PQL scoring and routing

Most PQL models fail for the same reason. Somebody picks the events that feel important in a meeting, sales gets flooded with accounts that were never going to buy, and within a quarter the flag is ignored. The fix isn’t a better tool. It’s deriving the definition from conversion data and then being ruthless about how few accounts qualify.

This guide covers finding the predictive events, building the threshold, routing by value, and measuring the false positives that quietly destroy trust.

Which usage events actually predict paid conversion

Find them by comparing cohorts, not by asking the product team. Export every account that converted to paid in the last two quarters and every account that signed up in the same window and didn’t. Then count how often each cohort performed each tracked event in its first 30 days.

You’re looking for lift, not volume. An event that 90 percent of both groups performed tells you nothing, even though it’ll be the most common event in your dataset. Keep events where the converted group is three times or more likely to have done the thing.

That table is illustrative structure, not measured data, but the shape is what you’ll see. Three or four events survive. The rest are noise dressed up as engagement.

Four signal patterns that recur across B2B SaaS

The specific events differ by product, the patterns don’t. Four show up again and again.

Multiple seats invited. One person exploring is evaluation. Three people in the same account is a project with internal momentum, which is why Slack, Figma and Notion all treat seat expansion as the central growth motion.

An integration or data source connected. This is the highest-effort action most users take, and effort is intent. Someone who has connected their CRM has made a small internal commitment and told a colleague about you.

A usage limit approached. Hitting 70 to 90 percent of a plan’s cap is the cleanest commercial signal in the whole model, because the product itself is creating the buying conversation.

A high value action repeated across sessions. Once is a trial. Four times across three weeks is a workflow. Repetition separates evaluation from habit better than any single event.

Building the definition: thresholds, account rollup, recency

Three design decisions turn events into a usable flag, and teams get the middle one wrong most often.

Threshold. Don’t build a 100 point weighted score on your first attempt. Start with a rule: any two of the four qualifying events within the window. Weighted scores are harder to explain to sales, and a rule sales can recite is worth more than a model they can’t audit.

Account rollup. Score the account, not the user. Two users each doing one qualifying action is a stronger signal than one user doing both, and a user-level score misses that entirely. Roll individual events up to the company domain and deduplicate free email domains carefully, because gmail.com will otherwise become your largest account.

Recency. Put a 7 to 14 day window on it. An integration connected 40 days ago with no activity since is not an intent signal, it’s a historical fact. Without decay, your PQL list becomes a slowly growing pile of accounts that once looked interesting.

Routing: who gets a human, who gets an email, who gets nothing

Route on expected account value, not on score. A perfect PQL at a 4-person company with a $600 expected annual value should never reach a salesperson, because a single 30 minute call plus follow-up costs more than the account is worth in year one.

Lane	Trigger	Action	Response time
Self serve only	PQL, expected ACV under $3K	In-product prompt plus lifecycle email, no human	Automated
Automated nudge	PQL, expected ACV $3K to $15K	Personalised sequence from a named rep, calendar link	Within 24 hours
Human outreach	PQL, expected ACV above $15K	AE call or personal email, plus account research	Within 4 business hours
Hold	High usage, poor firmographic fit	Stay in product, re-evaluate at seat 5	None

The hold lane matters. Plenty of accounts use the product heavily and will never be worth a sales conversation: agencies, students, single-person consultancies. Flagging them and leaving them alone protects both the sales team’s time and the user’s experience.

Speed matters most in the top lane. A PQL is a moment, and the moment passes. Four hours is a reasonable service level; next-day follow-up on a limit-approach signal arrives after the person has either upgraded or moved on. Tooling for the routing itself is usually your CRM plus a workflow layer, and the mechanics are covered further in lead scoring for B2B SaaS.

The handoff: what the first message says and when it lands

The worst PQL outreach opens with “I noticed you’ve been using our product”. It’s creepy and it’s generic at the same time, which is impressive. The good version references the specific thing the account did, and offers something that only makes sense given that action.

If the trigger was a seat invite, the message is about getting the rest of the team set up. If it was an integration connect, it’s about the configuration step people usually get wrong. If it was a limit approach, it’s about which plan fits their volume and whether annual billing saves them anything.

Timing follows the trigger too. Limit-approach signals are urgent. Seat invites are urgent within a day. Integration connects can wait until the user has actually used the integration once, because reaching out mid-setup interrupts the exact behaviour you want.

There’s a full library of trigger-specific messages in the product qualified lead email plays, and a deeper model-building walkthrough in building a PQL scoring model.

Measurement: PQL to paid, and the cost of being wrong

Two numbers, and the second one is the one that saves the program.

PQL to paid conversion rate is the headline. Whatever it is, it must be meaningfully higher than your baseline signup-to-paid rate. If your signups convert at 4 percent and your PQLs convert at 6 percent, you have not built a lead score, you have built a slightly biased sample.

False positive cost is the number nobody tracks. Take the PQLs that went nowhere, multiply by the hours sales spent on each, and price it at loaded cost. A model producing 400 PQLs a month at 25 percent conversion is burning 300 accounts of sales attention. At 40 minutes each that’s 200 hours a month, which is more than a full-time person.

Those bands are practitioner ranges, not measured benchmarks. Compare them against your own history before treating them as targets.

The position worth defending: PQLs should be rare

If you take one thing from this page, take this. PQLs should be rarer than your MQLs, and considerably rarer than your signups. The instinct in every organisation runs the other way, because a bigger number looks like a better quarter and because marketing gets measured on volume.

Resist it. The moment sales works fifty PQLs and closes two, the flag loses credibility, and credibility is the only thing that makes the routing work. A tight definition that surfaces 30 excellent accounts a month beats a loose one that surfaces 400 mediocre ones, even though the loose one looks better on a slide.

The distinction between the two flags is worth being precise about, which is why it’s covered separately in MQL vs PQL in B2B SaaS and in MQL vs PQL generation. For definitions, see product qualified lead, the related PQL metrics entry and PQL generation.

What to do this week

Pull the cohort comparison. Don’t design anything until you’ve seen the lift table, because the events that survive will not be the ones you’d have guessed and that surprise is the point of the exercise.

Then write the rule in one sentence, backtest it against last quarter, and check what share of signups qualify. If it’s over 15 percent, tighten it before anyone in sales sees the output. The rest of the funnel context sits in the SaaS lead generation hub.