone 3. Besides, since we only have 4 developers to track, all working on different things, we can do our side of the equation in our heads. Our partners have 20 developers to track, so they probably can't answer the question at all. And, it turned out, there was a whopping 0.6 work weeks spent on milestone 3 in the last month, with around 1.6 milestone 1 work weeks missing. So the error was greater than the estimate! A DilbertMomentsAvoided just narrowly.

My position is that we have done all the estimating it is humanly possible to do by now, and there is nothing to do but sit back and wait for results to trickle in (or not) over the next two months.

Interesting to note that she, the project manager, did not create the nasty questions. It was her boss, trying to prepare for the next sponsorship meeting with external sponsors, trying to make sure he could show a report to them and answer all their questions.


I prefaced our weekly WorkQueueReport? with a three paragraph cover letter. The first two paragraphs were boilerplate that explained how to interpret the attached spreadsheet. The third paragraph, labled something like New In This Report, explained in two or three sentences what I had learned as I collected the progress data for the week. That's right, I wrote only two or three original sentences for a weekly status report. But they were very good sentences, and they were read. -- WardCunningham

View edit of October 24, 2002 or FindPage with title or text search