Account scoring ranks a company by firmographic fit, headcount, funding stage, tech stack, industry, to decide whether it belongs in a pipeline at all. That's a different question than lead scoring, which ranks a contact's behavior inside an account already judged worth pursuing.
Account scoring evaluates the company, not the person. Headcount, funding stage, tech stack, industry: these are mostly firmographic, and they barely move month to month. Lead scoring asks something else entirely: did this specific contact open three emails this week, visit the pricing page twice, request a demo. That's mostly behavioral, and it decays fast. Conflating the two into one number is the most common way a scoring model breaks, because it lets a bad-fit account with a lot of activity outscore a great-fit account that's quiet.
The stronger version of account scoring runs as a yes-or-no gate before any point math starts: does this company clear the two or three firmographic facts that actually predict a closed-won deal, not the ten a team wishes were true. Fail the gate and the account doesn't get scored at all, no matter how many demo requests come in later. That's the structural fix for the blending problem, and it's also why account scoring and lead scoring need to live on different tables. Score a person on a company-level table and every contact from the same account gets double-counted. Score a company using person-level activity and one enthusiastic intern's browsing can spike the whole account's number.
A lead score answers who to call first inside an account. An account score answers whether the account is worth calling at all. Both usually run, joined, in the same pipeline, but they're answering different questions and deserve different refresh cadences: firmographic fields hold their value for months, so an account score doesn't need re-checking nearly as often as the behavioral fields feeding a lead score.
Run the account-level gate first, before any lead-level behavior gets scored, so enrichment and scoring spend never goes toward a company that was never going to qualify. Keep the fit criteria to two or three facts that actually separate closed-won from closed-lost in a team's own data, not a longer list that feels more thorough.
Teams build one blended score instead of two separate ones, then wonder why an active but wrong-fit account keeps outranking a quiet good one. The other mistake is running the model on the wrong table: scoring a company-level fact against a person-level row structure, or the reverse, which either double-counts contacts or lets one person's activity stand in for the whole account.
Tell us how your motion runs today. We'll show you what we'd engineer.
Contact us