Adotas

Where media buyers start online

Featured Author

Author Photo

Gary Angel brings over twenty years of experience in decision support, CRM, and software development, to SEMphonic. Gary co-founded SEMphonic and is president and chief technology officer. He's responsible for leading SEMphonic's development of Web analytics and SEM decision making tools for web marketing professionals. In addition, he helps companies like WebMD, Intuit, American Express and Charles Schwab maximize their web channel marketing through intelligent use of Enterprise Web Analytics

More articles by Gary Angel





Features

Measuring the Intangibles: Don’t “Blink” Now, But You Might be Gravely Misoptimizing

Written on
April 2nd 2007
Author
by Gary Angel  |
Feed
   XML Feed

It’s for this reason that the two most common measures for impression value — Page Views and Uniques — are deeply flawed. Both measures encourage poor site optimization techniques. In each case, a site can greatly increase measured traffic without significantly increasing brand impact. How? One easy method is with a PPC program. If you are evaluating your sites on Uniques and rewarding site owners for driving traffic, then they are going to use tactics like PPC. Well and good. But since the measure is just traffic, they will have no incentive (indeed — they will have a positive disincentive) to attract qualified traffic. Why? It’s pretty simple. Qualified traffic is nearly always much more expensive than unqualified traffic. So if you are optimizing for traffic then you are always going to end up buying more and more unqualified traffic since it will give you more traffic for the buck. In PPC terms, this means you’ll end up buying words like “free stock quotes” instead of words like “portfolio management advice.”

When sites are optimized this way, it’s not unusual to see truly horrible campaign results — with the vast majority of traffic looking at only a single page on the site and spending less than 15-20 seconds (measured time) on the site. That isn’t an impression, it’s a blink.

Misoptimizations don’t end there. When you do web site tuning, the bias will be toward tools that increase total page views (if page views is the metric). Unfortunately, this often leads to optimizing for the most cumbersome and inefficient paths for users — with disastrous effects on site satisfaction. On the other side of the spectrum, if unique visitors is the only metric then site optimizers won’t work toward either more views, more time or more visits. Surely that can’t be right!

A good measure of impression value will need to encourage all of the following: relevant page views, repeat visits, and time in key areas. By doing this, it will force site owners to really optimize their reach — not simply increase their traffic.

Here’s a sample impression value calculation:
Site Impression Value = ∑visitor.engagement.scores * weighted impression value
Visitor.engagement.Score = ∑((engaged.Visits — AVG(baseline.engaged.Visits))/STDDEV(baseline.engaged.Visits) +
((pageEngagement — AVG(baseline.pageEngagement))/
STDDEV(baseline.pageEngagement));

This is a mouthful for sure. But it’s not really that complicated. In essence, it says that we are going to score each visitor based on how often he/she returns to the site and has a qualified visit plus how many relevant pages and relevant page time they consume in those visits.

The definition of a qualified visit will vary by site, but will probably be something like — ‘Viewed at least x relevant page(s) and spent more than x seconds on relevant content.’ Where only pages considered to be part of the sales/branding message will be considered. Driving to repeat usage does have some potential to misoptimization. But the potential is much lower than for most other strategies — especially when you force the sessions to contain relevant content and significant time on site.

In addition to the visit score, we have a pageEngagement score. The idea is to reward views and time spent on key pages on the site. It’s important that navigational pages not be included in this measure or else serious misoptimization will occur — with site performance appearing to benefit when users can’t find key content, spend lot’s of time navigating to content or simply land on the site.

In each case, the actual score is built by comparing the visitor score to an average and standard deviation calculated during a baseline period. Why do it this way? The goal is produce a normalized score that can be easily weighted. That means scores should cluster around a single point — and using Standard Deviations (a measure of variability around an average) will accomplish this. For most distributions, a Standard Deviation of +1 is quite a strong indicator of positive tendency. And almost all values will cluster between a range of +2 and -2 with the majority being close to zero.

This technique keeps outlier visitors from over influencing the score (we sometimes cap scores as well).
What’s important, however, is not the technical aspects of the score so much as the underlying concept that the essential task is to find a measure that will drive site owners to optimize appropriately for brand impact. For almost every site, that will require a measure of engagement that somehow includes the idea of key pages, repeat visits and the elimination of “blink” visits.

Once you’ve finalized your strategy for measuring brand impact you’ll have to translate that into a working implementation in your web measurement tool. This is by no means a trivial task — particularly with regards to tracking relevant visits and scoring visitors. Even high-end web measurement tools will sometimes make heavy weather of these tasks.

If you do have to make compromises here, it’s important to always keep in mind the underlying goal. Try to make sure that whatever compromises you make in measurement are ones that are least likely to result in bad optimization practices.

There is no one single right way to measure brand impact on your site. But if you’re site is about more than conversion, then it’s truly vital that you come to a decision about how you are going to measure that impact. If you don’t, you’re site owners and marketers are going to optimize to what they CAN measure. And when you do come up with a metric — you have to take special care that it drives to the behaviors that, ultimately, you really want. It all seems simple, but in the real world, it is anything but!



Pages: « previous page 1 2

Tags: , , and
Article Sponsor

More Features

Reader Comments.

No comments yet

Leave a Comment

Add a comment


Spotlight

Turn VP: Ad Network Shakeout “Inevitable”ADOTAS EXCLUSIVE – Turn bills itself as the world’s first Smart Market for online advertising. Turn’s VP of product and [...] more...

Adotas Partnership