Opens are a mirror, not a metric
An open rate of 52% feels like a win until you remember that Apple Mail pre-fetches images for privacy. It can register an open before the human does. So a chunk of your "engagement" is a server in Cupertino, not a customer with a wallet. Clicks are softer than they look too — a curious tap is not a decision to buy.
I don’t ignore these numbers. They’re useful for diagnosing deliverability and subject lines. But I never let them stand in for impact. A vanity metric tells you the email happened. A P&L metric tells you the email changed something — a reorder, a reactivation, an upgrade, a margin you can point to in a finance review.
Measure the delta, not the absolute
The single most common mistake I see: reporting the absolute. "This campaign drove 4,000 orders." How many of those people would have ordered anyway? The honest number is the delta — the lift over a holdout group who got nothing. That gap is the only revenue the CRM program actually created. Everything else is taking credit for the weather.
So I hold out a slice. I compare treated against untreated, and I read the difference in the outcomes that matter — activation, repeat rate, contribution margin per recipient, revenue per send. If the delta is flat, the campaign was theatre, however pretty the open rate.
A message no one receives has a conversion rate of zero.
When the join doesn’t exist yet, build the bridge
Here’s the catch: most send platforms can tell you who opened, but not what those people did three days later in the product. The two systems don’t talk. The honest answer is to build the join yourself — wire the send log to behavioral and transaction data, key it on the user, and follow them downstream. Be your own analyst. The SQL that connects "we sent this" to "they did that" is where the real measurement lives, and almost nobody writes it.
Once that bridge exists, the questions get sharp. Did the people we messaged reorder faster than the ones we didn’t? Did the discount we sent actually lift margin, or just subsidize buyers who were already coming back? You can’t answer either from an open rate. You can answer both from a join.
The Uber Eats lesson: measure the thing that matters
At Uber Eats I found a platform-level defect in our SFMC setup that was quietly suppressing deliverability — across the platform, it sat around 84%. No open-rate dashboard flagged it as broken, because the opens you did get still looked fine. The problem only surfaced when I measured the thing that mattered: were the messages actually arriving? We took deliverability to 95%+ across the platform. That fix moved more revenue than any subject-line test, because a message no one receives has a conversion rate of zero.
That’s the whole discipline. Don’t measure what’s easy to count. Build the path to the number that ends up in the P&L — then read the delta, not the applause.