Deliverability in year two of the sender rules

The Gmail and Yahoo bulk sender requirements took effect in February last year. Almost a year on, the programmes we look at fall into two clear groups, and the difference between them is not effort or budget.

One group treated the requirements as a deadline. Authentication went in, a one-click unsubscribe header was added, someone confirmed the complaint rate was acceptable, and the work closed. Those teams are now in a familiar pattern: placement drifts, a campaign underperforms, somebody opens a ticket, the ticket gets closed, and it happens again in six weeks.

The other group treated the requirements as the first version of a standing routine. Their records are no better. What they have is a monthly rhythm, a small number of numbers they watch, and a written rule about who gets mailed and who does not. They spend less time on deliverability than the first group and get better results from it.

The requirements were a floor, and everybody is standing on it

This is the part that surprises teams. Compliance stopped being a differentiator almost immediately, because compliance was achievable in a fortnight and everybody did it.

Authenticated mail from a domain with a published DMARC policy is now the baseline expectation, not a signal of quality. Once every sender in an inbox meets the same technical bar, receivers sort on the only thing left, which is how recipients behave when your mail arrives.

So the useful question changed. It is no longer "do we meet the requirements". It is "what does our recipient behaviour look like compared with the other senders competing for the same inbox".

That reframing matters because it tells you where to spend. Another month on DNS records buys nothing once the records are correct. A month spent on who you mail and how often buys a great deal.

Complaint rate is the number that governs everything else

If a programme watches one figure, it should be complaint rate, reported per mailbox provider rather than blended.

The blended figure is the problem. A programme sitting comfortably in aggregate can be well over the line at a single provider. That provider is where the complaints are, so it is also where the placement damage is happening. Pull the figure per provider or do not pull it at all.

Two practical points about reading it. First, the published threshold is a level at which you are already in trouble, not a target. Treat any sustained approach towards it as an incident, and investigate at a fraction of it rather than at the number itself. Second, the rate only falls if you actually remove people, so watch the trend over rolling weeks rather than campaign to campaign.

When the rate rises, the cause is almost always one of four things:

• A new acquisition source added addresses that never asked for marketing mail.
• Frequency went up without content changing, usually in the run-up to a trading period.
• An old segment was reactivated after months of silence.
• The unsubscribe route is hard to find, so the complaint button is doing its job.

Each of those has a different fix and none of them is a subject-line test. Diagnose before you optimise.

Reputation is per-stream, and most streams are still mixed

Sender reputation attaches to a domain and, to a lesser degree, to an IP. If your password resets, your order confirmations and your Saturday promotion all leave from the same domain, they share one reputation and the promotion is setting it.

The separation we recommend is by subdomain, one per stream: transactional, lifecycle, and broadcast marketing. Each with its own authentication and its own reporting. It is a week of work and it is the single most useful structural change available to most programmes.

The benefit shows up on the bad day. A promotional campaign that goes badly wrong should not be able to delay a password reset. With streams merged, it can, and the business impact of that is much larger than the campaign ever was.

If you have already done the staged DMARC rollout, you have most of the subdomain structure in place, since tightening policy per subdomain requires the same separation. The two projects overlap enough to be done together.

Keep transactional volumes on the transactional stream even when marketing would like to attach an offer. The moment a receipt carries a promotion, it is a marketing message with a receipt in it, and it belongs to the marketing stream.

Engagement-based sending is what the calm teams are doing

The teams who are fine are not mailing fewer people because they like smaller numbers. They are mailing fewer people because they have a written rule for who leaves the active file, and the rule executes without a meeting.

A workable rule has three parts: a window, an attempt, and an exit. No opens or clicks in a defined window, one deliberate reconfirmation attempt, and then removal from routine sending if that attempt gets no response. Write the window down, review it quarterly, and resist the annual pressure to suspend it before a peak season.

Suppression is the mechanical side of the same discipline, and it is worth treating as a process rather than a list. We set out how we run it in the suppression routine, and the short version is that removals must propagate everywhere within hours, including to systems marketing does not own.

Good list hygiene also means being honest about acquisition. Every source that adds addresses should be tagged at the point of capture, so that when complaints rise you can trace them to a source rather than guessing. Sources you cannot trace are sources you should not use.

Build the monthly review that catches drift early

Deliverability problems are rarely sudden. They are usually three months of small changes that nobody was looking at. A one-hour monthly review catches nearly all of them.

What we put on the agenda, every month, in this order:

1. Complaint rate by provider, current month against the previous three.
2. Delivered rate and bounce rate by provider, same comparison.
3. Active file size, plus what was added and what was removed.
4. Send volume by stream, flagged if any stream grew more than a fifth.
5. Any authentication change made anywhere in the business, including by teams outside marketing.
6. Open questions from last month, with owners.

The fifth item catches the thing that breaks programmes: a new tool bought by another department, sending as your domain, with no alignment and no suppression. Ask the question monthly and you hear about it in weeks rather than after the damage.

Keep the output to one page and keep the history. Deliverability arguments are usually arguments about what changed and when, and a year of one-page reviews settles them in minutes.

What to do this quarter

• Split complaint rate by mailbox provider and put the per-provider series on the same chart, so the divergence is visible.
• Separate transactional, lifecycle and broadcast traffic onto their own subdomains if they still share one.
• Write your inactivity rule down as a window, an attempt and an exit, and schedule the first run before peak trading rather than after it.