No email benchmark has a primary source

by Francis Rozange | Sep 8, 2026 | Email Marketing

Abstract 3D render: a precarious stack of translucent measuring plates floating over a void

You have seen the sentence, in a deck, an article or a sales proposal: “the industry average open rate is 21.5 percent”.

Take five minutes and walk back up the chain. The figure comes from a blog post, citing an annual report, published by an email platform vendor, calculated on its own customers’ campaigns.

There is nothing behind it. No institute, no authority, no independent panel. We looked: there is no non-commercial primary source on open and click rates.

Why nobody else can produce that figure

The reason is not laziness, it is structural. To measure open rates at scale, you need access to the sending logs of thousands of campaigns.

Those logs belong to the sending platforms. A public authority has no access to them, nor does a university lab, and the rare academic work on email covers authentication and filtering, not marketing performance.

That leaves the vendors. They are the only ones who can measure, and they are both judge and party: the figure they publish describes the market they sell into.

This is not an accusation of dishonesty. It is a methodological problem, and goodwill cannot solve it.

Three biases that alone disqualify the comparison

The sample is the vendor’s customer base. A platform reserved for large accounts and a self-service tool do not measure the same population. Neither measures the market.

The numerator is broken. The open rate rests on a pixel that Apple has preloaded systematically since 2021, for a large majority of the installed base. An open benchmark published today aggregates human opens and automatic preloads, with no way to tell them apart.

Sector segmentation is self-declared. “Sector: financial services” covers a retail bank writing to its customers and a broker writing to rented prospects. Their rates have no reason to resemble each other, and their average describes nothing.

The real problem is not the figure, it is the use

A benchmark answers one question: “are we good?”

That is a meeting question, not an operating question. It leads to no action. Learning that you are at 18 percent when the sector is supposedly at 21.5 percent tells you neither what to fix nor in what order.

The questions that produce decisions are of a different kind. Did this campaign do better than the last, on the same list? Has this segment been degrading for three months? Did this subject line change produce an effect above the noise?

All of them are answered with your own data, with no external reference. None of them needs a benchmark.

What to measure instead

Your own time series, on an indicator not manufactured by a third party.

The click is acceptable, with the caveat of security gateways visiting links. Downstream conversion, on your site, is the only signal nobody else produces on your behalf. The unsubscribe, the bounce and the complaint are acts, measurable without ambiguity.

And a control group, as soon as the decision matters. Holding 5 percent of the list out of a send costs 5 percent of the volume and makes the effect measurable. It is the only spend in this trade that produces a certainty.

These measurements compare to nothing outside. That is precisely their value: they depend on no sample whose composition you do not know.

The general rule, which goes beyond email

There is a hierarchy of sources, and it is simple to apply.

At the top, normative texts and operator documentation: IETF RFCs, guidelines published by Google, Yahoo or Microsoft, legal texts and regulatory decisions. These sources describe enforceable rules, and they can be checked in one click.

Next, peer-reviewed academic work, with its sample and its method.

Only then, and clearly labelled, vendor figures, when nothing else exists.

Nowhere in that hierarchy: comparison sites, rankings, articles citing articles. A figure whose source is an article citing a source is a figure with no source.

What to do tomorrow morning

Open the last deck used to decide something at your company. Take the three figures it contains and trace each one back to its original source, not to the page that quotes it.

Note, for each, who produced it, on what sample, and on what date. The sorting takes care of itself.

Then apply the same standard to what you publish yourself. A figure you cannot source has no place in a document that carries your name, and “everyone says so” is not a source.

Sources

LaFactory works email on the evidence: headers, DNS records, rejection logs. No open rate promises, ever. Get in touch for a deliverability audit.