Product-Qualified Leads: Scoring, and PQL vs. MQL
A product-qualified lead is a user who has already used your product and behaved in a way that predicts they'll pay. Not a fit on paper. Not someone who downloaded a whitepaper and got a score bumped by ten. A person who signed up, poked around, invited a colleague, bumped into a usage limit, and by doing so told you more about their buying intent than any form field ever will.
That's the whole idea in one breath. The rest of this piece is about drawing the line cleanly, scoring it without kidding yourself, and knowing when the thing you actually want isn't a lead at all but an account.
What separates a PQL from an MQL
Here's the difference stated plainly: an MQL is inferred intent, a PQL is demonstrated intent. A marketing-qualified lead is scored on who someone appears to be and what marketing they touched. Job title, company size, email opens, a webinar registration. A product-qualified lead is scored on what they did inside the product. Those are different kinds of evidence, and one of them is a lot harder to fake.
The conversion gap is not subtle. Gainsight's 2025 benchmark work puts PQL-driven teams at roughly 25-30% conversion against 5-10% for MQLs. Multiple PLG benchmarks land in the same neighbourhood, which is unusual enough in this industry that I trust it more than a single vendor's number. It gets better as deals get bigger: Optifai's 2025 product-led growth guide reports PQLs converting near 30% for products with a $1,000-$5,000 average contract value, and closer to 39% once ACV climbs into the $5,000-$10,000 range.
So why isn't everyone doing this? Because it's more work. You have to instrument the product, agree on what "value" means, and get sales to trust a signal they can't see in the CRM out of the box. ProductLed's 2025 benchmark, drawn from more than 600 B2B SaaS companies, found 58% now call themselves product-led and 91% of those plan to spend more on PLG. Yet only about a quarter actually use PQLs. There's a big gap between believing in product-led growth and having a working definition of a qualified lead, and most teams are sitting in it.
| MQL | PQL | |
|---|---|---|
| Core signal | Demographic and firmographic fit, marketing engagement | In-product behavior |
| Evidence type | Inferred intent | Demonstrated intent |
| Typical inputs | Job title, company size, content downloads, email opens | Key action completed, teammates invited, limit hit, features adopted |
| Data source | Marketing automation, forms | Product event stream |
| When it fires | Before the person touches the product | After meaningful usage |
| Reported conversion | ~5-10% | ~25-30% |
| Fails when | Buyer looks right, never uses anything | Power user with no budget or authority |
Notice the last row. Neither type is complete on its own. A PQL who loves your tool but can't sign a purchase order is a real limitation, and pretending otherwise is how PLG teams end up with a pile of enthusiastic hobbyists and no pipeline. Fit still matters. It just stops being the only thing that matters.
The distinction almost nobody draws: PQL vs. PQA
Now the part that trips people up. A PQL is a single person. But most B2B software is bought by accounts, and an account is not one person having a nice trial.
A product-qualified account (PQA) is an account where enough individual users have crossed the bar. Inflection's 2025 write-up sets the common threshold at 50% or more of an account's users being individually PQL-qualified, sometimes paired with a floor like three active users in the last seven days. The unit of analysis changes, and so does the sales motion.
Think about a restaurant reservation. One person browsing the menu online is a PQL. A table of six who've all looked at the menu, asked about the wine list, and started arguing over the tasting course is a PQA. You'd staff those two situations completely differently, and you'd be foolish to send the same follow-up to both.
Why does this matter in practice? Because a single power user trialing your tool across their personal projects is a very different commercial situation from three departments quietly adopting it across forty seats. If your product is bought per-seat by teams, PQA is your real qualifying unit and PQL is the raw material that rolls up into it. If you sell a single-seat tool, PQL is the end of the line. Get this wrong and you'll either spam individuals inside an account that sales is already working, or miss a groundswell because no single user scored high enough on their own. The account was ready; you were watching the wrong row.
A scoring rubric you can actually adapt
Let me give you something concrete instead of the usual hand-wave. Score four families of signal. I weight them deliberately, and the weights are the argument, so change them for your product but change them on purpose.
Depth: how far into value a single user has gone. Completed the core workflow, created a saved report, ran the thing your product is for. This is the heart of the score; weight it heaviest.
Breadth: how much of the product surface they've touched. Connected an integration, used a second and third feature, invited a teammate. Breadth is the strongest predictor of stickiness, because a user wired into three parts of your product doesn't churn on a whim.
Velocity: how fast and how recently. Three sessions in the first week beats three sessions spread over two months, and a login yesterday beats a burst of activity that went cold in April. Recency is doing quiet, heavy lifting here.
Fit: the ICP check that keeps you honest. A perfect usage score from a student on a hobby project is not a sales lead. Fit is a gate as much as a score; a great behavioral profile with zero fit should cap the total, not average out to "medium."
A worked example. Say you cap the score at 100:
| Signal family | Example trigger | Points |
|---|---|---|
| Depth | Completed core workflow once | 25 |
| Depth | Created and saved output (report, project, dashboard) | 15 |
| Breadth | Invited at least one teammate | 20 |
| Breadth | Connected an integration | 10 |
| Velocity | 3+ active days in first 7 | 15 |
| Velocity | Active in last 7 days | 10 |
| Fit | Matches ICP (company size, role) | gate, not additive |
A trial user who finishes the core workflow, saves something, invites a colleague, and shows up three days in their first week lands at 75 before you even look at fit. That's your PQL threshold candidate. A user who logged in twice, clicked around, and never completed the core action sits near 25 and stays an ordinary trialist.
One honest wrinkle on the numbers-versus-binary question. Pocus's 2025 scoring guide makes the point that PQL qualification often works cleanest as a binary outcome, PQL or not, even when there's a point total underneath ranking who to call first. The score sorts your list. The threshold decides who's a lead at all. Don't let a pretty 0-100 gradient trick you into treating a 48 as "almost qualified" when your data says 70 is where conversion actually turns.
Where this goes wrong
I've watched more PQL programs wobble than land cleanly on the first try, so a few failure modes worth naming.
The first is scoring activity instead of value. Logins are not a signal. Time-in-app is not a signal. I once sat with a team whose top-scoring "PQL" was a user who left a tab open for six hours a day, which their model read as deep engagement and their sales team read, correctly, as noise. Count the actions that mean the user got the thing your product promises, and be ruthless about the rest. Vanity metrics love a scoring model because a scoring model launders them into something that looks rigorous.
The second is a threshold nobody validated. Picking 70 because it's a round number is guessing. Look at users who converted and users who didn't, find where behavior actually diverged, and set the line there. Then revisit it, because the line drifts as your product and pricing change.
The third is the one-directional handoff. A PQL fires, sales pounces, and the moment the user goes quiet nobody tells anyone. Intent decays. A lead that qualified three weeks ago and hasn't opened the app since is not the same lead, and your model should be able to say so.
Where the score lives
None of this works on a whiteboard. You need the product event stream feeding a scoring model, and you need the result somewhere sales will actually see it. Roughly three shapes of tooling do this, with honest trade-offs.
Product analytics platforms with a PQL angle — Amplitude, for instance, publishes patterns for building PQL signals and pushing them into a CRM. You get deep behavioral data, but you're often assembling the scoring and routing yourself.
Dedicated product-led sales tools — Pocus, Correlated, Endgame — are purpose-built for exactly this: pull usage signals, compute a PQL or PQA score, surface accounts for reps. Sharp at the job, another vendor to buy and wire in.
Broader analytics-plus-CRM platforms, where the audience segment and the messaging live in one place. Kixo sits here: it does product analytics (events, funnels, cohorts) alongside audience and CRM segmentation, so a "PQL" can be an audience segment defined from the same events you already track, with campaign tooling attached. The trade-off is scope. An all-in-one won't match a specialist's depth on scoring sophistication, and you're consolidating rather than best-of-breeding.
Whatever you pick, the mechanics don't change. Define the qualifying behaviors, weight them, gate on fit, roll individuals up to accounts if you sell to teams, and validate the threshold against real conversions. The tool is where the score lives. It is not what makes the score correct. That part's still on you, and it's worth getting right before you buy anything, because a PQL model is only as good as the definition of value underneath it. If you haven't nailed down what "value delivered" means for your product, that's the same discipline behind a good north-star metric tree, and it's the work to do first.
Common questions
Is a PQL the same as a free-trial signup? No. A signup is the start of the trial, not a qualified lead. The PQL is the subset of trialists who actually reached value and behaved like buyers. Treating every signup as a PQL is how sales teams learn to ignore the label.
How many signals should a PQL definition use? Fewer than you'd think. A handful of high-signal behaviors beats a sprawling model. Start with the one action that best correlates with conversion, add breadth and recency, and only expand when the data asks you to.
Do I need PQAs if I already have PQLs? Only if you sell to teams. Single-seat products live fine on PQLs alone. Multi-seat, per-account contracts need the account roll-up, or you'll misread a whole department's adoption as a few disconnected individuals.
Where should the PQL threshold come from? From your own conversion data, not a benchmark. Benchmarks like the 25-30% figures tell you the prize is real. They can't tell you where your line sits, because that depends on your product, price, and buyer.