Most ad reporting stops where your product begins.
Your ad platform can tell you somebody clicked and filled in a form. For a software business that's barely the start - what matters is whether they finished onboarding, whether they reached the thing your product is actually for, and whether they paid. Paid media for software has to be measured inside the product, not at the edge of it.
Book Your Impact WeekThe paid media playbook
GTM Engineer allocation
25%
The spikiest load on your plan. Watching a live account is the small part; standing a new one up takes a week, and because placements get killed and replaced, that week keeps coming back.
- Every click followed into the product - onboarding, first real use, and payment.
- Visuals, copy, landing pages and funnels all under test, one variable at a time.
- Judged on cost per paying user, never on cost per lead.
Where this playbook comes from.
Buying ads for a software product is a different job from buying ads for anything else, because what you're selling gets used rather than just bought. A cheap click that ends at a signup can be worth nothing at all, and a more expensive one that ends with somebody in the product every week can be the best money you spend that quarter. You cannot tell those two apart from an ad dashboard.
So the measurement has to reach into the product before the spending starts. The click, the signup, the point in onboarding where people stall, the moment your product does the thing it exists to do, and the payment - each of them attributable back to the ad, the audience and the page that produced it. Wire that first and every decision after it gets easier.
A Designli go-to-market plan runs several channels at once, and which ones you run is decided in the Impact Week that opens the engagement. Paid earns its place when you already know who your buyer is and want to find out quickly whether your message moves them.
The platform will always tell you it worked.
Every ad account reports conversions, and every one of those numbers is generous. It's routine for an account to claim seventeen conversions in a month where only eleven records exist, and for a few of those eleven to have arrived from somewhere else entirely.
The platform isn't lying. It's counting a different thing - a click that happened near a form, a view that preceded a visit - and counting it in whichever way most justifies the spend. The only numbers that survive are the ones sitting in your own product.
So nothing gets bought until the whole chain has been proved: a real submission pushed through every tagged link, followed into the CRM, and followed again into the analytics your product writes. It's an unglamorous half day, and it's the difference between a quarter of learning and a quarter of a story.
How a paid experiment actually runs.
This is the scientific method with a budget attached. A hypothesis, one variable, a fixed window, a result that points at a cause - then the next one, on a drumbeat, for as long as the channel is running.
1. Wire the funnel, then prove it with a fake lead.
Every step gets instrumented before a click is bought - the landing, the signup, the point in onboarding where people stall, the first real use, the payment. Then somebody submits a test lead through each tagged link and checks it arrives correctly at every one of those steps.
2. Hand the platform your own customers.
The audience your plan already built goes up first, and you find out how much of it the platform can actually match. Expect a fraction of what you uploaded to come back - eight thousand contacts returning eight hundred matches is an ordinary result, and worth running anyway with the budget capped.
3. Publish the page the ad is promising.
No campaign opens until its page is live, and each angle gets its own page instead of sharing one. If the ad names the situation your buyer is in, the page has to be about that situation and nothing else.
4. Change one thing at a time.
The visual, the headline, the landing page, the funnel behind it - each takes its turn under test while the rest hold still. Move two together and you've bought a month of data that can't tell you which one did it.
5. Follow the click all the way in.
A conversion is not a customer. Read how far each cohort got into the flow, whether they reached the moment your product is actually for, and whether they paid - then hand that back to the campaign that produced them.
6. Change the offer before you change the bid.
When a page takes traffic and returns nothing, the targeting is rarely the problem. Eighty visitors and not one submission is an offer problem, and it gets solved by asking the people who own that audience what their members actually want.
A typical ramp-up plan.
Paid is the only channel on your plan that stops the day the card stops, and the same wiring that makes that true is what makes it quick - it also starts the day you start. A message that isn't working gets found out inside a month rather than inside a quarter. What that speed doesn't shorten is the setup.
Month one. Almost none of it is spending.
Accounts get opened, your customer list goes up and comes back matched at whatever number the platform can find, the funnel gets instrumented end to end and proved with a test lead, and the first page and first three angles get built. Expect one campaign to come back rejected and go up again the next day.
Months two and three. The rankings split.
The learning window closes and the first read lands, and the placement that sent the most signups turns out not to be the placement that sent the ones who stayed. The testing drumbeat starts properly here - a new visual, then new copy, then a different landing page, then a different funnel.
Month four on. The account becomes a rotation.
One placement has earned a small permanent line and everything else is on the clock, so the next test gets funded by switching something off rather than by asking for more money. The audience that worked carries onto the next surface instead of being rebuilt.
The number you report, and the number you act on.
Cost per paying user. Impressions, clicks and form fills are all steps on the way there, and not one of them is the answer. A placement that fills your CRM with people who never finish onboarding costs you more than a quiet one that produces three customers, and on an ad dashboard the two look identical.
Because the funnel is instrumented inside the product rather than at the form, every cohort can be read the whole way down: how many landed, how many signed up, how far into the flow they got, how many reached the point where the product does the thing it's for, and how many paid. Each of those steps traces back to the ad, the audience and the page that produced it, which is what makes the next test worth running.
And there is always a next test. The visual, the copy, the landing page and the funnel take turns under the microscope, one at a time, each running its full window before it gets called. That's the drumbeat - a hypothesis, a change, a read, a decision - and it's the same loop your product team is already running on the build side.
The honest limits. Paid stops the day you stop paying - there's no residual, and anyone selling you compounding on this channel is selling something else. Attribution stays approximate no matter how carefully it's wired, because the customer often arrives weeks later through a different door. And a channel you can scale by spending more can be scaled into a wall, which is why the ceiling comes from your own unit economics rather than from whatever the platform recommends.
What it does better than anything else on your plan: it tells you whether a message works while you still have time to change it. Search takes a quarter to answer and partnerships take two. Paid answers in weeks, and a wrong answer that arrives in weeks is worth more than a right one that arrives in nine months.
Let's find out what your ads are actually buying you.
It starts with a free 30-minute call. Tell us what you're spending and what you can currently see for it, and we'll tell you what's missing from the measurement, what it would take to follow a click all the way to a paying user, and whether paid is the right place for you to spend anything yet.