Offline conversion import exists because of this. Tell Meta to find you leads and it will do exactly that: find the people most likely to complete a form. Within a fortnight your cost per lead is down, your sales team is furious, and the pipeline has not moved.
The algorithm did its job. It was given the wrong job.
The gap
A B2B software sale happens in stages: form fill, qualification call, demo, proposal, close. Weeks or months apart. The ad platform sees the first event and nothing after it, because everything after it happens in a CRM the platform has never heard of.
So it optimises for the only outcome it can measure, which is the one that correlates least with revenue.
Closing it with offline conversion import
Both Google and Meta accept conversions uploaded from your systems. You send back the stages that matter — qualified, opportunity created, closed won — matched to the original click or lead.
Once that is running, bidding learns from outcomes instead of intentions. The sources and audiences that produce qualified pipeline get more budget; the ones that produce form-filling enthusiasts get less.
This is the single highest-leverage piece of work in most B2B paid accounts, and it is routinely skipped because it needs a CRM integration rather than a campaign change.
Choosing the event to optimise against
Closed won is the truth. It is also often too rare to optimise against — forty deals a quarter gives an algorithm almost nothing to learn from.
The craft is picking the earliest event that correlates strongly with revenue and happens often enough to be useful. Frequently that is "qualified by sales" or "demo attended" rather than "demo booked". Work it out from your own funnel rather than copying someone else’s.
A rough rule: you want at least a few dozen of the chosen event a month. Below that, go one stage earlier.
Send values, not just events
A meeting with a five-person company and one with a five-thousand-person company are not the same conversion. If you can attach expected deal value by segment, value-based bidding pulls toward the accounts your sales team actually wants.
Without values, you optimise toward whoever books most easily, which is almost always the smallest.
What has to be true first
- Stages get updated reliably. If your sales team does not keep the CRM current, none of this works. Fix that before anything else.
- Qualification is defined in writing. Marketing and sales must agree what qualified means, and it must be a field somebody fills in.
- Click identifiers are captured and stored with the lead so the upload can be matched back.
Set the expectation before you start
Cost per lead will rise. That is the point — you are buying fewer, better ones. The number that should fall is cost per qualified opportunity.
If nobody has prepared the people who read the weekly report, the first fortnight looks like a failure. Put it in writing before the change, not after.
And the sales cycle problem
With a ninety-day cycle, a default seven-day attribution window is meaningless — the ad that started the deal gets no credit. Set windows to your real cycle, and keep a free-text "how did you hear about us" field on the form. For considered purchases it is frequently more honest than any model.
More on how we run these accounts on the SaaS page. To check whether the economics support the spend at all, the CAC payback calculator is the place to start.
Questions people ask
What is offline conversion import?
Sending events that happened outside the website — a qualified lead, a demo attended, a closed deal — back to the ad platform, matched to the original click or user. The platform then optimises towards those outcomes instead of towards form fills.
Which event should I send back?
The earliest one that correlates with revenue and occurs often enough to train delivery. For most SaaS accounts that is a qualified opportunity rather than a closed deal, because closed deals are too rare and arrive too late.
What has to be in place first?
A CRM that records the lead source and a click identifier or hashed email, a definition of qualified that the sales team applies consistently, and a reliable way to push the data back daily or weekly. The data discipline is the hard part, not the integration.
Does the sales cycle length break this?
It slows it. If deals close in ninety days, the platform learns from decisions made three months ago. That is why most accounts optimise towards a mid-funnel event that happens within the attribution window rather than towards revenue itself.