Buying signals B2B: who decides how much they count

Your account model just returned a 78. A sales leader asks the only question that matters: “why is this a 78?” And nobody in the room can answer. Not because the team is careless. Because the number came from weights someone chose, for a reason nobody wrote down, and that someone left the company two quarters ago. Every team has this moment. The score looked solid in the dashboard; it just couldn’t survive a single follow-up question. 

What decides who gets the lead isn’t the signal itself. It’s how much the signal was allowed to count, and that weighting decision is the one thing most teams have never examined on purpose. This article is about who makes that call, how, and how to do it deliberately. 

A score is an argument, not a measurement 

A measurement has a unit and a referent. Ten kilograms refers to a mass; the kilogram is a fixed unit anyone can check. A composite score has neither. A 78 isn’t 78 of anything. It’s a compression of several kinds of evidence (a webinar attendance, three pricing-page visits, a topic surge from a third-party feed) squeezed into one number. And the compression is a choice. Someone decided the webinar was worth more than the pricing pages, or the reverse. 

That makes a score an argument, not a measurement. It says: given what we believe about how our buyers behave, this account looks more ready than that one. A measurement you trust; an argument you interrogate. The trouble starts when a team treats the second like the first. 

Watch what happens when the reasoning goes invisible. In Salesforce’s 7th edition State of Sales report, 46% of sales professionals using AI agents say data quality issues are hurting their sales efforts. A rep who doesn’t trust the number stops using it, quietly. They work their own list, the one in their head, and nobody logs that decision. The score keeps producing 78s. The field has already moved on. 

The three ways weights get set, and what each optimizes for 

There are only three ways the weights in a scoring model get set. Each is right under certain conditions. Naming the conditions is the whole job. 

1. Vendor default is fast, and often a reasonable start. But a default encodes the vendor’s data strengths, not your buying reality. A provider with deep third-party topic coverage will weight topic surges heavily, because that’s what it can see best. That’s not a flaw; it’s the tool being honest about where its data is strong. The risk is mistaking “what the vendor measures well” for “what predicts a deal for us.” Use it when you’re new to signal scoring. Replace it the moment you have enough closed deals to argue from. 
2. Regression against closed-won is more defensible. You fit weights to deals you actually closed, so the model reflects your business rather than someone’s catalog. The catch: regression optimizes for the past. It learns the buyer you had. The moment the product changes or you enter a new segment, those weights describe a world that no longer exists, and they drift without announcing it. Use it when your motion is stable and you have real volume. Question it right after a pivot, a launch, or a move upmarket. 
3. Explicit business judgment is the slowest and the most defensible. Someone argues, signal by signal, how much each should count and why. It’s the only method that survives a real argument with sales, because the rationale is on the table. The cost is time and the discomfort of committing to a reason in writing. For anything that routes pipeline, it’s usually worth the effort. 

Most teams use a blend, and should. The problem isn’t picking one. The problem is when nobody can say which one produced the number in front of you. 

Why data quality beats model sophistication 

When we surveyed 1,500 B2B marketers globally, data quality and accuracy ranked as the #1 driver of campaign performance, chosen more often than creative, messaging, targeting, or technology. The people running the campaigns are telling you where the leverage is, and it isn’t the model. 

The consequence is uncomfortable. A sophisticated model over weak inputs doesn’t fail loudly. It produces confident wrong answers faster. Add more signals, tune more weights, and you get a more precise number built on the same shaky evidence. That’s precision the field will trust exactly until they check it against a deal they know. 

One case matters more than the rest. When the underlying signal can’t name a person, when all you have is “an account is researching,” no weighting scheme can recover the missing name. Accounts don’t buy. People do. A perfectly weighted account score still can’t tell a rep who to call. That’s a data problem, not a scoring problem, which is why some teams need contact-level data before the weighting conversation is worth having. 

What the unit of scoring should be 

The contact is the unit. The account is the roll-up. Not the reverse, and the arithmetic shows why it matters. 

Take two accounts, each scoring 60. In the first, that 60 comes almost entirely from one person: a single researcher hitting your pricing page at midnight, reading three comparison posts, downloading the technical brief. In the second, the same 60 is spread across six people in different roles: a director, two managers, someone in security, someone in finance, each doing a little. 

These are completely different situations. The first is one motivated individual who may or may not have a committee behind them. The second is a buying committee warming up across functions. One rep should call the researcher directly; the other should build a multi-thread play. An account-level model can’t tell these apart, because the information that decides the right next move (how the score is distributed across people) got compressed away before the rep ever saw it. 

Build from the contact up and that information survives. You can always roll contacts into an account view. You can’t recover the people once you’ve averaged them into a single number. 

The data backs this up. In our own campaigns, accounts engaged by two or more tactics surface 72× more buying-committee members than single-tactic accounts. That isn’t a rounding difference. It’s the gap between seeing one name and seeing the committee, and it only shows up when you score at the person level first. 

A method for setting weights on purpose 

Here’s a method you can run this quarter. It’s deliberately boring. The point isn’t cleverness; it’s that every number has a reason attached, and the reasons are written where sales can see them. 

1. Name the single decision this score serves. Routing, sequencing, spend allocation, or sales alerting: pick one. A score built to route leads to the right rep is a different score from one built to decide where to spend next quarter’s budget. A number that tries to serve all four serves none. Write down the one decision. 
2. List the behaviors that plausibly precede that decision. Not every behavior you can track, only the ones that actually come before the outcome you care about. For routing a sales-ready lead: demo request, pricing-page visits, repeat visits from the same contact, high-intent content downloads. 
3. Assign a weight to each, with a one-sentence written rationale. The rationale is the deliverable, not the weight. “Demo request = 40, because in the last two quarters it preceded a booked meeting more often than any other action.” If you can’t write the sentence, the weight isn’t ready to ship. 
4. Test the ranking against real deals. Pull a sample of closed-won and closed-lost from the last two quarters. Run them through your weights. Do the won deals rank above the lost ones? If your top-scoring accounts are full of deals you lost, the weights are wrong. Better to learn that against history than against next quarter’s pipeline. 
5. Fix a review cadence and an off-cycle trigger. Put a date on the calendar, quarterly is a reasonable default, and name the events that force an early review: a price change, a new product, a move into a new segment. Regression-derived weights drift silently, so the calendar is what catches the drift. 

A worked example, so “argued rationale” is concrete. Say you’re scoring for one decision: route to an account executive versus hold for nurture. 

Behavior	Weight	Rationale (one sentence)
Demo or pricing request	40	Directly precedes a booked meeting more than any other action last two quarters.
3+ pricing-page visits, same contact	25	Repeat pricing views from one person signal an individual evaluating, not browsing.
High-intent content download	15	Correlates with active evaluation but is noisier than pricing behavior.
Third-party topic surge	10	Real signal of category interest, but can’t name the person, so weighted low.
General site visit	10	Weak alone; matters mainly as corroboration alongside the above.

You may disagree with these numbers. That’s the point: you can disagree with them, because the reasons are on the page. Compare that to a 78 nobody can explain. The version you can argue with is the version sales will actually use. 

The discipline, not the tool 

If you can’t say in one sentence what your score is for, it isn’t ready to route anything. A number without a stated purpose can’t be checked, defended, or trusted, and an untrusted score is one the field quietly ignores while everyone upstream keeps admiring it. 

What most teams discover the first time they run this exercise is that the conversation itself is the value. The moment sales and marketing sit in a room and argue about whether a pricing-page visit should outweigh a topic surge, they’re building a shared definition of “ready” that no vendor model can give them. That shared definition is what makes a score usable. Interpretation outlasts prediction. 

Two things break this in practice, and they break differently. If your weighting is sound but the underlying signals can’t name a person, the fix is upstream: you need contact-level data from a source that can name the person before the weighting conversation is worth having. 

If your signals are solid but you’re only reaching a fraction of the committee, the fix is reach. Programmatic Nurture is built to surface more of the people inside an account so your weights have something to work with. 

Talk to Digitalzone about the signals feeding your score before you tune another weight. 

Frequently asked questions 

What are B2B buying signals? B2B buying signals are the actions and behaviors that suggest an account or a person is moving toward a purchase: pricing-page visits, demo requests, content downloads, event attendance, third-party topic surges. No single signal tells the whole story; the useful information comes from reading them together over time and knowing which ones actually precede a deal for your business. 

What is a lead scoring model? A lead scoring model assigns a number to a lead or account based on the signals it has produced, so teams can decide who to route, sequence, or reach out to first. The number depends less on the raw signals than on the weights applied to them, which is why how the weights get set matters more than which signals you collect. 

Should you score at the contact level or the account level? Score at the contact level and roll up to the account, not the reverse. A single account score hides how intent is distributed across people: one obsessive researcher and a six-person committee can produce the same number while demanding completely different sales plays. Building from the contact up keeps that information available; starting from the account averages it away. 

How often should you review signal weights? Set a fixed cadence, quarterly is a reasonable default, plus triggers that force an off-cycle review, such as a price change, a new product, or a move into a new segment. Weights fit against past deals drift as the business changes, and a review schedule is what catches the drift before it misroutes pipeline.