Give Me Estimates Now Story

From GiveMeEstimatesNow:

Had a bit of a nightmare with this one: I was heading up a ChargingTeam? in a bank that for a couple of years was the golden squad. After one awesome year, we were allocated 10% of the sales credits. Of 110 IT staff, I was the only one with a desk on the trading floor - at the business's request and resisted by my management. I was once in a situation where the business had changed subtly and so the business rules in our engine needed an emergency recoding. It looked like a nightmare and with 30min notice and no clear articulation from the business of what they were after, I was thrust into a meeting with a load (20) of senior bods (senior to me anyway) who needed a date right then.

I wriggle so I don't get hung because you don't give estimates like this, right? It was bound to turn into the delivery date, and this was over Christmas and all the skilled staff were going to be off. Unfortunately, one of the senior bods thinks it is easy, because I worked with her on something very very similar that was easy a few years before. Unfortunately, there is a vital subtlety she is missing that completely changes the game, so I say there are important differences and I need some time to talk to my business analysts to clarify some issues. " Yeah, I hear you, but how much work is it?" keeps coming back and leaves me looking like a drongo because we're trying to resolve a crisis here, so to get things moving, it is a crisis after all, I give a fairly defensive estimate of 2 weeks to allow a day or two for each of requirements, dev, testing, parallel run, rollout + a bit of contingency. I thought it sounded pretty reasonable.

In earlier, preproduction days (ie no change control or regression test requirement etc) that thing we had done took about 15 minutes. It was essentially adding a mapping table. So this senior manager thinks what she's asking for now is the same, but it's horribly different because it changes the feedback loops in the control system that is now used for calculating the P&L for the whole firm.

Nothing much happens and a few adjustments get posted into the trade data and with a bit more analysis it turns out that the system could already support one of the interpretations of what they were after.

My reputation got trashed after that meeting because she told the other senior staff I was yanking their chain and it would really only take 5 minutes. I was demoted soon after, and that charging team, slowly fell apart (the last remnants of it are still around 2 years later) and never delivered anything further of any significance.

Thanks for sharing... what a story. I'm currently (Oct. 2010) in a GiveMeEstimatesNow situation...

I also have the risk of my reputation getting trashed by her for overestimating, but what can you do. Realistically estimate with no data? It takes a certain amount of time to estimate, unless you're on top of everything continuously to be pulled into a room of senior staff to provide shoot from the hip estimates. If they leave you in peace for long periods, they expect you to know it cold.

Sometimes I wonder if it really would turn out better if I lied, said it would be easy, and then just reported on the "unexpected difficulties and problems" that crop up and make it longer. They say that it's easier to ask for forgiveness than permission.

It's nice to be a consultant in such a situation (I Don't Live Here). When faced with this squeeze I can honestly say I don't currently know; in X days I will give you an estimate that will be accurate within Y percent of total time with Z percent certainty. If you don't give me any time at all to do research then I say it will take infinite time and cost you infinite moolah. If you don't like that then use somebody else for the next milestone. No prob.

Bleep happens in this biz. The complexities of interactions are often subtle yet significant; a kind of e-butterfly-effect. Keep your nose clean and you will eventually be promoted back up again. The gal who claimed it was easy is probably naive enough that she'll eventually hang herself on something else. If it happened to me a second time, I'd probably say something like, "There is factor X, Y, and Z that are different this time that make me hesitant to use the prior duration. If X, Y, or Z don't pop up, then her estimate is reasonable. If they do pop up, we're looking at several weeks."

Sometimes I break down my estimates like this:

60% less than 2 days 30% between 2 and 15 days 10% longer than 15 days

--BlueHat

Hmm. This still seems kinda fuzzy and indistinct. What are the components of generating a good estimate?

Identify the requirements. This seems obvious, but sometimes the devil in the details will byte you right on the tender spot if you aren't ready for it.
Isolate the things you don't know from the stuff you are sure of, and clearly define what the boundaries are. At least this gives management types something they can hang their hat on when you say you need time to investigate. If you already know how to handle X but are unclear on Y, then maybe they'll cut you a little slack.
Determine how long it will take you to do discovery on the things you don't know. We're not talking about an optimum, released product solution, just something that works. Once you've discovered what you need to know about Y to effect some minimal solution then you are ready to proceed with an estimate.

If you can't get the time to do this stuff then you're hosed. Start calling the borkers for another gig, man, 'cause this one is gonna die a horrid death.

--BlackHat

It sounded above like the author didn't have time for such detailed analysis and introspection. Note "now" in the title. A "gut-level" probability curve may have to do sometimes.

--BlueHat

CategoryStory