Why did the campaign underdeliver, and who did it actually reach?

Marketing numbers have a habit of disagreeing with each other. The campaign report says one thing, the segment says another, and the list that was actually sent to is a third. Most of that is not error; it is different questions answered by different counts. An assistant that reads marketing data is only useful if it knows which count answers which question — so here are three questions, and the calls that answer them.

“Why did last week’s campaign underdeliver?”

A marketing manager on Monday. The assistant reads the campaign summary — the delivery funnel, what went wrong and of which kind, sends per day — and whether the sending server is configured and reachable at all:

GET /cmp-runs/summary
GET /cmp-sending-identities/mta-health

The funnel is built to be read by something that takes numbers literally. All four of its measures come from one query, not four — asked separately, they are four counts of the same rows at four different moments, which is how a funnel ends up reporting more opens than sends. Test sends are left out, because a test that reached one inbox is not a delivery statistic. And the failures are grouped by kind, so the answer can say which kind of failure the shortfall was — and, with the server check beside it, whether the problem was the list or the plumbing.

If the answer is bounces and complaints rather than configuration, the problem is reputation rather than software, which is its own post: deliverability is a reputation problem.

“How many real people will this segment reach?”

GET /crm-segments/{segment}/members
GET /crm-segments/{segment}/audience?channel=email

Two calls, two honest answers to different questions. Members is the report grid: one row per record, so two leads for the same person are two rows, deliberately. Audience is the send list: one row per contactable human, duplicates merged, people already merged away left out, and only those reachable on the channel asked for.

The gap between the two numbers is the answer the manager actually wanted. A segment of twelve thousand records that reaches eight thousand people by email is not a broken segment; it is a list with duplicates and missing addresses, and knowing the size of each is the difference between cleaning the data and blaming the campaign.

“What did the catalogue gain this month?”

GET /cms-pages/summary
GET /cms-pages

The website summary counts the catalogue by category and the products added each week, and the page list includes drafts. So the answer can say what went live, what was added, and which pages are still sitting in draft — the ones a campaign linking to them would send people to a missing page for.

Why pressing send stays with a person

Marketing is the area where an assistant is most tempting to let off the lead. It could draft the email, pick the segment and send it before lunch. Rutba Assistant will not do the last part. Anything that reaches your customers in your name is on the short list of things it will never be allowed to do, whoever approves — because a message people have read cannot be unsent, and “the assistant sent it” is not an apology a customer accepts.

• Sending a campaign. Yours to send, always.
• Publishing pages. Changing what the public sees on your site is a change, so it is a plan a person approves.
• Editing a segment. Changing who a campaign reaches changes the campaign. The assistant can show you who is in it and why; changing the rules is yours.