Tracking for Decisions in SEM

In the traditional retail business, people understand there’s a big difference between the tracking, data analysis, and analytics done for decision-making purposes and those done for accounting purposes. We can learn a lot from the brick-and-mortar world.

As interactive marketers, we often get spoiled with data access and therefore may become overly reliant on online-only data — even imbuing it with more importance than it deserves. Search engine marketers in particular easily fall prey to this over-reliance, because of the vast amounts of data at their disposal.

Still, we must understand offline realities. Not all buying behavior occurs online, and most companies offer consumers several interaction choices. For example, of the top 50 retailers recently named in Internet Retailer magazine, 36 have a telephone order option and 35 have retail stores (most with a store locator option).

Offline retailers provide excellent examples of ways to collect, process, and use data. Grocery and drug stores issue frequent-shopper discount cards to track buying behavior using bar codes. However, they know the information collected is a statistical sample. It’s not representative of all shoppers and may need to be corrected to remove biases. It may even need to be supplemented with additional data from Information Resources Inc. (IRI) and other scanner data resources.

After the necessary corrections are made, that data can be used for making decisions. Similarly, catalog merchants recognize that even with coupon codes and customer IDs printed on direct mail catalogs, data collection is imperfect. Yet, it’s highly valuable.

Search marketers should keep in mind offline buying behavior will occur on very different products or keywords than the online conversion behavior. That means the ratio of phone orders to online orders may be very different by product or type of customer.

When we measure conversions to phone orders for our clients, we often find items ordered by phone are more expensive, require more involvement (or complexity), or are heavy and require expensive shipping. Personal factors are important, as well. Some older customers may prefer to deal on the phone, for example.

Don’t know anything about your offline customer behavior? Then, chances are you’re making the wrong campaign decisions.

When you factor in offline behavior, online data becomes a sample (perhaps a skewed sample) of total conversions. All gatherable behavior data, both online and off-, must be factored in.

Confining thinking to online doesn’t solve the data-loss problem, either. It only introduces another set of issues. Even a pure online retailer such as will lose a certain amount of data because a product page or shopping cart is printed out and taken to another computer.

There’s no such thing as 100 percent data accuracy. Chances are, you spend a significant amount of time and energy trying to get the reporting to reconcile. Even if the result of every click from paid search could be measured, the data still may not reconcile with that provided by the search engine. Here are some reasons:

  • Engines automatically credit (and remove) clicks for any click-fraud they find.

  • Engines may be running on a different time zone than you are, resulting in different cut-off times.
  • Engines often automatically deduct a double or triple click made by a single user within a set time.
  • Engines will deduct clicks that look like affiliate network fraud, even though those clicks will be seen by your systems.

It’s easy to get data-fanatical in search engine marketing (SEM), particularly in the paid-placement environment. To some extent, this is useful. However, there are an almost overwhelming number of variables under your control that can dramatically impact the success or failure of your campaign, including:

  • Search engine

  • Price per click
  • Position at the time of click
  • Match type (standard, phrase, or broad)
  • Time of day of click
  • Day of the week of click
  • Title creative (your message)
  • Description creative (your offer)
  • Landing page variation (searcher’s experience after arrival)
  • Competitive effects (listings or offers the competition ran near your listings at the time)
  • Price (including competitive pricing at the time)
  • Inventory shortages

Measuring the impact of all these variables is nearly impossible. Even if it were possible, it would be prohibitively expensive, particularly if data is collected continuously. But don’t get caught up in trying to measure everything. Some data loss is OK.

For data types that can be measured reasonably easily, each variable’s impact on results can be regressed (once sufficient data has been collected). After the impact of each variable is known, you can develop campaign strategies to maximize profit based on your particular situation. But even here it’s important to take offline into account. It would be very unusual if you could achieve a perfect campaign strategy based solely on online data.

There will always be some data loss, both online and offline. Your mission is to make sure the very best decisions are made with the data you have. Don’t confuse data availability with data validity and actionability. Make informed, intelligent decisions based on valid data, and you’ll have a great campaign.

Related reading

site search hp