People no longer only type a search; they ask an assistant where to buy good coffee nearby.
Story · Solution
The marketing lead runs a campaign and sees whether it moved anything
I send a campaign, sales go up a little, and I cannot say whether the campaign did it or the month would have been good anyway.
Sound familiar?
I cannot say whether a campaign moved sales, and a discount can reach someone who complained last week.
Why it hurts
I run marketing for a small coffee roaster. I send a campaign, sales go up a little, and I cannot say whether the campaign did it or the month would have been good anyway. I also worry about who receives it: sending a discount to someone who complained last week only makes things worse.
How it is solved
Send a campaign to the right people, with the owner's approval, and afterwards see what changed against the weeks before it.
- The marketing lead starts the campaign. It is drafted from what is selling, with its discount code and its tracking tag, and stored as a draft. Nothing is sent.
- The marketing lead chooses the audience. The app finds the subscribers who have stopped ordering. Anyone among them who has complained recently is left out of the audience and named, so a discount does not land on an open complaint. Nothing is sent and nothing is stored.
- The marketing lead stages the drafts for approval. Each email and design request is held for approval, and each email carries the recipient count the app works out from its audience. A draft scheduled in the past, or with a photo still missing, is refused with its reason. Nothing is approved or sent here.
- The owner approves the send. Nothing is sent until the owner approves it, and only the owner may approve a large send. The approval is recorded with who gave it, when, and the recipient count. In the demo the send is simulated and labelled as such.
- The marketing lead measures the result. When the campaign's window has passed, the app counts how many times the code was redeemed and the subscriptions signed up with it. It sets each pushed product's revenue against the weeks before and the same weeks last year. Every change is marked as inferred, never proved to be the campaign's, and there is no total across products.
- The marketing lead reads what worked and what did not. The growth marketer, an AI agent, writes one paragraph on the result. The app fills in every figure, and an inferred change is worded with care. A paragraph that breaks those rules is held with its reason.
Works with
Stripe
The result
Both ends of the campaign are covered. Before it: the audience of lapsed subscribers, with anyone who complained recently left out and named, and the send approved by the owner. After it: the code's redemptions and sign-ups, each pushed product's revenue against the weeks before and the same weeks last year, and a plain account of what worked and what did not.
- If a product's revenue rises against the weeks before, the change is shown, marked as inferred rather than proved.
- If sales move no more than they would have anyway, the comparison shows it.
- If a subscriber has complained recently, they are left out and named, so no discount lands on a complaint.
Try it
The demo runs on WidgetCo Coffee Roasters, an example business: a coffee roaster with an online store and a wholesale side. You will see it as Priya, the marketing lead.
Connected to this story
The app
Read next
Why: Now possible, also about marketing, with an example app you can try today