The Odd Result notification is triggered when there is a sudden change in the recorded conversion rate in an experiment. The exact criteria is as follows:
If the maximum conversion rate of the day has increased or decreased by more than 25% when compared to the maximum conversion rate of the previous day.
You will be notified by email and by an in-app message if this occurs.
If an Odd Result is recorded, it is highly likely that something external to the experiment you have set up is affecting your results. This could be having a positive impact, such as a sudden increase in exposure from a new Adwords campaign, or a negative impact, such as some aspect of your website working incorrectly or experiencing downtime.
You should visit your website and simulate some customer journeys to see if any issues are arising that would cause a sudden change in recorded conversions.
You should also look at your analytics data (through Google Analytics or whichever provider you use) to check if a sudden change in traffic is affecting your test results.
If you receive an Odd Result notification from a sudden increase in the conversion rate of one of your variations, you might be tempted to ignore it as increased conversions are the aim of the game. However, in the context of the specific test you have set up, a change in a variable outside of the test itself will make it difficult to ascertain whether to accept or reject your hypothesis.
Therefore, you might falsely conclude that variation B results in a better conversion rate when its results were unfairly affected by the external factor.