By clicking “Accept All Cookies”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.
Revenue Operations

The Two Scores Your GTM Will Rest On: Account Scoring and Contact Scoring

swirled squiggle accent

Most go-to-market teams have spent the last two years pouring AI investment into one place: outbound volume. AI SDRs, automated sequences, scaled messaging. And the result, as anyone paying attention to response rates already knows, has been a race to the bottom. Digital channels are saturated. The volume goes up. The signal goes down.

‍

There is a different use for AI in a GTM system — one that goes in the opposite direction. Instead of pushing more messages out, it uses AI to understand more precisely who deserves to receive them in the first place.

‍

In a recent RevOps Co-op webinar, Matthew Volm sat down with Anthony Goldbloom, CEO of Sumble, to walk through the data science logic behind account scoring and contact scoring — and why getting those two things right changes what's possible across the entire revenue organization. An opening audience poll asked how confident attendees were that their current scores actually reflected who they should be selling to. The majority landed in one of two buckets: either not scoring at all, or not very confident. One attendee reported being very confident. Anthony, characteristically, said he was very confident in Sumble's account scoring — and still actively fine-tuning contact scoring. It was an honest place to start.

‍

Why Precision Scoring Is Different Now

The case for scoring accounts and contacts isn't new. What's new is the level of granularity that AI makes possible — and how dramatically that changes the usefulness of the output.

‍

Anthony opened by walking through a single job posting from Walmart. From that one document, Sumble's system could extract which team posted it, what platform migration was underway, which technologies were in use (Cassandra, Docker, Kubernetes), and which organizational structure the team rolled up into. That's meaningful intelligence from one data point. Now multiply it across two years of job postings from a single company, then layer in blog posts, podcast appearances, executive interviews, GitHub activity, and LinkedIn profiles — and the picture of what's happening inside a prospect account becomes extraordinarily detailed.

‍

"You can piece that together and come up with a very granular understanding of what's happening in an account. This was not possible in a world before AI because nothing could process and structure and draw intelligence from these job posts." — Anthony Goldbloom

‍

This is the precision he was referring to. Not "this company has 500 employees and uses Salesforce." Rather: which specific teams exist inside the account, what they're building, which technologies they're actively adopting, and who the relevant people are for a given seller.

‍

The other half of the precision argument is data quality. Anthony described the compounding error problem that every RevOps professional has encountered but rarely names this clearly: if company record accuracy is 80%, person-company match accuracy is 80%, people data freshness is 80%, job function classification is 80%, and seniority classification is 80% — the ability to do something as basic as count the number of designers in a company ends up at 33% accurate. Matthew captured the audience reaction precisely: "If someone tells you 33%, you're like, 'Well, that is garbage.'" The degradation is real; it just happens quietly at each step of the chain rather than announcing itself.

‍

Historically, achieving high-quality GTM data at scale required the Bloomberg approach: thousands of people in lower-cost geographies continuously cleaning records. That was never economically viable for go-to-market data vendors. AI agents change the equation.

‍

This connects directly to a broader pattern the RevOps community has been grappling with — data management at the foundational level matters far more than most teams acknowledge until a downstream system fails because of it.

‍

How Account Scoring Works in Practice

Sumble's account scoring model is built around three primary dimensions: size, growth, and concentration.

‍

Size is intuitive — larger accounts support higher average contract values. Growth matters for two reasons: a fast-growing company will be larger in a year or two, and growing companies encounter broken systems more often, which means they reach for new tools more frequently. Concentration measures what percentage of an account's employees are actually relevant to what you sell. A company with 2 million employees where the sales team is a tiny fraction of headcount is a very different opportunity than a 500-person company where go-to-market roles represent a substantial share of the organization.

‍

Within each dimension, the scoring drills down to genuinely granular data points. For a company like Gong, Sumble's system might identify how many teams use GTM enrichment and signals, whether they're using Clay (a signal of modern data practices), how large the RevOps team is, how fast the sales development team has grown over the past year, and what fraction of total headcount sits in revenue-facing roles. Anthony walked through a specific example: Gong's RevOps team had added 67 people over the previous year. Sales makes up 26% of their overall headcount. These aren't estimates derived from firmographic categories — they're inferences drawn from a continuous stream of job postings and profile data, each traceable back to a source.

‍

That traceability matters for a reason beyond accuracy. When a data point is wrong — and Anthony was direct that mistakes happen — the ability to click through to the original source means trust isn't lost. Users can see exactly what the system inferred and what evidence it was based on. That transparency is what separates a scoring system people actually believe from one that gets ignored. For more on what happens when data quality erodes trust in the systems built on top of it, this piece on CRM data management covers the downstream consequences in detail.

‍

The scoring model is also configurable. Out of the box, Sumble provides a default score based on its understanding of what a given company typically cares about. But customers can weight the components differently — more emphasis on growth, less on size, or vice versa — and those adjustments update the ranked list immediately. The system can also be fitted to past closed-won accounts, using historical win data to calibrate what a good account looks like going forward. Anthony noted the limits of that approach: if a company sold primarily to SMBs early on and is now moving upmarket, past closed-won won't reflect the target. In those cases, the alternative is to have the team collectively nominate their dream accounts and fit the model to those instead.

‍

Contact Scoring: Who Inside the Account?

Once an account is identified as high-value, the next question is who inside that account deserves attention. This is where contact scoring comes in, and it's the dimension Anthony described as less widely deployed than account scoring.

‍

Contact scoring in Sumble works similarly to account scoring in structure — a ranked list, fully decomposable, configurable — but applies to people rather than companies. The system scores individuals based on job function, seniority, specific skills and technologies they mention in their professional profiles, the initiatives they're connected to, and their relationship to job postings that signal relevant work.

‍

The configurability goes deep. In the demo, Anthony showed how a team could set the slider so that a CMO and a director of GTM engineering receive equivalent scores — calibrating the system to reflect that both personas are equally valuable for a specific selling motion. Adjust the slider, and the ranked list of contacts updates immediately to reflect the new weighting. That kind of real-time interactivity is what makes it possible to get cross-functional alignment on who the target contacts actually are.

‍

One of the more subtle contact-scoring features Anthony highlighted: a junior person at a highly scored account may be more valuable than a senior person at a low-scored account. The system allows sellers to weight account fit into the contact score itself, so territory doesn't just optimize for seniority in a vacuum.

‍

For teams thinking about how contact scoring connects to lead routing and prioritization decisions, the linkage is direct: when every person in a prospect universe has a rank, any inbound lead can be instantly contextualized against that rank. Anthony showed this in Sumble's internal system — a lead comes in, and the system surfaces that this person is the head of RevOps at an account ranked 102, and the sixth most important contact at that company. That context arrives before the rep has made any judgment call.

‍

The One-Score Rule: Why Scoring Discipline Matters

Matthew raised a trap he'd seen teams fall into — and admitted falling into himself. When scoring becomes popular inside a GTM organization, the instinct is to apply it to everything. Accounts get scored. Contacts get scored. Leads get scored. Meetings get scored. Events get scored. And once every entity has a score, the scores stop meaning anything.

‍

Anthony's response was unambiguous: each entity should have exactly one score.

‍

"A man with one watch knows exactly what time it is. A man with two is never sure. If you have two scores for different elements of an account, like when you're prioritizing that account, which score do you listen to?" — Anthony Goldbloom

‍

The resolution he proposed isn't to score fewer things — it's to roll all relevant components into one number, then allow users to decompose that number as deep as they want. Someone who just needs to know which account to work next looks at the top-line score. Someone who wants to understand why a particular account ranked where it did can drill through the components: size, growth, concentration, and within each, the specific data points that drove the sub-score.

‍

The scoring discipline extends to what should be scored at all. Anthony's test: does a score drive an action or a prioritization decision? If yes, scoring is useful. If not — and his example was a "meeting score," where the action driven by the score is unclear — the score probably shouldn't exist.

‍

This principle connects to a pattern worth naming. The instinct to measure more, track more, and add more scores to dashboards is a version of the same problem as adding more reports for the sake of having more reports. Measuring things without a clear action attached is a form of ops debt, and scoring systems are not exempt from it.

‍

What Downstream Becomes Easier

The argument Anthony returned to throughout the session is that account scoring and contact scoring are not just useful in themselves — they're the foundation that makes a surprising number of other GTM functions cheaper and more accurate to execute.

‍

Field marketing and event selection. If you have a ranked list of every person in your addressable market, you can monitor where those people are speaking, attending, and appearing. Sumble surfaces events by cross-referencing the contact universe against known speaker and attendee lists. The question "which event should we sponsor?" becomes answerable without instinct-based guessing: deploy dollars where the highest-ranked people for your business are showing up. The same logic applies to podcast sponsorships.

‍

Territory planning. With reliable account scores, territory equity becomes something you can measure rather than argue about. Anthony described using the KS distance — a statistical measure of distributional balance — to quantify how equitably a territory set is distributed. When accounts are ranked and that rank is trusted, territory disputes shift from qualitative disagreement to quantifiable imbalance. Reassigning an account updates the equity metric in real time.

‍

TAM sizing and white space discovery. The same account score applied to your CRM can be applied to everything outside your CRM. The result is a ranked list of accounts that should potentially be added — ordered by how good a fit they are, not by how recently someone thought to look them up. Anthony showed a real example: a hot AI-native company that had been missed in Sumble's own CRM surfaced immediately when they ran this exercise.

‍

GTM headcount planning. If you know how many highly ranked accounts exist and how many are unallocated to current reps, you have a principled basis for hiring decisions. This connects directly to the capacity planning work that RevOps teams often do from less reliable inputs.

‍

Lead routing. With every person in the market pre-ranked, inbound leads arrive with immediate context. Rather than routing based on geography or round-robin assignment, teams can route based on how highly the inbound person ranks — and ensure that the highest-ranked inbound contacts go to the reps best positioned to handle them.

‍

The case study that anchored these downstream benefits came from Nooks, a customer Anthony described as a power dialer that had just closed a $500 million round. Nooks worked with Sumble to apply their account score to their full addressable market rather than just their existing CRM records. They found 700 high-quality accounts that hadn't been in their system. That discovery was substantial enough to justify doubling the size of their enterprise AE team.

‍

Who Owns the Score — and How Often Should It Change?

A second audience poll asked who owns the definition of a "good account" and a "good contact" inside their company. The majority answer was what Anthony expected and what anyone who has worked in a mid-to-large GTM organization will recognize: everyone has their own version.

‍

The reason this happens, Anthony argued, isn't that companies don't care about alignment. It's that previous tools didn't make it possible to get to a version that anyone was really happy with. RevOps would build one for territory planning. Marketing would work from something different for campaign targeting. Sales reps would operate from their own instincts.

‍

The conditions for genuine alignment around a shared score now exist in a way they didn't before — not because the alignment problem has become easier politically, but because the underlying data has become granular enough to be convincing, and the tooling makes it possible to sit in a room and debate accounts interactively. When someone asks "why is OpenAI ranked fourth?" and can get an immediate answer grounded in actual data points, the conversation is no longer about whose intuition to trust.

‍

The question of how often to change score weights came up in the closing discussion. Anthony's framing was useful: there are two distinct moments that should trigger a fresh look. The first is an ICP shift — a new product, a move upmarket, a new persona added to the target. These are moments when the components of the score itself may need to change, not just the weights. The second is ongoing feedback from the people using the scores. His recommendation was practical: maintain an open channel where reps and marketers can flag rankings that feel off.

‍

"I think it's always interesting when a ranking feels off to somebody. Take a look at what drove it — why does account three rank lower than account one. And it doesn't necessarily mean change it, but use that as a moment to say, 'Okay, yes, they implicitly think size matters more than growth. Do I agree?'" — Anthony Goldbloom

‍

He gave a real example from that same day: a Sumble rep had flagged that his number-three account seemed like it should outrank his number-one account, because the third was larger and the first was simply growing faster. Anthony walked through his reasoning, decided not to change the weighting, and explained why. The rep's concern was legitimate; the response was engaged rather than dismissive. That exchange — feedback given, genuinely considered, decision made and shared — is what builds the kind of trust between RevOps and sales that most teams say they want but rarely operationalize.

‍

Matthew's framing for the RevOps function here was equally direct: don't be defensive about scores. The value of the exercise isn't that the score is always right. It's that thinking through why you have it weighted the way you do, regularly, reinforces what actually matters to the business — and proves it's embedded in how you're prioritizing.

‍

The "Surround Sound" Effect of Shared Scoring

The through-line of the session that deserves its own treatment: what happens when marketing, RevOps, and sales all work from the same account list and the same contact list.

‍

Anthony described this as "surround sound." The age-old marketing principle — that someone needs to encounter a brand seven or eight times before it sticks — is well-known. What's less often operationalized is the insight that this effect only compounds when all those encounters are aimed at the same people. If the field marketing team invites the top-ranked contacts in a city to a dinner, the SDR team is doing outreach to the same people for the same accounts, and LinkedIn ads are targeting the same individuals — the reinforcement is real. Each touch lands in a context where the other touches have already been made.

‍

The inverse is also true, and more common: when marketing, sales, and RevOps each operate from different definitions of who matters, the coverage is scattered. High-value people get missed while low-value ones get saturated. The problem isn't the channels; it's the absence of a shared ranked list that everyone trusts.

‍

Wherever that shared score lives — Salesforce, a data warehouse like Snowflake or Databricks — the same account score and contact score should be what drives marketing campaign prioritization, RevOps lead routing and territory planning decisions, and rep-level account prioritization. The score is only as powerful as the breadth of adoption behind it. Aligning revenue teams around shared KPIs is the broader organizational challenge this sits inside, and account scoring is one of the most concrete ways to make that alignment operational rather than aspirational.

‍

Key Takeaways

  • AI enables a new kind of precision in account understanding. By processing job postings, case studies, podcast appearances, and other public signals at scale, it's now possible to build a detailed picture of what's happening inside a prospect account — which teams exist, what they're building, and who is relevant — in ways that weren't feasible before.‍
  • Data quality errors compound. If each step in a chain of contact data processing is 80% accurate, the end result is roughly 33% accurate. This is why it's worth investing in accuracy at every stage rather than accepting each step's error as acceptable in isolation.‍
  • One score per entity. Each account, each contact, each event should have exactly one score. Rolling multiple factors into a single decomposable number is more useful than maintaining separate scores for separate dimensions — the decomposition is for understanding, not for parallel decision-making.‍
  • Scores are only worth having if they drive an action. The test for whether something should be scored: does the score drive a prioritization decision or a specific action? If not, it probably shouldn't be scored.‍
  • Surround sound requires a shared list. The compounding effect of marketing, SDR outreach, and digital advertising reaching the same people only works when all three teams use the same ranked account and contact lists. Organizational alignment around scoring isn't just an ops goal — it's a revenue multiplier.‍
  • Score governance is never “done.” Create a channel for stakeholders to flag scored records that seem “off.” Investigate the flagged records, decide whether to adjust the score, and explain the reasoning behind your decision. This process builds trust with reps and strengthens the scoring logic over time.

Getting account scoring and contact scoring genuinely right — precise, transparent, trusted, and updated continuously — is the single investment that makes the most other GTM work easier. Territory planning, lead routing, field event selection, white space discovery, headcount planning, and cross-functional alignment all become more tractable when everyone agrees on who matters and can see why.

‍

Learn more about how Sumble helps go-to-market teams score their universe with AI-powered account and contact intelligence, so every part of the revenue organization is working from the same trusted list.

‍

Looking for more great content?

Check out our blog, join our community and subscribe to our YouTube Channel for more insights.

Related posts

Membership Options

Pick your path

Join 21,000+ RevOps pros. Start free, or unlock the full community and everything that comes with it.

Free 👋
Free forever

A newsletter full of best practices and lessons learned, plus a heads-up on upcoming in-person and digital events.

Get started
Starter 🕺 Popular
The full community

Our community of thousands of RevOps pros, plus a peer-to-peer matching program.

Join Now
Rise 🚀
Everything, leveled up

Everything in Starter plus course and conference discounts, our knowledge hub, career coaching, and more.

Join Now

Reach your RevOps goals

Our average member has 5+ years of RevOps experience — so you'll have real-time access to seasoned pros. All we ask is that you're generous with your knowledge in return.

Become a Member →