Thursday, 13 November 2008

New Insights On Time Of Day For Email

This is a really interesting post on time of day:

The interesting thing is that we have come full circle. In the early days of email it was a free for all but people quickly realised that email was all about getting the right message to the right people at the right time. It was the last part that caused all of the worry.

How would we as marketers know when somebody was likely to be receptive to our message? So, we assumed that all consumers care about is the weekend and applied some "logic" about their behaviour as it related to the weekend and decided Wednesday would be the best day to send.

Some of the more clever email marketers realised that if every body was sending on a Wednesday it would be too hard to achieve cut-through so they overlaid this thought with the previous thinking on the importance of the weekend and said Tuesday was better (although some argued for Thursday).

Throughout the Great Day of the Week Debate, individually we struggled to answer the question, "I have changed to the 'optimal' day and my results have gone down." Luckily as marketers we had a ready made answer to this dilemma. "Maybe our customers are different. We should test it." These individual conversations eventually lead to a groundswell until most everybody was in agreement that we should rigorously test for the optimal day of the week and hey, let's throw in time of day as well.

What we never understood was why. Now the why is not important if your customer base is homogeneous, but it is if you have more than one customer. The why allows us to understand the behaviours of different segments and target them appropriately.

A standard day of week test randomly puts your readers into buckets so you end up identifying the optimal day based on the average. This would hide the fact that your most profitable segments might respond better at other times. In fact, the optimum time identified by the test may turn out to be rubbish.

By looking at day of week by segment we can optimise the response across the board. I know you are thinking: "Wait a minute. By looking at this on a segment level the sample sizes would be too small to be statistically significant." This is where I am going to pick up a string of thought proposed by Dela Quist of Alchemyworx. Dela has never been a fan of the organised day of week test. He proposes that if you look back over the past twelve or twenty four months you will have probably sent on each day of the week. There are some hazards to this approach but if you were to structure your testing going forward where launch each segment on the same day over the next seven campaigns you would be able to normalise for any outside influences.

No comments: