[jib]

The best lesson I ever got about goals I got from learning to play (6-max, limit, texas hold'em) poker. Poker has stupidly high variance to profit ratio, and an endless amount of metrics you could measure to track progress.

Obviously, the goal of playing is to make money, and almost equally obviously, focusing on how much money you made is actively harmful if done for sample sizes lower than 100k hands or so.

The solution to improvement through quantitative analysis in that environment is to find key drivers that you believe impact your overall goal, analyse trends you see in those drivers, and based on those find a few areas that you want to qualitatively drill into to figure out changes you can make. I.e. - "I open 20% of the time from early position and 25% from late position - if I compare to other successful players that is too low a difference, so lets see if I can find some situations where I may be making bad decisions and plug those holes".

That is what we want to do with goal setting in business as well. Yes we have some high level KPIs that we want to improve, but actively working "to improve your revenue" is harmful and will no doubt lead to selling cars that explode etc. To be successful you set your overall KPI(s), then you forget about it and focus on smaller things that will over time incrementally add to your overall goal.

If you have managers and employees who cant make that separation or want to take short cuts to directly affect the overall goal without focusing on the smaller bits, or comparing to some kind of "best known" practise, then you end up with exploding cars, tools that has all features and no usability etc.

[tunesmith]

This seems like a clear recipe for premature optimization, or "the root of all evil" as Knuth likes to say.

Ending up with an exploding car is also an end result of premature optimization. It means you're focusing on revenue to an extent that is counter-productive.

That means that revenue isn't the overall goal, either. The overall goal is "an ever-flourishing company".

KPI is only one way to try and marshall the resources of a company to move in the right direction, but it seems easy to use wrongly, leading to balkanization of departments, etc. The other way is to use a system like Theory Of Constraints, which encourages a systematic view of the company's overall goals, that is transparent down through the implementation levels, and do regular constraint analysis to identify the (ever-changing) root constraints that block progress towards the overall goals.

[amirmc]

> "To be successful you set your overall KPI(s), then you forget about it and focus on smaller things that will over time incrementally add to your overall goal."

I don't see how this works in practice. Once you expose KPIs like this, then everything in your org works to increase them. Promotions probably get handed out on this basis. As you say, this leads to exploding cars and bad tools. The challenge (which you don't mention) is how you create an environment that doesn't hold the metrics higher than other things. I'd argue that values-driven orgs probably do this better but I'm not sure.

[jib]

Yeah that is the challenge. If I knew the right answer that solves it I would certainly share, but I dont.

What I try to do is talk about it over and over, and emphasize for each new initiative we roll out how we dont care about how it will impact the KPIs, we care about how it will impact our overall goals that the KPIs are trying to measure.

That said, #1 resistance point for most initiatives is still "how will affect the KPIs", so we are certainly far away from solving that problem.

Things we've tried to de-emphasize the KPI importance is going away from having an "exceeds" measure, now we are just "meets objective" or "doesnt meet objective", with "doesnt meet" set fairly low so that the people who like to min-max things dont have another clear target to min-max towards, and doing things like moving the focus to team oriented goals rather than individual goals to lower the link between KPIs and rewards, and instead linking individual rewards to general appraisals.

[epaladin]

For those less initiated in management-speak, KPI is apparently "Key Performance Indicator" - http://en.wikipedia.org/wiki/Key_performance_indicator