How to build a rolling email benchmark and use it every month
Build your benchmark from the last 13 months of sends, weighted by volume and split by audience and purpose. Rebuild it every month. Then judge each new send against its own benchmark, not against a hunch or an industry average.
Why build your own benchmark?
Industry averages tell you how email does in general. They can't tell you whether your invite to cold leads did well. Your own history can, and it's the fastest way to catch a problem before it turns into a trend.
How do you build a 13-month rolling benchmark?
- Export the last 13 months of sends from your email platform.
- Drop any send with fewer than 100 delivered. Tiny sends are noise.
- Split by audience and purpose. A newsletter and a promo aren't comparable, and neither are customers and prospects.
- Roll any purpose with fewer than five sends into one "other" row.
- Weight by volume. Add up the raw delivered, opened and clicked counts. Don't average per-send rates, or a 200-person send counts as much as a 100,000-person one.
- Each month, add the newest month, drop the oldest, and show the change against last month's benchmark.
Why 13? You get a full year of seasons plus the month you're reporting on. The catch is that the benchmark can drop because a strong season rolled off, not because you got worse. Don't panic at the arrow.
Rebuild from a fresh export every time. The platform keeps counting late opens and clicks, so older numbers shift after you first pull them. Patch last month's file and yours will drift too.
Where do timestamps come in?
They got me. At the last company I worked for, I was using the same export to find the best time to send. The timestamps ran three hours ahead of Eastern.
My first pass, built with AI, shifted the clock the wrong way and made evenings look best. I caught it and we flipped it. With the right times, the best window was 10am to noon Eastern. It was the busiest slot, and it had the lowest bounce rate of any slot with real volume.
Same day, different hour, big gap. Wednesday sends from 10am to noon opened at about 25%. Wednesday sends in the late afternoon opened under 19%. Plenty of sends were landing in that late slot by default.
The check is easy. Send a test record through and compare its timestamp to the clock on your wall in every tool you use.
How do you use the benchmark in your regular review?
Same routine every month:
- Upload the fresh 13-month export. Have the AI confirm it reads cleanly, summarize what changed since last month, and flag data problems before anything gets built.
- Rebuild the report by audience.
- Read deliverability first: delivery, bounce and spam rates. If those are off, you can't trust the engagement numbers.
- Then engagement: open rate, click-to-open and opt-outs, each against its benchmark with an up or down arrow.
- Check the oddballs by hand. Flag any small list with a bounce rate over 10%, and any newsletter far outside its usual range.
- Rebuild the send timing analysis from the same file.
- Save this month's benchmark values as next month's "prior" reference.
Conversion comes after that. Lead counts for a month look low right after month end, because attribution takes two to four weeks to catch up.
How does AI help?
It handles the tedious parts: the volume-weighted math, the formatted report, and the comparison to last month. A rebuild takes a few minutes instead of an afternoon. When the window rolls, it also tells me what changed. Old sends dropped off, new ones came in, and a few hundred existing sends had updated numbers.
Two things made it work. I gave it standing rules once (subtract three hours, skip sends under 100 delivered, weight by volume, keep the audiences in the same order), so every month started the same way. And I never took a number on faith. When something looked odd, I asked it to check the underlying data and tell me whether it was real or a data problem. The time zone mistake is why.
What best practices did the benchmark back up?
- Send between 10am and noon. Move sends out of late afternoon, especially midweek.
- Lead subject lines with something specific or a little tension. Skip labels like "Webinar:" or "Success Story:". The subject line has one job, which is to get the open.
- Match the subject line to how warm the audience is. Customers like affirming. Cold leads respond to tension.
- Test subject lines first, since they're the cleanest signal. Test body copy on the next send.
- Resend to non-clickers after 48 to 72 hours. Change the subject line, keep the body nearly the same, and stop at two sends to protect your sender reputation.
- Be careful sending twice in one day. It can raise complaints and opt-outs.
What should you do this week?
Export the last 13 months of sends, drop the tiny ones, and calculate one volume-weighted open rate for your biggest audience. Then compare your last send to it. It takes about an hour, and you'll know whether your last send was normal.