Accelerated Keyword and Engine Testing, Part 2

Keyword expansion can be critical to maintaining campaign efficiency and scale for paid placement search. When I test new keywords, I know the list will have winners and losers. Last week, I discussed aggressiveness when adding new keywords and engines to a campaign and how testing costs extend to testing new creative (ads and landing pages).

Generally, I like to set preliminary bids as high as possible, within reason. The easiest way to determine reasonable aggressiveness is by using existing data to model a test. When testing new keywords within an existing engine, look at your current campaign and find similar keywords. What do you pay for those keywords now? More important, what would you be willing to pay for the top position? Though the new keywords won't get exactly the same conversion rate or be the exact same quality, using current willingness to pay provides some guidance.

Engine expansion success is the most difficult to predict. Tier-two engines in particular get their traffic from various sources; those may or may not deliver the right mix of quality and volume. When testing new engines, I look at my clients' competitors' bids. Are they being smart with bidding based on return on investment (ROI) or other success metrics? If so, I take a cue from their bids.

When testing new creative, use power keywords first. Power keywords are high-volume keywords, brand keywords, or generic searches (shorter keyword phrases that are early buying-cycle phrases). To give new creative a chance to outperform the old, boost price/position at least a bit (except if you boost from position 2 to 1). Unless you conduct a multivariate test, you'll often get an answer to your creative "survivor challenge" in a short period. If you test many variables simultaneously, however, the amount of required traffic rises dramatically. That often requires more time and investment.

To get a statistically valid data sample for one keyword/engine/creative combination, expect to pay at least twice your allowable. Assume you want leads at $75 each or bring in orders at a given ROI that translates into $75 (based on a cost-per-order or a return-on-ad-spend percentage and average order size). The test requires a sufficient number of clicks to know if the new media combination will work. If the clicks are inexpensive and the conversion rate low, expect to use lots of clicks to get an order. If your clicks are more expensive, you'll need a high conversion rate to hit targets. You can use existing data to help model the test.

A good rule of thumb is you must spend 110-130 percent of your target allowable to reasonably presume the media pricing and creative aren't working. Even when you overspend the target by 20 percent (here, $90) without getting a lead or an order, it still may not be enough data to deem the test a complete failure.

Nothing hurts your business more than giving up on a potentially important keyword within an engine that can deliver enough click volume to make a difference, assuming you can find a way to hit your media objectives with that keyword. By giving up on a keyword, you give visibility, leads, and, potentially, customers to your competitors.

That said, some competitors may use much more aggressive ROI targets than your own, factoring in data that supports lagged or offline conversion or using branding metrics or buy-cycle data. Or, they could just be lunatics bidding crazy prices. Regardless of a competitor's motives, you're stuck with them in an auction environment.

When a keyword tested at a particular price/position with specific ad creative and landing page selection doesn't deliver, you may be tempted to give up on it. Here are some reasons you can't give up on a keyword with only a small data set on the first go-around:

• If you can get the keyword to work, the upside is very high.
  
  
• At a lower position, the same listing is distributed to a different set of search engines.
  
  
• As position changes, page placement often changes radically, resulting in a different type of searcher seeing and reacting to your ad. A top position may attract "compulsive clickers'" attention.
  
  
• Creative, keyword, and price/position work together for success. If one or more isn't optimal, the combination fails.

Continuous testing is a requirement of a pay-per-click (PPC) search campaign, or any marketing campaign for that matter. Clearly, top marketers, including your competition, likely reserve budget for testing, knowing some of that money will be spent learning does and doesn't work. To succeed in this marketplace, you, too, must continuously allocate a budget toward testing. I prefer to allocate 15 to 20 percent of the budget to exploring efficiency improvement and testing. Half goes toward new media (keywords and engines), the rest toward trying existing media with different creative (landing pages and ads). A well-constructed test is never a waste, even if all you learn is what doesn't work.

Meet Kevin at Search Engine Strategies in Toronto, Canada, May 4-5, 2005.

Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.