Monday, April 1, 2019

Count and Measure


Why do you count?

What do you count?
Do you want the thing you're counting to get bigger or smaller?
What is important to measure?
How do you measure it?
What do you measure against in order to know if the number is moving in the right direction in an absolute sense?

Things change: for example, if you're counting the number of machines that aren't configured up to your standard and you add a thousand new machines, would you expect the number of non-standard machines to go up or down?

Assuming the machines are brand new, you'd expect them to be released to the current spec, so you'd hope the count of non-standard machines to remain the same, while the proportion of non-standard machines to the total would go down.

Even though the proportion looks better, you're not allowed to call this a victory, because you haven't actually fixed anything.

Similarly, if the count of nonstandard devices goes up while youre releasing those 1000 new machines, you should stop.  Stop and figure out why your brand new machines aren't going out with the correct configuration.  Stop the bleeding.  Fix the process,  release the rest of the machines with the correct configurations.  Unless it's easy to wipe them and start over, it's usually a bad idea to try go back and do a special fix for the machines that already went out with the bad configs.  It's cheaper and easier to simply treat them the same way you are treating the misconfigured machines you already had.

The earlier in the lifecycle you can fix an issue, the cheaper and faster it is to fix.  So if you see bad work, stop and fix it right away.  Then figure out what to do with the broken stuff.

The only way you know something is broken is if you're measuring it.

No comments:

Post a Comment

Patient Gardening

I was pulling weeds in my garden last weekend, and it struck me that there are a lot of parallels between gardening and cybersecurity.  I’m...