This article was originally published on ClickZ.com on February 27, 2009. It remains current and relevant, because the consistent use of analytics to inform PPC search campaigns remains highly spotty in the industry. As I note in the text, PPC search advertising enterprising marketers a wealth of data that they can (and should) take maximum advantage from.
Many marketers don’t even bother looking at the data they have. In paid search, however, data drives decisions, so the more data you have, the more options you have. You can elect to use this data or decide that the incremental data either doesn’t have sufficient value to warrant analysis or can’t be fed into the decision-making process you use for bidding, segmentation, campaign structure, or keyword refinement.
If you’re lost about how much data to use, you aren’t alone. Studies and surveys of marketers differ on exactly how many use analytics to measure the results of marketing campaigns. The audience polled by these surveys has a huge impact on the estimates of level of analytics use.
One recent survey by Alterian reported that: “less than half (47 percent) actually use analytics to measure their campaigns.” Yet, if you ask a different audience and phrase the question differently, you’ll get vastly different results. The CMO Council asked a similar question last year and reported that: “when asked how they tracked and measured return on marketing spend, nearly 20 percent of marketers said they did not.”
You must ask yourself whether data is at your disposal that will allow you to either manually or automatically improve your campaign. Larger data sets in the hands of sophisticated analysts with powerful software can give rise to some great epiphanies.
But most marketers don’t have large enough data sets, even when they choose to capture a lot of information about every visitor. It takes a lot of data to build an accurate predictive model that isn’t out-of-date by the time you want to use it. Marketers of all sizes can benefit from at least evaluating whether additional data fields in their conversion data might be useful for building better models and, in the case of the largest marketers, influencing real-time bidding.
On most e-commerce and lead-generation platforms, you capture a lot of data during the sales process. You may even have pre-existing data from prior orders from the same individual. Of course, you already know all the details about the current order or lead generation session as well.
Because all this data is accessible to your systems, with the right coding, your “thank you” page has the ability to pass a variety of data into a third-party campaign management, analytics or business intelligence package. Some of this data’s importance is often overlooked.
You may want to consider looking at data you might have overlooked before. The obvious data point that one collects on the “thank you” page is the fact that someone made it there, i.e. accomplished a yes/no binary conversion. This fact alone is valuable in tying this event back against all the targeting variables at your control, particularly:
- Keyword and listing ID (exact searched query and therefore match type)
- Time of day
- Engine (not just the account, but the specific search engine or site source)
- Ad creative (some engines)
- Landing page (assuming you’re doing a A/B split test or fractional factorial experiment)
The real value accrues when you start moving beyond the convert/don’t convert data and look into variable data. For example, the following variables give you more options on how to manage and optimize your campaigns. I’ll cover the e-commerce data points first and then address some of the data points that service providers, B2B (define), and other marketers might have available as part of a lead-generation campaign.
- Revenue: With revenue data, you can manage your campaign based on a return on ad spend (ROAS), also often called revenue per-dollar spent.
- Net immediate profit: On either a dollar basis or as a percentage of revenue, many e-commerce systems can calculate the profit (or profit margin) of a sale. This ties in more to the bottom line than to the top line and many ROI-driven marketers (where return is defined as profit) would prefer to manage to an ROI (define) ratio.
- Number of items in shopping cart: The breadth of the shopping cart may be indicative of, or a predictor for, lifetime customer value (LTV).
- Use of coupon code and coupon-code type: May also predict LTV or whether the customer is sensitive to price (could be regressed against promotional or non-promotional creative elements).
- Repeat customer: Is the customer a repeat purchaser or a new one? This may be interesting if you want to know the different keywords used by new versus returning customers or it could be regressed against any of the targeting variables above.
- ZIP code: Does the stated ZIP code match the ZIP code derived from the IP address?
- Lead score data: For lead generation, were there variables that can be used to construct a lead-quality score? For example:
- Company size
- Time to purchase
- Title (for a pull-down list)
Each business has its own data set and I can’t cover them all in one column. In some cases, I can’t even begin to imagine the diversity of those data points and how powerful they will be in honing your campaign. Yet if you fail to take advantage of the extra data available at your conversion pages, you’ve definitely got a sub-optimal campaign running.