The Uncomfortable Truth about Segmentation
Everyone says you should segment your database and it’s been repeated so often it’s taken as read. It’s just common sense, right? But the latest data analysis from email marketing agency Alchemy Worx shows that this commonly held belief is wrong. The uncomfortable truth is that any open rate improvement is relatively small until the segment size drops below 5% of your list. If you are using four or five segments there is a good chance the extra campaign creation effort is not bringing you any reward, certainly not enough to cover the additional segmentation costs.
Alchemy Worx analysed data covering 224 major brands with very large lists such as Amazon, Groupon and About.com, looking at open rates across their campaigns over an 18 month period.What makes this insight unique is their analysis of performance against segment size. Here’s the most important chart from the data: showing lift in Open Rate as a function of relative segmentation size It shows how open rate performance is not substantially improved until list segment size drops to under 5% of the total list size.
When Alchemy Worx showed me their analysis, the implication hit me immediately. The message is that segmentation doesn’t make a meaningful impact until you have at least 20 segments and to really make an impact you need to be pushing towards 100 segments. Given this result is so shocking I’m not happy to take it as read, so Alchemy Worx are letting me crawl over the data to see if there is any chink in the armour. The above chart is the first output from the analysis currently in progress and we couldn’t wait to share, so we’ve released the early findings now. But we’ve lots more questions that need answering.
Why is segmentation with just a few segments seemingly not working?
How does email volume and frequency relate to segmentation?
Are brands with more segmentation sending lower volume?
What’s the trend?
Have brands being moving to more segmentation?
How are those brands with high volumes of small segments doing it?
Does segmentation have an impact on inbox placement?
The data analysts are hard at work and there’ll be a white paper published with more extensive analysis and insight into segmentation strategy you won’t find anywhere else.
Methodology and chart interpretation
The analysis draws on data from eDataSource. This covers billions of emails sent from 224 major US brands with the largest customer databases.The send size of each campaign was compared against the maximum send size for each brand to create segment size for each send. The horizontal axis shows the percentage of the list to which the campaign was sent. Highly segmented campaigns are at the left and full list broadcast at the far right.
The green open rate line is plotted against the left vertical axis. Rather than being absolute open rate the vertical axis is based on the difference from a re-indexed open rate. The difference from a re-indexed open rate is used because brands have different levels of average engagement. So the axis shows uplift (positive values) or reduction (negative values) against the brand specific average. This method levels the playing field across all brands and allows a like for like comparison. The lower shaded area and right hand vertical axis gives the volume of email sent. It shows that overall a relatively small amount of volume is sent without any segmentation, though the uptick at the far right shows that some brands are sending to their whole list.
About the Author: Tim Watson is an Email Marketing Consultant and founder of Zettasphere. He’s over ten years of email marketing experience, working with a variety of B2C brands on improving email strategy. His approach to strategy development is through maximising use of data for insight and metrics to close the performance loop.An international speaker, DMA Email Council member and regular blogger about all aspects of email marketing from strategy to deliverability.
Guest Post by Tim Watson, Email Marketing Consultant and founder of Zettasphere
Last updated: Jul 27, 2017
