Well, it’s over and I am tired.
It was a great event. My head is still spinning.
One of my favorite parts was talking to Avinash Kaushik of Intuit and Web Analytics is Dead fame. This is a very sharp guy.

In any case, I had just finished presenting on report design (big shock) and he wanted my opinion on using UCLs and LCLs (Upper Control Limits and Lower Control Limits) in charts and I had a DUH! moment.

Using UCLs and LCLs (in this case, based on standard deviation) is a great way to show your report users what to ignore or not worry about. Normal variation (or as Avinash called it - biorythm) can appear quite extreme to the uninitiated so they might leap into action when action is not needed.

As a bonus, using constants like limits allows you to eliminate Axis grid lines (chart junk).

I always enjoy listening to the guys from Future Now about the Persuasion Architecture. The part that I like best is the strategic approach and thought process but it’s all good. We had John Quarto-vonTivadar, CTO, talking today and he took us through the persuasion architecture but also presented an XML tool that Future Now has developed for piping the correct configuration into your web analytics package (just HBX to start) for measurement of your persuasion architecture.

The one thought of his that I want to leave you with is this: Plan first. And that is not just planning your measurement, it means planning the persuasion path from A to Z before you even think about how you will measure it. Put it another way, its useless to measure a process if you haven’t planned the process before hand (otherwise how can you be sure what the data is telling you).

Other cool thoughts:
The data is inaccurate and it is still true (paraphrasing Jim Sterne)
The data is inaccurate - get over it (paraphrasing Avinash)
Note: the point of both these statements is that even with the inaccuracy inherent in web data, as long as we measure consistently we can still use it effectively

I’ll be at the Emetrics summit the next few days. Eric Peterson has promised to blog so hopefully I won’t have to correct him ;~)

This will be my first time at Emetrics as well as my first WAA annual meeting and I am looking forward to both.

Don’t use ‘em…

…Unless you have multidimensional data that requires 3 points of definition (even then it’s a risky idea).

3-D charts create perspective by tricking the eyes and these tricks skew users’ perception of the data.

Even multidimensional data doesn’t necessarily need a psuedo-3D or 3D rendering. The archetype being Minard’s chart of Napoleon’s march on Moscow. Each point in the plot has at least four (x and y location on a 2-D plane, direction of travel, and size of the army) dimensions and some (the retreat) include time and temperature. No fancy 3-D here and yet, it’s probably the best chart ever made (I happen to whole heartedly agree with Tufte on this one).

It has been my experience that, especially in business, 3-D charts are a gimic - they offer a quick hit of supposed professionalism or ‘design esthetic’ to a chart. They are very tempting because the software authors put some more design work into them than they did the other chart types. But at what cost? If the slickly designed chart skews understanding, doesn’t that potentially lead to the wrong decision and the wrong action?

So watch out for 3-D charts, they don’t typically add any value or better portray the information you are trying to convey and, in the worst case, can be harmful to the decision-making process.

Irregular posting will resume upon my return. ;~)

Cheers.