Traditional lead scoring rewards form fills and email opens: a whitepaper download is worth 5 points, a pricing page visit is worth 10, a demo request is worth 50. That model was built for a world where the sales conversation is the first meaningful interaction a prospect has with your product. Product-led growth teams don't have that problem, or that luxury. The prospect is often already inside the product, generating far better signal than any marketing form ever could, and a scoring model that ignores usage data is leaving the best data on the table.
Why marketing-activity scoring fails PLG teams
In a PLG motion, the highest-intent action a user can take usually isn't visiting a pricing page, it's inviting a teammate, connecting an integration, or hitting a usage limit on the free tier. A model built around content downloads and email engagement will systematically under-rank exactly the accounts most ready to buy, because those accounts are busy using the product instead of clicking around your marketing site. Meanwhile, a single enthusiastic marketer who reads six blog posts can rack up a "hot lead" score without a single engineer at their company having touched the product.
Signals that actually predict PLG conversion
The scoring inputs that correlate with real conversion in a self-serve or freemium software product tend to cluster into a few categories:
- Activation depth: did the account get past initial setup to a genuine "aha" moment? For most products this is a specific, definable event: first successful API call, first dashboard built, first integration connected. This single signal usually predicts conversion better than any demographic or firmographic data point.
- Team expansion: an account where 1 person signed up and stayed alone for three weeks looks very different from one where 6 teammates were invited in the first 10 days. Multi-seat activity inside a single account is one of the strongest indicators of organizational buy-in.
- Usage frequency and trend: daily active usage climbing over two weeks says something very different from a burst of activity on day one that flatlines. Trend direction matters more than a single usage snapshot.
- Limit-hitting behavior: an account bumping against a free-tier or trial limit (seats, API calls, storage, integrations) is telling you directly what they need next. This is often the single cleanest upgrade-intent signal available and it's completely invisible to a scoring model based only on marketing activity.
- Technical integration signals: for a dev-tool or API-first product, connecting GitHub, setting up a webhook, or shipping to a production environment indicates the kind of committed technical investment that rarely gets abandoned.
A practical scoring model
Rather than a single blended score, most PLG teams get better results from two separate scores that get combined into a routing decision: a fit score (does this account match your ideal customer profile, based on company size, industry, and stated use case) and an intent score (is this specific account showing behavior that predicts conversion, based on the usage signals above). A high-fit, high-intent account should route to a sales rep same-day. A high-fit, low-intent account is a good target for a lifecycle email nudge, not a cold call. A low-fit, high-intent account (a student or hobbyist power-user, for instance) usually isn't worth sales time regardless of how engaged they are.
Building this without a data science team
You don't need a machine learning model to get most of the value here. A rules-based score using 4 to 6 weighted signals, seat count trend, feature adoption of your 2 to 3 core workflows, integration connections, and limit-proximity, will outperform a purely demographic model in almost every PLG context. The key infrastructure requirement is getting usage events out of your product and into the same place your sales team works, so the score updates automatically as behavior changes instead of being a stale snapshot from the last time someone ran a report.
Softmatica ingests usage events from Segment, Amplitude, or a data warehouse connection directly onto the account record, so fit and intent scoring can be built as automation rules that update live rather than a monthly export job. When an account crosses your intent threshold, routing to a rep (or a specific rep, based on account size or territory) can happen automatically, with the usage context that triggered the score already attached to the account, so the rep's first outreach can reference something real: "Noticed your team just hit your seat limit on the Fathom Systems workspace," lands very differently than a cold "checking in."