We covered defining the test in part 1 of this article. In this second post we’ll cover building the test defined in part 1.
Building A Test
Depending on your goals, your test structure may vary. The most common and easiest format to set up is the A/B test. An A/B test compares the performance of single elements within a creative. This format is great for comparing small tweaks to the same creative or dramatically different variations.
This A/B test focuses on different messaging types for a product. For example, a positive message about a toothpaste brand might emphasize a bright, winning smile while a fear-based message might focus on the likelihood of tooth decay and gum disease if toothpaste is not used.
Multi-variant tests test multiple elements in the same creative at the same time. These are more difficult to set up, but allow the tester to understand the inter-relationship of different components within a search or display creative. This results in a deeper understanding of which combinations perform best and why.
This multi-variant test focuses on different messaging types and different ways of communicating the message for a product. In addition to testing the message, the format of the message either as quote from a person or as a standard message are tested.
Once variations are built, testing can start.
Running Your Test & Measuring Results
Testing in online marketing is a difficult thing due to traffic volatility. There are a large number of tools online that will help measure whether enough impressions, clicks, and conversions are captured to be “significant”. Marin always recommends using click-to-conversion rate or ROI for most display campaigns and CTR, conversion rate, or ROI for search campaigns. The volatility in the quality of impressions and clicks, particularly with display means that any measure involving impressions is less likely to be repeatable or accurate. In our experience, assigning both a significance goal and a daycount floor (Marin recommends at least a week) is the best way to assure significant results. Make sure that your multi-variate test runs until all variables are considered significant, not just the first one. One important note: Equal rotation for creative while a test is running is recommended. Frequently, the publishers will reduce traffic much too soon for lower performing creatives.
Testing should never cease. You can be sure your best creative will grow stale over time as views increase and competitors copy your successful ad. Continuous iteration will help the diligent advertiser stay different and ahead of the competition.