Ad Testing: Research & Findings

Add Your Comments

Anton Konikoff with Acronym Media kicked off the morning session with a broad overview of ad testing methodologies, strategies and tactics. While much of the initial content was intuitive and common sense, it was helpful for the beginners in the audience. For the savvy PPC professionals, Konikoff did address a few common questions with specifics. He suggested testing 3 to 4 ads for smaller ad budgets and 5+ ads for larger budgets. In terms of data size, Konikoff suggested a minimum of 100 clicks with an ideal range of 1,000 clicks.

In terms PPC campaign test elements, Konikoff suggested evaluating headlines, display URLs, title, body copy, offers and messages. More granular test elements include punctuation, capitalization and dynamic keyword insertion (DKI). Advanced test elements include geo-targeting and day parting, messaging by target audience. He provided helpful examples and case studies to illustrate his points and reminded the audience to start small, and then grow test complexity and scope before making meaningful investments in PPC.

Hugh Burnham with Rare Method provided one detailed case study (OnAir.ca) to illustrate various PPC strategies & tactics. Specifics included his agencies client recommendations based on a comprehensive audit: changed broad to exact matches, raised daily budget, developed more diverse ad copy, expanded ad groups, turned PPC off during weekends and expanded keywords to include plurals and synonyms. While the case study didn’t include conversion metrics, all other metrics pointed to a successful campaign (including initial resulting business paying for the entire year’s ad budget).

Jonathan Mendez with OTTO Digital provided a simple test methodology: to design tests around a single question or thesis. He illustrated his point with two examples where one simple hypothesis was proven (or not) and additional questions that came out of those insights. Perhaps the most insightful element of his presentation was a detailed example of an L9 Taguchi Orthogonal Array in action and the resulting success of the text: a 52% lift in conversions for that client. Key leave-behinds included: ensuring a large differentiation between test elements, account for changes in temporal behavior and collect enough data to reduce margin of error (a common problem I’ve seen personally with many PPC tests).

Gord Hotchkiss with Enquiro (the heat mapping experts) reminded the audience that our job as SEM professionals is to decipher user intent. Unfortunately, there is not much data around keywords enabling insight into intent, so we must test thoroughly via PPC. Hotchkiss also reminded us to do more up-front research before developing test campaigns. He shared eye tracking reports that clearly demonstrate user behavior results in an F-shaped “golden triangle” or area of greatest promise and a broader consideration set (basically a larger rectangle that encompasses the top 1 to 4 sponsored listings and 1 to 2 organic listings.

Hotchkiss proposed and tested a single question: intent should impact scanning behavior. He provided an example scenario involving booking a reservation at Bellagio vs. learning more information on the hotel (research). The resulting eye-tracking charts indicated that 85 percent of users clicked on a result in the first 5 seconds. Additionally, purchasers spent three times longer on the results page (23 seconds vs. 7.3 seconds) and researchers spent a greater percentage of time in sponsor listings than purchasers. In research queries, 100 percent of the click-throughs were organic while purchase queries were 44.4 percent of click-throughs. After the testing was complete, the average interaction with a search engine results page was roughly 8 seconds.

Andrew Goodman, the session moderator, provided additional feedback to round out the discussion. He elaborated on findings from the panelists, including the fact that price points in ads are critical in some industries (where it’s expected). He gave a specific example where they tested price point breakdowns by day and week as well as altered the price point (raised the price of the product 3x). In the end, the increased price did not alter conversions. A very interesting finding indeed. Let’s all go raise our prices!

About the Author

Add Your Comments

  • (will not be published)