Signal Scoring: How to Prioritize Buying Signals So You Work the Best Leads First

If you are doing signal-based selling right, you have the opposite of the old problem. Not "no leads," but too many. Once you monitor hiring, funding, job changes, competitor engagement, and tech stack changes across a market, and filter them against your ICP, you still end up with more qualified accounts than any rep can work in a day. In one of our own runs, 4,774 raw signals became 341 ICP-qualified leads. That is a good filter. It is not a call list. A rep cannot work 341 accounts today, so the real question is which ones first.

That question is signal scoring, and most content about it stops at "assign points and prioritize," which is not a method. This is the actual rubric: points by signal type, a recency multiplier, and stacking, run end to end until 341 becomes a top 30 a rep works this morning.

ICP fit is a filter, not a priority order

The most common mistake is treating ICP fit as the ranking. Fit is close to binary: a company either sells into your buying committee, is the right size, and is in the right place, or it does not. That gets you from 4,774 to 341. Every one of those 341 fits. Fit cannot rank them, because they all passed the same test.

Scoring is what happens after the filter. It ranks the qualified pool by how much is happening and how recently, so the rep spends the morning on the ten accounts most likely to reply today, not the ten that happened to load first.

Keep the two steps separate. Filtering answers "should this be on the list at all." Scoring answers "where on the list." Conflating them is why so many scoring models quietly re-score fit and never actually rank intent.

The rubric: points by signal type

Not all signals predict a deal equally. Weight them by how close they sit to a real buying decision. A starting rubric, meant to be tuned against your own data:

→ Job change into an ICP role (a new decision-maker in the seat): 30 → Funding round, Series A to C: 25 → Hiring for an ICP-adjacent role (building the function you sell into): 20 → Competitor engagement (following or engaging a competitor's presence): 15 → Tech stack change (adopt, drop, or switch in your category): 10

The logic: a new decision-maker with a mandate and fresh budget is the closest thing to a buyer actively deciding, so it tops the scale. Funding proves budget. Hiring proves the function is being built. Competitor engagement and tech changes are real but softer and earlier. These weights are a defensible starting point, not gospel. The point is that they are explicit and consistent, so every account is scored the same way.

The recency multiplier

A signal's value decays. A funding round from last week and one from eight months ago are not the same signal, and a model that treats them the same will send reps at cold accounts. Multiply each signal's points by how fresh it is:

→ 0 to 14 days old: times 1.0 → 15 to 30 days: times 0.5 → 31 to 60 days: times 0.25 → Past 60 days: times 0 (it is history, not a signal)

So a Series B (25 points) that closed nine days ago is worth the full 25. The same round at 40 days old is worth about 6. Recency is not a tiebreaker, it is a core term, because acting fast is most of what makes signal-based selling work.

Stacking: sum the account, do not pick one signal

Score the account, not the individual signal. If a company has a new VP of Sales (30), closed a Series B three weeks ago (25 times 0.5 = 12.5), and is hiring SDRs (20), its score is not 30. It is 62.5, and it should sit far above an account with a single strong signal.

This is the whole case for stacking in one number. One signal is a reason to look. Three fresh signals pointing the same way is a priority account, and the summed score is what surfaces it to the top without anyone eyeballing the list.

A worked example: 341 into a call list

Run the 341 through the rubric and the shape of the day changes.

A handful of accounts score above 50. These have multiple fresh signals stacked. They get deep, researched, multi-touch outreach, because the odds justify the time.

A larger band scores roughly 20 to 50. One strong fresh signal, or a couple of softer ones. These get a solid, signal-referenced message, worked in volume.

The long tail scores under 20. Single soft or aging signals. These are not worked today. They sit in nurture and either pick up a second signal later, which re-scores them up, or decay to zero and fall off.

So 341 qualified leads become maybe 30 that a rep works this morning, 80 to 100 in the active queue, and a tail that waits for a reason to move up. That is the job of scoring: not to shorten the list arbitrarily, but to order it so effort follows odds.

Calibrate against closed-won

The starting weights are a hypothesis. Your pipeline is the test. Every quarter, look at what actually closed and work backward: which signals preceded real deals, and how fresh were they at first touch. If funding turns out to predict your closes better than a job change, raise funding's weight. If competitor engagement produced nothing, cut it.

A scoring model that never gets recalibrated is just someone's opening guess, frozen. The teams that win with signals treat the rubric as a living thing they tune against closed-won, not a config they set once.

Common mistakes

• Using ICP fit as the ranking. Fit builds the list. Scoring orders it. They are different jobs.
• Ignoring recency. A signal without a decay term sends reps at accounts that were hot two months ago.
• Scoring signals instead of accounts. Sum the stack per account, or you miss the accounts that matter most.
• Setting weights once. Untuned weights are a guess. Recalibrate against what actually closes.
• Over-engineering it. A five-line rubric you apply consistently beats a forty-variable model no one trusts. Start simple, tune with evidence.

Frequently asked questions

What is the difference between lead scoring and signal scoring? Traditional lead scoring mostly ranks fit and demographic attributes (title, company size, form fills). Signal scoring ranks intent and timing: how many buying signals an account has and how fresh they are. Fit says who could buy. Signal scoring says who is likely deciding now.

How do you calculate a signal score? Assign points to each signal type by how close it sits to a buying decision, multiply each by a recency factor so older signals count for less, and sum the fresh signals on each account. The account's total is its priority score.

How stale can a buying signal be before it is worthless? As a rule of thumb, most signals lose the bulk of their value within 30 days and are close to noise past 60. Treat any specific decay number as directional and tune it against your own data, since different signal types age at different rates.

How many signals should stack before you reach out? One fresh, strong signal (a new decision-maker, a recent raise) is enough to act. The point of stacking is not a minimum threshold, it is prioritization: more fresh signals means a higher score and earlier, deeper outreach.

How much weight should ICP fit get versus signals? Fit should be a gate, not a score. Ideally an account is already ICP-qualified before it enters the scoring model, so the score reflects intent and timing alone. Mixing fit back into the score usually just re-ranks fit and buries the timing.

How often should you recalibrate the model? Quarterly is a reasonable default, or whenever you have enough closed-won to see which signals actually preceded deals. The weights are a hypothesis; closed-won is the test.

Order the list, then work it

Signal-based selling does not fail for lack of signals. It fails when a rep stares at hundreds of qualified accounts and works them in whatever order they loaded. A simple rubric fixes that: weight signals by how close they sit to a decision, decay them by age, sum them per account, and tune the weights against what closes. Do that, and 341 qualified leads stop being a firehose and become a ranked list where the best thing to do next is always at the top.