Part 9 in a Series on Web Analytics and Search Engine Marketing Programs
[September is here! That means X Change is getting close. The time to Register is now - and we have even more great speakers. Don't miss this chance to work in small groups with our new Expert facilitators - Rand Schulman, Clint Ivy, Anil Batra and Judah Phillips! They join amazing group including Eric Peterson, Matt Belkin, Olivier Silvestre, Aaron Gray, Marshall Sponder, Manoj Jasra and many more. Details and registration on the Semphonic web site.]
In an earlier post, I described one of the most useful tricks in an analyst’s kit-box – assigning time to various events on the web. I’m going to re-copy that discussion here because it’s the essential infrastructure to what I’m going to talk about in this post.
“… If you can create campaigns in your tool, you can setup a campaign for a specific time period. You can then use Response to that campaign as the criteria for a VISITOR segment that includes significant chunks of time before and after the Response. Now, you can compare the types of Ad Groups, Search Terms and Sourcing mechanisms before a particular campaign response (or conversion) AND after.
If you can’t create ad hoc campaigns in your tool, we suggest tagging key events (like campaign sourcing) with a date in a custom variable. Adding a date in a custom variable allows you to use that date or a date range to build a visitor segment identical to that discussed above.”
By using date-based campaigns or adding the date of source to a custom variable, you can create visitor segments based on WHEN a particular group of visitors were sourced. Suppose, for example, you create a segment of visitors whose first visit was in January 2007 and whose first source was PPC.
Using this segment, you can track January PPC visitors over time. There are quite a few interesting aspects of this over time tracking. The first, and most basic, is to understand the mere fact of return behavior. To show return behavior, you can trend the number of return visits by month – both for the whole campaign, for each engine and for ad groups (and even key search terms).
Once you have this basic data, the next step is to repeat the trend using conversions and/or conversion value.
Why is this important? For a PPC Manager, understanding the amount of return behavior helps define the window against which you have to track. The trend for conversions and conversion value is essential for understanding how to model the lifetime value of PPC sourced visitors. And this information can help you define the sales cycle – the typical durations between first acquisition and first sale.
Of course, all of this data needs to be carefully vetted in light of the known issues with cookie deletion. One of the most basic impacts in cookie deletion is the loss of tracking on groups over time. So if you measure a 10% return rate in February, there is an excellent chance that you are significantly undercounting the true rate. Worse, this effect is not cumulative (the fact that 30% of your visitors delete their cookies each month doesn’t mean that in 3-4 months all of your visitors will have deleted their cookies). Instead, it is often the same 30% of visitors deleting their cookies. So to adjust month out numbers, you can’t assume a flat rate of cookie deletion or a pure extrapolation.
Depending on your web site, you may be able to ascertain the true cookie deletion curve (at least for your customer population) by measuring the percentage of a specific set of customer logins or ids is associated with a “new” WA cookie over a period of time. With this curve, you can adjust the raw return visit counts given in the analysis above.
Using this curve for everything carries with it some dangers (if you think your customer base has significantly different cookie deletion patterns than your prospect or SEM pool) – but at least it’s a start.
These trends for return visits, conversions and value are all useful in and of themselves. They provide the SEM Manager a deep understanding of the overall impact of sourcing a visitor (and how it varies by Campaign and Keyword). I don’t think anyone would doubt that this is good to know. This knowledge can also shed light on how much information is getting lost when 3rd Party time-limited cookie based tracking is being used (as is usually the case with Bid Management tools).
For many sites, the biggest impact of this analysis will likely be around the assessment of lifetime value (and the cross-channel studies discussed in previous posts). Measuring lifetime value can raise serious optimization issues; because if lifetime value turns out to be significantly different for “converters” from different campaigns, then you can't use initial conversion as a good optimization metric. You also can’t optimize a PPC campaign based on waiting six months to see how lifetime value plays out. This problem is especially telling for ad-based sites who need to predict lifetime value based on projected return visits and page consumption. You need to optimize campaigns immediately. And while an analysis like this could spur you to make one-time immediate adjustments to a campaign – how do you go about the business of day-to-day optimization?
Fortunately, this is one place where you probably don’t have to worry so much about cookie deletion. Unless there is some reason why converters (or page consumers) sourced on one campaign are more likely to delete their cookies than visitors sourced on another, then the lifetime value numbers derived from our analysis of behavior over time are comparable (and therefore useful).
To tackle this long-term optimization issue, you need to carve out yet another time-period segment. For this analysis, you’ll want to segment based on a narrow time window of sourced visitors (often a single day). Now, you need to search for “lifetime-value” proxies in either the first visit or a short window after initial sourcing. The idea here is to look for patterns of behavior (conversions, pages, visits, etc.) that correlate with the long-term life-time value classifications.
Sometimes, you’ll get lucky and these will be relatively obvious. If so, it means that you can optimize your PPC programs based on the lifetime-value proxies you’ve discovered. This makes it possible for your PPC buys to be optimized against very short-run data – a big advantage. Sometimes, of course, you won’t. There just aren’t always good short-run behaviors that are predictive of lifetime value – and when they don’t exist you’ll have to use a combination of short-run and study-based optimizations when doing PPC buys.
Whenever you use proxies (be it for conversion or lifetime value) it’s also a good idea to check yourself with a deep-dive study every so often (perhaps twice a year) to make sure that the patterns you’re using still hold. Organizations tend to take the results of these studies and institutionalize them – making them into a kind of ritualized knowledge that nobody questions. But no matter how accurate these proxies may be for a given snapshot, they can age and become increasingly less useful as the business environment changes.
Tracking SEM visitors over time is a powerful and under-utilized method for SEM Analytics. It can provide PPC Buyers with a better understanding of channel interactions; a good sense of the true length of the sales-cycle; better perspective on the degree of error inherent in the PPC tracking from Bid Management or the Search Engines; and a much better measure of the lifetime value (and cost) of visitors sourced by PPC.
In the next post, I’m going to start looking at some types of measurement that are focused on true SEM optimization.
Other Posts in this Series: Introduction, Searchnomics Issues, Getting Setup for SEM Analysis, SEM Data vs. Web Analytics Data, High-Level Search Engine Reporting, Analyzing Search Traffic in more Detail, Measuring Search Effectiveness for eCommerce Sites, Measuring Search Effectiveness without Conversions, Measuring Search Engine Marketing as a Channel and Measuring Search as a Channel Part II.
Comments