Partly sunny, chance of showers, bring an umbrella

Everyone knows that my 10 rules of performance are measure, measure, measure... etc. 10 times 🙂

OK, well, if you didn't know, now you do.

But does that mean that every time you tell someone about performance problems and/or risks that you should be doing so in quantitative terms?

I think not.  Although that got me a little friendly scolding today -- "What's up with this commentary with no numbers Rico?"

I'll tell you what's up:  In a performance report, like any form of communication,  it's important to know your audience and give them data in a form that is both useful and consumable.  What does that mean?  Well it means that if someone asks you what the weather is going to be like this afternoon probably giving them a description of location of local isotherms isn't what they had in mind.  It's a lot more likely that they just want to know if they need to bring their umbrella. 

Now you don't want to be redendering your opinion on the weather, or anything else for that matter, without reasonable data to base it on but that does not mean that you need to present every bit of the data or even any of it for any particular audience.

One of the things that I often get asked to do is to provide a gut-level feel as to where we are likely to run into trouble, and why.  This helps to provide some coarse steering and also to set the level of urgency.  This is an important function and it's not one that needs to be encumbered with vast amounts of data.  In fact, vast amounts of data can do the opposite -- make the conclusions less approachable, less likely to be read, and less likely to be acted upon.

It's important that when asked you can back up your reasoning with hard numbers but it's not always important, or even a good idea, to encumber executive summaries with a lot of data.  And it's certainly a good idea to keep your executives duly summarized.

Comments (5)

  1. Travis Simon says:

    An idea for an upcoming article: I would like to see measurements (and explanations) on LINQ’s performance. I am currently stuck in .Net 2.0 land, but am still curious for when we make the transition.

    Gut feeling: slow. But, who knows?

  2. AlSki says:

    Another issue here is that both performance and weather are localised. I commute over a hundred miles a day so this morning I got up put on my raincoat, and met my fellow travellers who included a young lady with summer shoes in her bag. Down in London my sunglasses come out.

    Performance can be just as variable. I develop lots of code on a single core laptop, and a dual core desktop, then release through various test environments into a production system. Our UAT system is (currently) faster than our production. So any performance measurement that I can make has to always be subjective. 56 seconds on one box can be five on another.

    The good news is that performance between environments is relative, unlike weather between locations. What I need is to hear is that its going to slow while my laptop simulates a three box environment, that it will run fine in UAT, but I have to know what factor it will be slower when it gets to production. If I run an isolated performance test in UAT will it be 2x or 50x faster than Prod at the busiest time of the day, and so will my 56 seconds become 2 mintes or 45 minutes.

  3. Norman Diamond says:

    Hey wait a minute.  The OPTIMAL response to partly sunny, chance of showers, is to bring part of an umbrella.

  4. While writing an Outlook addin, lots of people feel that they should try to help with Outlook performance

  5. Leggere docx (Office OpenXML) via Linq

Skip to main content