Email: stop what dirties the base before you clean it

Most teams meet a high bounce rate and reach for the purge. Delete the bad addresses, send again, watch the number drop. Within days it is back.

The purge was never the fix. If something is feeding bad addresses into your base every week, cleaning it is a task you repeat forever. You are mopping with the tap open.

We ran this on a French food media brand sending to a base of around 200,000. It was bouncing 8.62% of what it sent. Past a few percent, mailbox providers start reading your domain as a spam source, and the emails that stop arriving are the ones people asked for. After the work, a full-base send bounced 0.13%.

Here is the order we do it in. The order is the point.

How it works

1. Audit the intake

Before touching a single address, find every route by which an unverified one gets in. Imports, integrations, forms without confirmation, old lists nobody owns any more.

In this case the answer was blunt: 15,000 addresses were being injected every week with no verification. Every cleanup had been undone within days. Nobody had been bad at cleaning. They had been cleaning the wrong end.

2. Stop the source

Cut the injections. Park the flows that mail the oldest part of the base. Fix the sending records: here, two DNS records on the sending domain were misconfigured.

None of these is dramatic on its own. Together they produce a base that degrades faster than anyone can repair it. Until this step is done, nothing downstream holds.

3. Clear in bulk, reversibly

Now clean. Suppress through the platform's bulk API, in batches, and never delete.

We processed 62,984 profiles in 631 jobs. A suppressed profile keeps its history and can be restored. A deleted one takes with it the reason it went bad, which is exactly what stops you making the same mistake next quarter. Abandoned lists were renamed and archived, not removed.

Be careful what you clear. A subscriber who never engaged is not a bad address. That segment, 54,619 addresses here, was left intact and still receives.

4. Prove on the full base

Send real campaigns to everyone, and read the bounce rate.

Do not test on a sample. Bad addresses cluster, so a sample tells you nothing about the whole. We sent two full-base campaigns four days apart: 173,247 recipients at 0.35% bounce, then 171,353 recipients at 0.13%, with 99.87% delivered.

Frequently asked

We cleaned last quarter and it came back. Why would this be different?

Because you cleaned without closing the intake. If the rate returns on a schedule, something is refilling the base on a schedule. Find that first. The purge is the symptom, not the problem.

Why not just delete the bad addresses?

Deletion is irreversible and it destroys evidence. Suppression gets you the same deliverability with the history intact and a way back. One warning: segment counters keep counting suppressed profiles, so a cleanup can look like it did nothing when it worked perfectly.

Can we cut the bill by blocking unsubscribes too?

We tested it, on 50,311 profiles. Every one was ignored, because the platform had already suspended them and was not billing for them. It is a non-move. Check what your platform already does before you automate around it.

How do we know it actually worked?

You send to the full base and you read one number. Not a dashboard figure, not a projection from a test segment. If you did not measure it on everyone, do not claim it.

If this is your base

This is one narrow job, done in the right order, with the numbers read before and after. The full account, with the figures, is in the email deliverability case study.

If your base bounces more than it should, bring us the problem. Thirty minutes, and you leave with a map of what to fix first.