• by March 25, 2011

Traditionally, "time lag to conversion" has been defined as the amount of time between the last online display ad the consumer interacted with, to the first conversion activity that was performed.  And most ad servers have delivered a "Publisher Time Lag Report" to their clients detailing the number of users that convert within an hour, a day, a week, a month, etc., by publisher.  This was very meaningful when consumers clicked on online ads and in so doing clearly "marked" their intent, and it allowed marketers to make meaningful conclusions around which publishers refer users who convert and/or convert quickly.

However, over the last couple of years, there has been a huge shift in online user behavior.  Users have demonstrated a more sophisticated behavior to consuming online advertising -- choosing to click less and less on ads, but at the same time transacting online more than ever.  In other words, the "intent marker" that has been the anchor of online advertising is quickly becoming irrelevant, while the overall value of the online channel itself is actually increasing.  In fact, there has been a steady increase in dollars spent on online display advertising in order to influence consumers, who are clearly spending a lot more time online than ever before. This has resulted in most brands seeing their proportion of "view-through conversions" (those produced after only an ad impression) go through the roof while their "click-through conversions" (those produce after a click) drop to the floor.

This shift in online user behavior is one of the fundamental reasons why attribution management as a marketing strategy is gaining traction.

We know users don't click on ads as much as they used to. And the time lag report from the ad server only represents the time lag from the last impression to the conversion. As a result, many ask: is the reports even useful?

As a part of an attribution methodology that leverages the entire stack of marketing touchpoints experienced by the user - as opposed to just looking at the last ad - a whole host of metrics can be calculated.  One of them is time lag, but broken down further into a variety of "path markers," such as:

·       Time lag from First Impression to First Conversion

·       Time lag from Last Impression to First Conversion

·       Time lag from First Click to First Conversion

·       Time lag from Last Click to First Conversion

·       Time lag from First Website Visit to First Conversion

This data is extremely impactful in illuminating opportunities for optimization. Let's start with the time lag from the last impression (the metric the current ad server reports address).  The question this data answers is, "Of the users I managed to convert, how many made the visit to my site in the same browsing session that they saw the last impression."

Imagine that median time lag of a group of consumers from last impression to the conversion is 1.75 days. This means that 50% of your converters converted after 1.75 days and 50% converted before 1.75 days. So another 50% of them did not visit your site in the same browsing session as their last impression. Now imagine that only about 17% of those consumers converted in the same session. So about 83% of them actually came back to the website during a session different than the one in which they saw the last ad to convert.

First, this speaks volumes about the relevance (or the lack thereof) of "last ad"-based attribution. Next, it means your converting users are not likely to abandon their browsing experience to come see what you have to sell.  But they have obviously chosen to engage with you in other ways - arriving at your site at a time of their choosing and through a channel of their choosing (typically search or direct navigation).

What this means from an optimization standpoint is, you need to create messaging that supports a deferred action. Assume your users are not going to navigate to your site right away, so speak to them in a way that they can perform your desired action at a time of their choosing.

In the second part of this article we'll address the optimization implications of other types of lag time.