Case study $18,000 on tests: The other side of successful campaigns.
Katrina L

Any affiliate marketer knows that getting a profit is just the tip of the iceberg.

Behind successful cases, huge and little profits, there is always a lot of work, including: testing offers, creatives and traffic sources.

On the world wide web you rarely find cases with a negative ROI, but there are plenty of exclusively green statistics. We wrote about the expediency of such publications in the article “Bad “good” statistics. Is what we see true?“.

What most often remains in the shadow of publications about successful cases? Resources and time spent on tests prior to finding working bundles.

In the article we will discuss: how much can you spend on tests? Common mistakes and recommendations for testing, and of course the case of a webmaster, whose test budget was $18,000.


How much is “normal” to spend on tests?
On average, a test offer is allocated from 20 conversion costs. For example, if the payout for a lead is $2, then the minimum budget for a test in an advertising network is $40.  
Tests are an experiment to understand which of the combinations "plays" better.
It is clear that everyone has different budgets, and different offers, different GEOs, verticals. However, according to our estimates, the average cost for tests is between $50-500.
Recommendations for testing from 0.
  • Start with High activity cohorts, that is, with users who are subscribed to push notifications from 1 to 3 days.
  • Don't test 200 different GEOs at once. Start with 1-2 GEOs.
  • At $50, you can't get an idea of the quality of traffic on a Tier 1 network. Therefore, if you have a limited budget, do not choose large GEOs, as there is high competition, and, accordingly, a high cost per click.
  • It is more appropriate to choose Tier-2 or Tier-3 if you have a small budget for the test.
  • Quickly eliminate all ineffective sources and targeting and focus on no more than 1-2 offers.
Bad advices
  • Didn't test at all. Relying on luck is not a systematic approach to work, and if you are lucky with one offer, it does not mean that you will be with the next. Affiliate marketing is not a casino.
  • Test a million creatives at a time; At the other extreme, don't test creatives at all.
  • Not having an action plan. A strategy in affiliate marketing is half the way to a successful campaign.
  • Rushing it and finishing too soon;The test should be run for as long as it takes to collect enough data to draw conclusions. No more, no less.
  • Trying to copy someone else's successful case and expect at least the same result. The same applies to the use of other people's creatives.
  • Launch multiple GEOs in one campaign. This will not give correct data for analytics.
The cost of tests can be tangible, but there must be a system and understanding in everything. Если лить, как и на что попало, то и If you launch, as and on anything, then the result will meet the demand.
One of the affiliates working with Vimmy kindly shared the statistics of his tests, which he conducted in October last year. He spent $18k on tests, which came back as 0. The amount invested is equal to the amount received. Let's agree - this is also a good result.
Look at the amount of "red" lines. It is unlikely that such statistics motivate or inspire anyone.There is a desire to say that this offer, traffic or vertical does not work - the end.. But it is on the basis of this data that you can create white-list campaigns, choose better creatives and landing pages. Then launch and scale those very successful campaigns.
Wasted on nothing?
Sometimes, before getting a profit, tests alone can take a lot of budget. It also happens that it will not pay off, you also need to be prepared for this. It's pointless to give up here. Making money is not the original goal of the tests. Your goal is to test different layers of traffic on the network and find the ones you need.
The secret of success is simple and complex at the same time - tests, error analysis, retest, research, scaling.
How much do you spend on tests before drawing final conclusions?