Have you ever had a thought wasted on conversion latency?
Is the conversion rate for a day or period just measured at this time or within this period, or is it the result of different influencing factors of a much larger, previous period?
If so, why do we not take this into account while calculating our conversion rate?
Feel free to have a look at my recent thoughts and please let me know what you think about it:
Thank you in advance
@CouchPsycho this is super interesting! Thanks for sharing.
Are you running an experimentation program that takes into account conversion latency? How has that changed the way you design and run your experiments, or look at your results?
i am still at the beginning. The most interesting thing is how you interpret the results. From my point of view: One move the time frame which is the basis for the results. After you have started your experiment you should remove the first days ( which equals your sites latency value) from the calculation, since these visitors were not seeing their variation all the time. At the point, where you would stop the experiment, you should change now change the allocation to put all new visitors in the original bucket. After a period which equals your latency value, you see the final result. Why? Because now all users which have been bucketed into a variation should have finished their "journey".
This means, that their is a shift regarding the results interpretation.
I think the longer your shops latency value is the more valuable is the effect of this change in conversion calculation.