Offline conversions for B2B: send closed-won deals or earlier SQL stages?

Picture a SaaS company with plenty of demo requests that closes only a small share of them, while the ad platform keeps optimizing toward the form fill. Sending sales outcomes back as offline conversions fixes that, but the hard part is picking which stage to send.

Closed-won is the cleanest signal. The catch is that if you close only a few dozen deals a month, bidding has almost nothing to learn from, and with deals taking a quarter to close, the feedback shows up way too late. SQL or other qualified-stage events give you more volume sooner, but they only work if sales marks stages the same way every time.

My view is that below a few dozen closed deals a month, SQL is the better target, because a noisy signal usually beats a starved one. That only holds if the stage definitions don't drift over time, though. How do you deal with that when reps update the CRM differently?

Who sets the stage matters more than what it's called. If a rep moves a deal to SQL whenever it feels right, the signal shifts with every new hire or quarter, and you won't notice until bidding starts acting strange.

My rule of thumb is to send the earliest stage that a system sets, not a person. A meeting logged with a real calendar timestamp, or a qualification field that has to be filled in before the stage can save, drifts much less than a dropdown does. Closed-won is still the cleanest signal, but below a few dozen deals a month, a stage with a hard trigger usually beats a vague one.

Agreed on system triggers, but a calendar event still drifts if reps book calls with anyone who replies to an email. The trigger needs a filter too, like a meeting with a decision-maker at a target account. Do you filter meetings that way before they count?

Agreed, but the filter itself is worth a look. If the decision-maker test depends on a rep reading a job title, the drift just moves from the stage to the title field. A check the CRM can verify without judgment holds up better, like the account being on your target list plus a title match from enrichment. Keep the unfiltered meeting event too, and compare both against closed-won over a quarter. If the filtered version gets more volume but closed-won doesn't move, the filter is the problem, not the bidder. Where does your title data come from, enrichment or manual entry?

Depends on your sales cycle. If most deals close within a month, a quarter of closed-won data is enough to judge the filter. On a six-month cycle that comparison shows up too late, so look at meeting-to-opportunity rate first.

The meeting-to-opportunity fallback has the same weakness. Opportunity creation is usually a rep decision too, so that rate drifts the same way the title field does, and you'd be judging the filter with a metric that has the same problem.

On a six-month cycle, I'd split the quarter into creation cohorts and compare filtered and unfiltered meetings by early outcomes, like whether they get a second meeting or a proposal. Early losses show up fast, and a bad filter usually reveals itself there well before closed-won matures.

Agreed, and I'd be just as wary of the proposal stage, since reps usually decide that too. A second meeting on the calendar is a cleaner early outcome. When you cut the cohorts, measure both arms at the same number of days after the first meeting, say 30. If you read the newest cohort at 10 days and the older one at 60, the filtered group will look worse only because it hasn't had time to book a second call.