A user who has crossed a usage threshold chosen because it predicts buying — qualified by what they do, not by what they downloaded.
In practice
A PQL qualifies on a behaviour inside the product rather than on a company size or a job title — that's the whole difference from an MQL, a marketing-qualified lead, and the reason it predicts better: the person has already done what a demo would have tried to convince them to do. The part most definitions miss is that a good threshold has to hold on two axes, not one. Lift — how much better the qualified group converts — is the one everybody measures; list size is the one everybody forgets, and two thresholds with identical lift can produce, respectively, about ten high-density contacts a day and a mailing list in disguise. Finally, a threshold expires: when the product changes, the behaviour that used to be a strong signal becomes routine, and re-deriving it quarterly is ordinary maintenance.
The formula
PQL lift = (conversion rate of users above the threshold) ÷ (conversion rate of users below it) — and the threshold is only useful if the list it produces is one a human can work
- Threshold
A behaviour inside the product — invited a teammate, imported real data, hit a plan limit — not a company size or a job title. That is the whole difference from an MQL, and it is the reason the number means anything.
- Lift
How much better the qualified group converts than the unqualified one. A threshold with a lift near 1 is not qualifying anybody, it is drawing a line through the middle of your users.
- List size
How many users cross the threshold in a period. The forgotten half of the definition: a threshold can have a beautiful lift and still qualify more people than anyone could ever contact.
Both halves have to hold. Optimising for lift alone produces thresholds so rare they qualify four people a quarter; optimising for volume produces a list indistinguishable from "everyone who signed up".
Worked example
Two thresholds, the same lift, one of them useless
- 5,000 free sign-ups in a quarter, 300 of whom convert to paid: a 6% baseline.
- Threshold A, "invited a teammate": 900 users qualify and 135 convert (15%); the other 4,100 produce 165 conversions (4.0%). Lift = 15 ÷ 4.0 ≈ 3.7×.
- Threshold B, "logged in three times": 3,200 qualify and 260 convert (8.1%); the other 1,800 produce 40 conversions (2.2%). Lift = 8.1 ÷ 2.2 ≈ 3.7× — identical.
- Same lift, and only one of them is a list. A is 900 people a quarter converting at 15%: about ten contacts a day, each with better-than-even odds of being worth the call. B is 3,200 people at 8%, which is a mailing list with extra steps.
Lift alone would have rated these two thresholds identically. The second axis — how many people it qualifies, and at what density — is what separates a signal a team can act on from one it will quietly stop opening after three weeks.
Orders of magnitude
- The claim you will meet everywhere is that PQLs convert several times better than MQLs. It is true and the comparison is rigged: a PQL has used the product, an MQL downloaded a PDF. Quoting the ratio proves nothing about your thresholds, because it would hold for almost any behavioural threshold at all.
- The only benchmark worth holding is your own baseline conversion rate, measured before any threshold exists. Without it there is no lift to compute, which is why this is the one number to get first.
- A threshold that stops working is normal rather than alarming: it means the product changed and the behaviour that used to be a strong signal became routine. Re-deriving it quarterly is maintenance, not a sign the first one was wrong.
How to improve it
- Derive the threshold from data, never from a workshop. Take the users who converted, look at what they did in their first two weeks that the others did not, and test each candidate behaviour for lift and list size.
- Check both axes before adopting one, as the example above shows. Write down the lift AND the quarterly count next to every candidate.
- Keep it to a single behaviour a person can read in one sentence. A score built from seven weighted signals cannot be argued with, cannot be debugged, and gets ignored the first time it is wrong.
- Decide what happens when someone crosses it, before you start counting. A PQL nobody contacts and no in-app offer greets is a database column, not a growth mechanism.
In the Tour
Do you know what percentage of users reach that moment?
- Yes, tracked precisely20 pts
- A rough sense of it7 pts
- No visibility at all0 pts
The Tour asks whether you know the percentage of users who reach your activation moment, for 20 points. A PQL threshold is that same measurement pointed at a commercial decision: both ask which behaviour separates the users who got it from the ones who did not, and a team that can answer 20 already owns most of the work. The order matters — defining a PQL before knowing your activation event usually produces a threshold chosen for how easy it is to query rather than for what it predicts.
Questions people ask
PQL or MQL?
They answer different questions and can coexist. An MQL says someone showed interest in the category; a PQL says someone got value from your product. If you have a free tier or a trial, the second is available to you and is far more predictive — the user has already done the thing a demo would have tried to convince them to do.
How do I choose the threshold?
Backwards, from your converted users. List what they did early that non-converters did not, then compute lift and list size for each candidate. Three or four candidates is usually enough to find one that is both predictive and workable, and the exercise takes an afternoon with a database.
Do I need a sales team to use PQLs?
No. The threshold can trigger an in-app offer, an email, or a change in what the product shows — all of which scale without anyone picking up a phone. A sales team lets you use a smaller, higher-lift threshold, but the mechanism works without one.
How many PQLs should we have?
As many as whoever acts on them can actually handle, which is a capacity question rather than a marketing one. Set the threshold so the weekly list matches the follow-up you can genuinely do; a threshold producing three times that capacity is producing a backlog and a false sense of pipeline.