How to Build an Intent Data Scoring Model for Outbound Prioritization

Your team has intent data. Nobody knows who to call first.

That's the real problem with plenty of "hot account" lists. A dashboard turns green. An alert fires. A rep is supposed to know what that means.

Plenty don't. They dial, get nothing, and move on. Do that for a few weeks and reps stop trusting the alerts completely.

The data isn't the problem. The lack of a system is.

An intent data scoring model fixes this. It weighs every signal you already collect against how close it sits to an actual purchase, then turns it into one number a rep can act on without guessing.

This guide breaks down exactly how to build that model: which signals to weight, how to stop old signals from outranking new ones, and how to make sure your top-ranked account is one your team can reach, not just rank.

What Counts as "Intent" in a B2B Scoring Model

Intent data usually gets split into two buckets. Third-party intent tracks what a company researches across the open web. Think co-op networks like Bombora, or comparison activity on G2. First-party intent tracks what someone does on your own site: pricing page visits, demo requests, content downloads.

Neither bucket is trustworthy alone. The same Forrester-backed research cited above found that median precision for topic-based intent signals sits at just 0.51 across 47 deployments studied. 62 percent of buyers report that fewer than 70 percent of flagged accounts show any matching CRM activity within 30 days. In plain terms: roughly half the time, a raw intent signal points at nothing real.

A scoring model exists to fix that. It combines signal types and weights them by how close they sit to an actual purchase decision. The output is one number a rep can act on, no data science degree required. Here's how that model gets built, layer by layer.

Step 1: Build the Four Layers of an Intent Score

Every account gets scored across four layers. Score each one independently first, then combine them into a single composite number. Skip a layer and the model breaks in a specific, predictable way.

1. Fit Score: Does the Account Match Your ICP

Fit measures whether an account belongs in your target market at all, independent of any behavior it's shown. This is your ideal customer profile translated into points: industry, employee count, revenue band, and geography.

Layer in firmographic data for company-level context. Add technographic segmentation to check whether an account's current tech stack fits, or if it's a clean displacement target.

Fit doesn't decay. A company's headcount and industry don't reset every month. Once you score it, it stays valid until something structural changes, like an acquisition or a pivot.

What to do: treat Fit as a gate, not just a score. Set a floor (we'll use 40 percent of the max in the example below). Accounts under that floor never reach a rep's queue, no matter how loud their intent signal gets.

2. Engagement Score: What Prospects Do On Your Own Site

Fit tells you whether an account belongs on the list at all. Engagement tells you what they're doing once they're on it. It's first-party behavior you already own: pricing page visits, demo requests, webinar attendance, and content downloads. Weight it by proximity to purchase, not by volume. One pricing page visit should outscore ten blog reads. Ten blog reads signal curiosity. One pricing visit signals someone doing math on a budget.

A practical note that plenty of guides skip: email open data has gotten noisy. Apple Mail's privacy relay pre-fetches images now, which inflates open counts without a human reading anything. Weight opens low. Weight clicks and replies far higher.

3. Intent Score: What They Research Everywhere Else

Engagement only sees what happens on your own site. Intent picks up everywhere else an account is researching, and it's usually the largest single input in the model. Bombora's topic-level intent network tracks which companies are researching which topics across a large co-op of B2B publisher sites. It's typically the first data source teams plug into a scoring model.

Layer in G2 buyer intent where it applies. G2 signals fire later in the buying process than a typical Bombora surge. Someone has to be actively comparing named vendors on a review site to trigger one. Weight G2 activity higher than a generic topic spike.

Two signal types belong here that plenty of scoring frameworks leave out entirely:

• Buying trigger events like funding rounds, new leadership hires, and expansion announcements.
• Job changes at target accounts. A former champion landing in a new VP seat is one of the highest-converting signals in B2B. It rarely has a dedicated field in a typical CRM.

4. Reach Score: Can You Contact the Buying Group

Fit, Engagement, and Intent tell you whether an account is worth pursuing. Reach tells you whether you can pursue it at all, and it's the layer nearly every intent data guide skips. It's also the one that decides whether your model produces pipeline. Or just a well-organized list of accounts nobody can get on the phone.

A perfect intent score is worthless if nobody on the buying group has a verified email or mobile number on file. I've watched teams build a beautifully weighted composite score. They rank their top 50 accounts, hand the list to SDRs, and then watch half of it sit untouched for two weeks. The reason was never the score. It was a generic info@ address and a phone number that rang out to an empty desk.

Score Reach on verified email and mobile coverage for the economic buyer and at least one technical evaluator. If coverage is thin, keep the account flagged as high-intent, but route it to an enrichment queue before it goes to a dialer.

This is also where B2B contact data decay quietly wrecks scoring models that otherwise look great on paper. HubSpot's Database Decay Simulation, built on MarketingSherpa research, puts B2B contact decay at roughly 2.1 percent a month. That compounds to about 22.5 percent a year. A model built on top of a decaying database will eventually rank unreachable accounts as your top priority. Nobody notices until pipeline dries up.

With all four layers defined, the next question is how many points each signal is worth.

Step 2: Assign Point Values and Weights

Once the four layers exist, resist the urge to assign points purely by gut feel. Start there if you have to, then correct it with real data within a quarter. Here's a starting 100-point framework:

This is broadly consistent with how traditional lead scoring allocates points, extended to account-level buying groups instead of single contacts. Bombora itself describes a version of this as FIRE scoring, short for Fit, Intent, Relationship, Engagement, applied at the account level. That's an established direction in the market. Reach is the layer we'd add to it.

Don't leave these weights untouched. Once you have 90 days of pipeline data, pull your closed-won accounts and check which signals they had before the deal closed. If your highest-weighted signal shows up in only 20 percent of your wins, the weight is wrong. This single quarterly check is worth more than any other tuning you'll do to the model.

Points alone aren't the full picture, though. A signal's age matters just as much as its weight.

Step 3: Apply Time Decay to Engagement and Intent

A pricing page visit from yesterday and one from two months ago are not the same signal. Plenty of spreadsheets score them identically anyway. The same 2024 B2B Buying Study found that 47 percent of intent records go stale within just 14 days. Buying committee changes and shifting project priorities are the biggest drivers. Decay isn't an optional refinement. It's the difference between a score that reflects reality and one that reflects last month.

Apply decay to Engagement and Intent only. Fit and Reach move on a slower clock, since company size and contact coverage don't shift week to week the way research behavior does.

Set your decay windows against your actual sales cycle, not a number copied from a blog post. A team closing deals in two weeks should decay signals hard within the first week. A team running a nine-month enterprise cycle can let intent stay warm for a month or more before it meaningfully fades.

Decay handles when a signal matters. The next question is who it belongs to.

Step 4: Score the Buying Group, Not One Contact

McKinsey's B2B Pulse research found that B2B buyers now use an average of ten interaction channels across a single purchase. That's up from five just a few years earlier. The buying committee behind that purchase is bigger and more scattered than it used to be. That alone makes single-contact scoring close to useless.

If your model only scores the one person who clicked an email, you're measuring one voice. Five to ten other people in that room also get a vote. Mapping the full buying group means rolling up signals from every identified stakeholder, not just whichever one engaged the hardest. That produces an account-level score that reflects organizational readiness rather than one person's browsing habits.

In practice, this means your CRM object for scoring lives at the account level, with contact-level activity feeding into it. Three different people at the same account researching the same topic in the same week is a far stronger signal. One person doing it three times isn't. Scoring tools default to contact-level tracking more often than not. You'll likely need to build this rollup logic yourself, or push your vendor for it.

Once every layer is combined into one account-level number, that number needs somewhere to go.

Step 5: Set Routing Thresholds and Outbound Plays

A score with no threshold attached is a number nobody acts on. Break the composite score into tiers, and give each tier a specific play instead of a vague "prioritize this" instruction.

1. Tier 1 (75+ composite, strong Reach): Immediate AE or senior SDR outreach, multithreaded across the buying group, same-day response SLA.
2. Tier 2 (50 to 74, or high score with weak Reach): Route to an enrichment queue first if Reach is thin. Otherwise, standard SDR sequence within 48 hours.
3. Tier 3 (30 to 49): Marketing nurture with content aligned to the specific intent topic, re-scored weekly.
4. Below 30: No active outreach. It sits in the general outbound prospecting pool for standard cadence treatment.

For any Tier 1 account with a wide buying group already mapped, this is where an account-based marketing motion earns its budget. Coordinated ads, executive outreach, and SDR sequencing hit the same account in the same week. That beats three teams working it independently.

Write these thresholds down somewhere reps can find in ten seconds, not buried three folders deep in a RevOps wiki. If a rep gets a Tier 1 account and doesn't know what that means, the tier system doesn't exist in any way that matters.

Even a well-built tier system won't stay accurate on its own. It needs a regular check-up.

Step 6: Review and Re-Weight the Model Every Month

Markets shift, and ICPs evolve. A topic that correlated with buying intent last quarter can turn into background noise once everyone's writing about it. Set a recurring monthly review where someone checks which signals correlated with closed-won deals against which ones just generated activity.

Track this against RevOps KPIs built for the purpose. Watch meeting rate from Tier 1 versus Tier 3 accounts, and pipeline value generated per tier. Also watch false-positive rate: Tier 1 accounts that went completely cold within two weeks of outreach. If that false-positive rate climbs past roughly 20 to 25 percent, the weights need retuning before you scale anything further.

Sales adoption data backs up why this matters. The same benchmark catalog cited earlier found something blunt. Only 41 percent of reps actively work intent-flagged account lists within the first 90 days of rollout. That figure jumps to 67 percent when scores surface natively inside the CRM instead of living in a separate dashboard. A model reps have to leave their workflow to check is a model the majority of them will quietly ignore.

This review cadence also protects pipeline generation numbers from a slow decay nobody notices in real time. Teams tend to catch a stale model three months late, when a VP asks why intent-scored accounts stopped outperforming the general pool.

A handful of recurring mistakes account for almost everything that breaks a scoring model, so it's worth naming them directly.

Common Mistakes That Break An Intent Scoring Model

1. Scoring Volume Instead of Proximity to Purchase

Ten blog reads shouldn't outscore one demo request just because ten is the bigger number. Weight signals by how close the behavior sits to an actual buying decision.

2. Trusting a Single Signal

One data point triggering full Tier 1 routing is how rep trust collapses inside a month. Require at least two independent signal types before anything reaches your top tier. Intent plus Engagement works. So does Intent plus a fresh buying trigger.

3. Skipping the Reach Layer

A scoring model with no Reach layer will happily rank unreachable accounts as top priority. This is the mistake I see more than any other. It's also the cheapest one to fix, since the data to fix it usually already exists somewhere in your stack.

4. Running the Model Without a Review Cadence

Call this Intent Score Theater. A dashboard that looks impressively data-driven in a QBR slide, and changes nothing about what a rep does on Monday morning. If nobody checks correlation against closed-won, the model is decoration, not infrastructure.

5. Scoring Contacts Instead of Accounts

B2B purchases get made by committees, not individuals clicking links. A model built around single contacts will consistently misjudge how close an account really is to a decision.

Avoiding these five mistakes is partly a discipline problem. It's also partly a tooling problem, and that's worth addressing directly.

Tools You Need to Build This

A first version of this model doesn't require a data science team. It needs three things working together:

1. A source of third-party intent, such as an intent data provider like Bombora or G2, or both.
2. A way to enrich and verify contact data, so the Reach layer means something instead of just looking good on a slide.
3. A CRM or workflow tool that can calculate the composite score and route accounts automatically.

This is one of the few places a consolidated sales intelligence platform earns its keep. It beats adding another login to your GTM tech stack. Native intent data, verified contact coverage, and CRM sync living in one place removes an integration layer. That layer would otherwise sit between your intent signal and your Reach score. Every extra system in that chain is another place data quietly goes stale before a rep sees it.

A quick gut check for your current stack: pull ten accounts your intent provider flagged "hot" this week. Check how many have a verified, working phone number for someone in the buying group right now. Under half, and the problem isn't your scoring logic. It's the data foundation underneath it.

A few questions come up constantly once teams start building this. Here are the ones worth answering directly.

The Bottom Line

A scoring model isn't a data project. It's a trust project that happens to run on data. The real test isn't whether the math is elegant. It's whether a rep looks at a Tier 1 account and believes it's worth the next hour. Or whether they quietly go back to working lists by gut feel, because the last three "hot" accounts wasted their morning.

Weight the four layers from real closed-won data. Decay what deserves decaying. Score the buying group, not one contact. And build the Reach layer everyone else skips. The best intent signal in the world means nothing without a working number behind it.

See how SMARTe pairs native Bombora intent data with verified mobile and email coverage across your buying group. Reach will never be the weak link in your model again.