FACT : the same add repeated over and over again looses some, if not all of its potency, slowly moving from the status of a working incentive to that of a convenient bookmark for readers/customers (who by the way, already know you). As part of the analysis of our ADD campaigns results, this is one point we clearly identified, and definitely want to take into account.

As an introduction to this article, I'd like to point you to the first article in this series (Some insights on our ADDS Campaigns - part 1 - introduction ) in which we adressed  the question : what do we want to measure ?

We chose to measure what we call "useful visits". Now, everyone wants do do that, and there's not one definitive math formula to extract "quality visitors" from your logs. We choose to call "useful" a visit that both :

1. lasts more than 120 seconds
2. gets at least 5 different pageviews

To implement this in our pipeline, we've gone for a very simple indicator (evidently flawed, but better than nothing) which attributes a note, varying from zero to (no upper limit, but 2/3 is already a fairly good score) to each unique visit we get. It goes :

NOTE = ($Page_Views / 11 )  +  ( $Visit_Duration / 945) / 2

We divide the number of viewed pages by 11 (the overall average of useful visits on our sites) and then add half of the visit duration divided by 945 (the average of useful visits in seconds). Half, because if ever you drop your mouse to answer the phone then resume your browsing, your phone call counts as time spend on the site. And there really is no way of knowing how long that shower of yours lasted. By halving the result, we keep the aberrant measures to a minimum (those can be spotted by a variance graph). We've tested this formula on 50 selected visits (from our logs) and it satisfied us. But you should try and play with the constants for yourself.

By the way, even if we use a homemade third party application for our stats, this is easily implemented in Google Analytics. (more on that next time)

That being said,

First easily measurable effect : the Wearing of Add Banners

The two following graphs show what we mean by wearing of Add banners :




Keep in mind that we're only taking one factor into account here : the influence of the variety of displayed adds on the percentage of valuable visits. But in this precise situation, we feel confident that the dramatic changes in add efficiency are related to the add themselves. Not the add boxes, the visit time,... all things we also measure.

When we started our add campaigns, we measured a 30% average of useful visits in the first days. We had an overall of 8 adds running (only small sized ; buttons, halfbanners, squares). We experienced 1 week at 30%, then a drop to 25%. We produced 4 additionnal adds (1 skyscraper, 1 banner, 2 leaderboards) and got back up to 30% for a week. Then a dramatic and inexorable drop. When we reached 20%, we decided to produce a series of 40 new adds, which instantly propelled the usefulness of visits to 37% (our highest score sofar). As with the preceding adds, the effect wears off, but much more slowly. After a 5 weeks period only, did we reach the 30%  score we've now set as our bottom limit.

So yesterday, I've spent an entire day designing a 40 new adds from some of Mirlikovir's drawings. We'll see next week if the effect still endures,  and will update these diagrams.

Enough for today :)

Cathbad.