Data Challenges for Small PPC Campaigns

Small PPC (define) marketers have it tough. I get a ton of calls from people asking me if my firm can help them manage their small (under $5,000 per month) campaigns. I respectfully decline and often steer them to one of the technology or service providers focusing on the small business market. Some small businesses want tools to help them manage their own campaigns. Most, however, want to know that a professional is handling it for them, because small business marketers believe search marketing to be difficult, and (depending on whom you ask) they may be entirely right.

Regardless of whether the small business chooses to manage its own paid search advertising campaign or to have it managed “professionally,” the reality is there are hundreds of thousands, perhaps millions, of businesses spending under $2,000 on PPC search and online yellow pages listings each month. These small business marketers face some unique challenges relating to their search spend. One big challenge they face is actually shared by larger-spending advertisers with very broad keyword sets and low data validity at the keyword (or perhaps even ad group) level.

Most of us have it easy. Spending sufficient funds on PPC means that we can watch conversions from our power keywords flow in on a daily or hourly basis. These conversions make it easy to calculate the appropriate bids for keywords given our ROI (define) or profit objectives, because we’re generating a statistically valid sample of data. The challenge with low spending accounts in general is that statistically valid sample data takes a long time to generate at every level of the campaign — from campaign to ad group, down to the keyword (or phrase) level.

With a bid at a specific level and an average position at a particular range, only three possibilities exist: the bid is too high, too low, or just right. How can one know if the data is missing or statistically suspect? One can’t, not really. But there are some solutions.

1. Wait it out: If you wait long enough, you’ll have statistically valid data (arithmetically) even on a keyword level. However, if your business is even slightly seasonal, the variation in the likelihood to convert (to an order, lead, or call) for clicks is changing over time. This makes waiting for valid data an iffy proposition.
2. Cluster “like keywords” together: Hopefully, you’ve done a good job organizing your campaign so that keywords in an ad group represent very similar search intent. This means searchers are likely to be very similar. Therefore, clustering the data at the ad group level or even combining ad groups may allow you to make a decision faster.
3. Cluster at the campaign level: The ad group level may not be enough for really low spend accounts. If the campaign was well constructed, one could make decisions at the macro level.
4. Try to attribute offline conversion behavior to the campaign at some level: While it may not be worth instituting phone tracking at anything other than the campaign level (or by simply asking customers which keyword they used to find you), chances are you don’t rank organically for most of your paid keyword choices anyway.
5. Use micro conversions or proxy conversions: If you have a site with a dozen or more pages, some pages may be used as a signal to indicate the click source was a good one, or at least better than average. For example, a “contact us” page or a product page might serve as a micro conversion. Using data from micro conversions will allow you to start differentiating between good clicks and poor clicks sooner.
6. Use the click-forward rate (opposite of a bounce rate): Users arriving on your landing pages can take only three actions: click back (bad), close window (bad), or click forward into the site. Like micro conversions, the idea of determining which clicks stayed (instead of left) can provide a valuable data set.

Even if you do all of these things, you probably won’t be changing bids nearly as often as a larger spending business would, so you may find that the bid landscape is swirling around you. Therefore, your bid velocities (speed of changing bids) are fairly low on a comparative basis. Eventually, as you learn more about what is working, your bids will align in the approximate correct range.

Dealing with data scarcity is a matter of tradeoffs. The more keywords for any low budget, the longer it takes to move individual keyword bids to an optimal level. Conversely, high keyword click concentration can help speed up the process.

I hope some of the above has helped you deal with the issue.