split testing ad titles

by 3 replies
4
Hey there !

I've launch yesterday a campaign testing 10 different titles.
Get 436 clicks for the total of the campaigns and the max clicks I get per title is 48, the lower is 40... I see no huge differences in terms of clicks between my 10 titles.
Of course, titles are not the exact same (change words, order of words, type of informations given about the offer...) but of course not the exact opposite.

What do you guys think ? Is that not enough data ?
#search engine optimization #split #titles
  • I think it is a horrible idea to test 10 titles at once. Test 2 at a time. No more.
    • [ 1 ] Thanks
    • [1] reply
    • Hi saylar,

      Usually I start out testing at least 4-8 ads at a time for each segment, however those ads often may include only 2-3 different headlines, I also vary the value propositions and CTAs (Call-To-Action). The more variables you test the more data you will need to reach a valid conclusion.

      You have not ran your test long enough. Generally, you need an absolute minimum of 2 weeks of data to reach validity. The statistical odds of your results being completely random, after only a single day of testing, is very high. You might have a couple headline stand out the next day of your test, and then another pair stand out the day after, and then back to nearly equal the day after that. Also, things will often fluctuate a great deal depending on the day of week, week of the month, and time of year.

      The length needed for a test to reach validity varies depending on the relative difference in the results. The fact that the relative difference in your test is small indicates that you do not have enough data to draw a valid conclusion from, keep running the test.

      Marketing Optimization: How to determine the proper sample size | MarketingExperiments Blog: Research-driven optimization, testing, and marketing ideas
      • [ 2 ] Thanks
  • That would be enough data to get some sort of idea about two titles, but for 10 titles it's just not enough. You could leave the campaign running for some time to get statistical significance, but I feel it would be more efficient to test the titles in pairs. Just start from some major difference in the message, and keep improving from there.
Join the discussion

Next Topics on Trending Feed