The metric stack that matters

Read every send in the same order, from the inbox outward. Delivered: how many emails actually arrived, after bounces — a falling delivery count is a list or reputation problem, not a copy problem. Clicks: how many people acted on the email's content. Orders: how many bought. Revenue: what those orders were worth. That stack, in that order, tells you where a weak send actually failed — nobody clicked is a different diagnosis from everybody clicked and nobody bought.

The single most useful number is revenue per recipient: total email revenue divided by the number of people the email was delivered to. It lets you compare a send to 2,000 people against a send to 20,000 on equal footing, and it punishes the vanity of big blasts to tired lists. Worked example: a campaign earns $4,200 delivered to 10,000 recipients — that's $0.42 per recipient. A smaller, better-targeted send that earns $0.60 per recipient is the stronger email, even though its total looks smaller in a screenshot.

Notice what's missing from the top of the stack: opens. Treat open rate as a soft, directional signal only. Apple Mail Privacy Protection preloads email images in the background, which registers "opens" for people who never read a word — on lists with many Apple Mail readers, open rates climbed without any human behavior changing at all. Opens can still hint that a subject line worked or that deliverability shifted, but no client decision should rest on them. Clicks and revenue are measured from real actions; build your reporting on those.

One more honesty habit: know your attribution window. When your platform says email "drove" a sale, it means the sale happened within a set number of days after a click or open — Klaviyo's default is a 5-day window, and accounts have been found running 14-day windows that credit email with nearly everything the business sold. Before you quote a revenue number to a client, know which window produced it. A marketer who can explain their own numbers' limits is the one owners keep.

A/B testing honestly: one variable, enough volume, a rule set in advance

An A/B test changes exactly one thing between two versions: the subject line, or the call to action, or the send time, or the offer framing. The moment version B differs from version A in three ways, B winning tells you nothing you can use next month — you don't know which change did it. This sounds obvious and is violated constantly, usually because someone "improved" the whole email for the B version. Discipline here is most of what separates testing from guessing with extra steps.

Before the send, write down the decision rule: which metric decides the winner (for a subject-line test, clicks are a sturdier judge than privacy-inflated opens), how the audience splits, and how long you'll wait before calling it. Decide it in advance because the losing temptation is to peek early, see the version you liked trailing, and invent a reason to let the test "run a little longer." A rule written before the send protects you from your own preferences — and it takes thirty seconds.

Volume is the other honesty check. Small differences need large audiences to prove anything. If a send produces about 200 clicks total, a version that gets 104 clicks against 96 has told you essentially nothing — that gap is noise, and next week's coin flip would reverse it. On a small list, test big swings (a plainly different offer, a clearly different angle) where a real difference is large enough to see, or run the test across several sends before believing it. Calling noise a result is how testing programs lose the client's trust.

Keep a test log for every client: the hypothesis in one line, what changed, the result, and the decision it produced. Over a quarter, that log becomes the most valuable page in your reporting — proof that the program learns. And test in the order money moves: offers and audiences first, subject lines and send times after. Subject lines are the easiest test to run and rarely the one that changes revenue; don't let easy wins crowd out the tests that matter.

Reading a month of results and saying what it means

At month's end, assemble the same picture every time: total revenue attributed to email (and, for a DTC brand, email's share of overall revenue — 20–40% is the common range for a healthy program), revenue per recipient across the month's sends, list growth and list health, and the best and worst sends with a reason for each. Same order, same numbers, every month. Consistency is what lets a client see the trend instead of a pile of screenshots.

Then translate. Every number you report gets a "so what" and a "so next": clicks on the welcome flow fell while traffic rose, so the first email's offer is stale — next month we test a new lead offer. Revenue per recipient rose on segmented sends and sagged on full-list blasts, so next month's calendar shifts one blast into two targeted sends. Three or four sentences like that are the actual product of your reporting. Owners don't need a data dump; they need to know what happened, what it means, and what you're doing about it.

The estimator below is the same arithmetic, pointed forward. Before you promise a client anything — or quote a campaign in a proposal — run the honest version: their real recipient count, a click rate in line with what their list actually does, a conservative share of clickers buying, their real average order value. It takes ten seconds and it keeps your proposals grounded in their numbers instead of in hope.