# What I Learned About Web Analytics By Watching The Olympics

I was watching the Olympics the other night, and the first race of the decathlon was on. As the announcers explained the way the decathlon is scored, I was reminded of its complexity. Then they proceeded to have a "race" in which they talked about the person who "won."

In fact, everyone who ran got a score, and that score accumulates towards the gold. They were explaining that one runner came in fifth but that was fine because the 100-yard dash was not something they did well on and they would make it up on other events. So, what was being presented (times in the race) was not actually the cumulative score. There is a strong analogy between this kind of scoring and Internet metrics, where we look at things like click-through rates, but in fact evaluate things based on ROI. So the numbers in the initial report are not necessarily going to decide the winner.

Let's go deeper into the decathlon, as I have come up with what I think would be a really cool way for the networks to use technology to let us know how well any given entrant was doing in their quest for the gold. First some background from Wikipedia and some help from Murky Blog

The current scoring system, (it has changed over the years) is as follows:

The decathlon combines ten different track and field events, so to come up with a final score we need some way to tally up all of the scores. You know what that means: an equation. Let's imagine that you finish the 100-meter dash in 9.9 seconds. Then your score in that event, call it x, is x = 9.9. This corresponds to a number of points, calculated according to the following formulas:

points = ?(x0-x)? for track events,

points = ?(x-x0)? for field events.

That's right - power laws! With rather finel -tuned coefficients, although it's unclear whether they occur naturally in any compactification of string theory. The values of the parameters ?, x0 and ? are different for each of the ten events, as this helpful table lifted from Wikipedia shows:

 Event ? x0 ? Units 100 m 25.437 18 1.81 seconds Long Jump 0.14354 220 1.4 centimeters Shot Put 51.39 1.5 1.05 meters High Jump 0.8465 75 1.42 centimeters 400 m 1.53775 82 1.81 seconds 110 m Hurdles 5.74352 28.5 1.92 seconds Discus Throw 12.91 4 1.1 meters Pole Vault 0.2797 100 1.35 centimeters Javelin Throw 10.14 7 1.08 meters 1500 m 0.03768 480 1.85 seconds

If you are still with me, I will expound upon my solution. Each entrant has in his or her mind a goal that will get her to her high score. The announcers have some perspective on this as they talk about an entrant needing to do x or y in a certain event to stay on track to their ultimate score. What if the networks quantified this and showed place rankings not based on points scored in the events to date, but indexed to how well someone was expected to do if they were on track to win the gold. If this were done, then the viewer would have a better idea of how each of the entrants was doing relative to getting the gold. Net-net, they are reporting accurately but not in a meaningful way. Now, if we could only get NBC to adopt this.

Bringing this full circle, think about this relative to what you are reporting out in your analytics. For example, let's say for a particular client that the data showed that rich media ads with high interaction rates, video plays, etc corresponded to a net high ROI and conversion rate. When a new campaign is live, if it is trending higher than previous campaigns, you could say it is on track to be a great campaign -- if the campaign scores well in the other areas as well. It could be a sort of early warning system or indicator of the health of a campaign so that you can take corrective action if need be.

Are you giving the client a bunch of numbers because that's what the system generates to date, or are you taking some time to modify the data and project out so that the client can judge their true projected ROI on the effort? Food for thought.