Most revenue teams have more data than they know what to do with. Dashboards report on conversion rates, customer acquisition cost (CAC), and pipeline coverage. Forecasts describe what happened last quarter. And yet, when a board member or private equity owner asks the essential question — where should we invest our next dollar of growth capital to generate the best return — the answer is rarely in the dashboard.
The gap between describing revenue performance and actually improving it is one of the defining challenges in modern go-to-market (GTM) operations. Closing that gap requires a fundamentally different orientation toward analytics: not just reporting what happened, but answering the hard questions that drive business outcomes.
In a recent RevOps Co-op webinar, Matthew Volm, CEO of RevOps Co-op, sat down with Stephen Diorio, Managing Director at Revenue Operations Associates, faculty member in the Chief Revenue Officer (CRO) program at University of Chicago, and author of the book Revenue Operations, to explore what it actually looks like to use analytics to drive revenue growth — not just measure it.
The session opened with a poll asking attendees which GTM question their teams struggled most to answer with data. The results were telling: forecast confidence led the pack, followed closely by pricing and packaging decisions and budget allocation. The spread across all five answer choices confirmed what Diorio has seen across clients for decades: the problem isn't a single blind spot — it's the entire frame of what analytics is supposed to do.
When a board or private equity owner allocates capital to go-to-market, they expect a financial return. That return — whether expressed as net recurring revenue, future cash flow, or growing customer lifetime value — should correlate with the sum of value generated across the entire customer base. That is the baseline expectation.
The problem, Diorio argued, is that most dashboards don't connect to this expectation at all.
"Analytics for decades, defined by great companies like Tableau, is: I have the information, I have a funnel, now a bow tie, I'm gonna describe what's going on there. That is antithetical to what private equity owners, boards, CEOs are asking." — Stephen Diorio
The bow tie funnel — demand at the front, conversion in the middle, customer expansion at the back — is useful for describing a system. But it doesn't tell you how to improve it, where to invest, or which customer segments are worth pursuing. It describes the water flowing through the pipes. It doesn't tell you which pipes to build next.
This is the fundamental reframe Diorio teaches CROs at University of Chicago: analytics should answer strategic questions, not just describe operational state. And there are, according to his framework, 12 of those questions — everything from "What offerings, through which channels, to which customers?" to "How much should we invest, and where in the revenue cycle?" to "How do we measure and improve performance, not just report it?"
Most RevOps teams, operating under bandwidth pressure and constrained organizational remit, are lucky to address five.
When Volm asked why most teams never get to all 12 questions, Diorio's answer was direct: "All of the above." Restricted remit, educational gaps, organizational silos, and bandwidth — all of it is real. But the single most pervasive constraint is structural.
"The problem with RevOps people is they spend 95% of their time fighting fires, keeping the machine running, and running reports. And that's the Lord's work — and you keep the company running — to the degree that you can fight like crazy to find the time to ask these questions." — Stephen Diorio
To help practitioners think about how to get beyond this reactive posture, Diorio introduced a framework he calls RICE: Report, Investigate, Collaborate, and Educate.
Most RevOps teams are firmly in the "Report" stage. The goal of RICE is to move deliberately through the other three:
That last point matters more than it might seem. As Diorio put it, even the most junior ops professional can ask smart questions. Asking "Should we be changing our pricing and packaging?" based on RFP data is not outside anyone's purview — even if implementing the answer is.
This connects directly to a broader challenge the community has been wrestling with: how RevOps can operate more strategically even when organizational constraints limit formal authority. The answer, in practice, often starts with better questions.
One of the sharpest exchanges in the session came when Volm named a pattern that runs through nearly every part of GTM analytics: teams default to measuring what's accessible rather than what's meaningful.
"We measure what we can, not what we should." — Matthew Volm
Diorio immediately confirmed this is one of the most persistent traps he sees with clients. The practical implication is that entire analytical frameworks get built on the data that happens to be available — rather than the data that actually explains outcomes.
The ICP example Volm offered illustrated this clearly. Historically, ICP definition has been shaped by what traditional data providers could supply: employee count, industry code, revenue range. These are proxies — filters that narrow the universe of companies to something manageable — but they often have little to do with the actual value a product or service delivers. A shipping company that handles heavy goods efficiently might serve manufacturers across twenty industries and five revenue bands. None of the traditional firmographic filters would capture that.
Diorio extended the idea by pointing to how legacy thinking about sales territories follows the same pattern. Territories were originally defined geographically — by how far a salesperson could drive — and the logic has simply persisted through decades of iteration. Industries and firmographics replaced geography, but the deterministic, top-down structure remained.
"Companies don't buy, people buy — and people buy because of needs, roles, signals, and triggers. So he uses the example of: we should be describing territories like dating apps. Why aren't we using probabilistic data about individuals and events and signals, and defining territories as if somebody with this problem in this role demonstrates this behavior — that's my territory." — Stephen Diorio
The practical implication for RevOps practitioners is that breaking free of legacy measurement frameworks requires a deliberate act of reimagination. ICP refinement isn't a one-time project; it's an ongoing process of asking whether the variables being measured are the ones that actually predict outcomes — or just the ones that happen to be in the database.
The second poll of the session asked attendees where most of the revenue-generating knowledge in their companies currently resides. Nearly two-thirds said "messy data in systems, but hard to trust." Only two respondents said their data was organized and reliable. Nobody selected "codified knowledge."
This distribution, Diorio argued, is not just an inconvenience — it's a strategic liability. And understanding the difference between data, knowledge, and codified knowledge is foundational to thinking about where analytics creates real value.
Data is an expense. Knowledge — meaning data that has been organized and contextualized well enough to reliably inform decisions — is a step up. Codified knowledge is something else entirely: it's the ability to use information to reliably generate revenue. Think of Amazon's recommender engine, or Spotify's playlist logic. Those systems don't just describe what customers have done. They use that understanding to drive future purchase behavior.
"Our job is to mine this ecosystem to manage, curate, and monetize data assets, and put it into formats where it could be monetized. A lot of the best knowledge in your business is between the ears." — Stephen Diorio
The organizational implication is significant. Most businesses assume their valuable data is structured and centralized — in the CRM, the marketing automation platform, the data warehouse. In reality, more than 90% of the information sitting in a central data repository is unstructured: email threads, Slack messages, call transcripts, proposal documents, and bid management records. That's where the actual knowledge about customers, deals, competitive dynamics, and buying behavior lives.
This is why tools that capture and analyze conversations — Gong and similar platforms — have become valuable not just for sales coaching but as a source of organizational intelligence. Transcripts that used to serve only sales managers now feed product marketers, content teams, and customer success organizations with information about how buyers actually talk about their problems. As Volm observed, that data is moving rapidly from a sales tool to a company-wide asset.
Data management in RevOps has always been foundational, but the emergence of AI-powered analysis of unstructured data is changing the equation. The challenge isn't just cleaning what's in the CRM — it's capturing, organizing, and activating knowledge that has never lived in structured form at all.
One of the most practically important threads in the session was a warning about a trap that is becoming more common as AI makes analysis faster and easier: confusing data availability with data relevance.
The session touched on an example from Diorio's work with clients on win-loss analysis. When examining deal outcomes, teams often lean heavily on whatever data is easiest to capture — digital engagement metrics, meeting attendance, sentiment scores from call recordings. These inputs feel rigorous. They're quantifiable, they're systematic, and they're available at scale.
But when Diorio ran actual closed-loop win-loss analysis for a client, the top predictors of purchase decisions were brand preference and relationship strength — factors that are genuinely difficult to quantify and that rarely make it into a predictive model. The easy-to-measure proxies (website visits, call sentiment, response time) were capturing noise, not signal.
"One of the biggest traps is not actively challenging what data is informing our model — and is it predictive? You always fall into the trap of over-interpreting available or sentiment data. People go way too far with NLP data, 'cause it's easy to mine." — Stephen Diorio
The danger is compounded by AI. Machine learning models will find correlations because they're designed to produce answers. If the training data is built on available variables rather than causal ones, the model will confidently produce output that looks rigorous but drives decisions in the wrong direction.
"AI is gonna find causality 'cause they wanna give you an answer. And you're allocating resources against accounts based on really bad models." — Stephen Diorio
The discipline required here is uncomfortable: the most predictive variables are often the hardest to measure, and accepting that discomfort is necessary. Messy, qualitative, difficult-to-quantify information — relationship quality, brand perception, buyer-side organizational dynamics — may matter more than everything that's cleanly structured in the data warehouse. As Diorio put it, "you can't say, 'Well, let me just find a lot of easy stuff that explains it,' because it's a false positive."
This connects directly to broader questions about AI readiness and data governance: AI can't fix assumptions that haven't been challenged, and a fast model built on bad inputs will confidently produce bad outputs at scale.
With the conceptual frame in place, Diorio walked through what he called "the menu" — a set of analytical applications where RevOps teams can generate measurable financial value, organized around the journey from messy data to codified knowledge.
Several of these deserve particular emphasis:
Response function analysis addresses over half of all GTM decisions, which are based on assumptions about cause and effect. Should we run three salespeople in a territory or one? Hit a prospect ten times with ads or four? Understanding the S-curve of how inputs translate to outcomes — and identifying the threshold below which effort is insufficient and above which it becomes counterproductive — is one of the highest-leverage analytical investments available.
Propensity-to-buy modeling, done correctly, moves teams past deterministic segmentation (firmographics, geographies, roles) and down to the specific individuals on a buying team, what they care about, and whether they're ready to engage. This is the "dating app" model Diorio referenced earlier — probabilistic, signal-driven, and focused on people rather than company profiles.
Pricing optimization is significantly underutilized by RevOps teams, partly because it requires collaboration with product and product marketing organizations that many ops practitioners don't feel empowered to approach. But Diorio was direct about the stakes: "Modulating the four Ps could be a game changer. Understanding willingness to pay could put a lot of money in your pocket." Won-loss data that traces deal losses to pricing, packaging, or positioning is available to most organizations — it's simply not being mined.
Marketing mix analysis expands the frame beyond the channels that are easy to measure (paid digital, search, social) to include partners, in-person events, thought leadership, and increasingly, large language model (LLM) visibility. As more B2B buyers use AI tools to research vendors, share of voice in LLM outputs is becoming a meaningful consideration — one that requires a more integrated view of investment across content, brand, and earned presence than most RevOps teams currently take.
Seller performance analysis replaces the blunt instrument of quota attainment with a more nuanced view of individual seller capacity, capability, and fit for the accounts they're covering. As Diorio noted, most organizations apply one-size-fits-all quotas to sellers with meaningfully different skill profiles and territory compositions. Sales capacity planning that accounts for individual seller productivity rather than uniform benchmarks produces better quotas and better outcomes.
The through-line across all of these applications is the same: moving from data that describes what happened to knowledge that drives what happens next.
The practical path from two-thirds of attendees saying their data is "messy and hard to trust" to something resembling the codified knowledge that Amazon or Netflix uses to drive behavior involves a series of deliberate steps.
Diorio outlined several:
Mine third-party signals and augment CRM profiles with first-party engagement data. Call transcripts, campaign engagement, and digital behavior all carry signal that rarely makes it into structured fields. Bringing that data into a central repository — tagged to account and contact identifiers — makes it available for analysis.
Implement trust scoring. Before relying on any data field for a decision, RevOps teams should be able to answer: where did this data come from, how current is it, and how reliable is it? AI can help automate this assessment, but the framework needs to be human-designed.
Automate deduplication and organize data around accounts and people, not products or geographies. This is unglamorous work, but it's foundational. Data organized around business units or product lines creates persistent barriers to the account-level analysis that drives actual sales decisions. Fixing broken data structures is prerequisite to extracting value from them.
Build closed-loop learning processes. The human-review step in any AI-assisted workflow matters enormously. As Volm described from a recent community session, the key to avoiding feedback loops that amplify incorrect inferences is ensuring that what gets fed back into the learning system is grounded in verified outcomes — not AI-generated inferences treated as facts.
Prioritize smarter, not just harder. The forecasting challenge most attendees identified isn't really a forecasting problem — it's a resource allocation problem. The question isn't whether to predict accurately; it's where to concentrate effort in the last 30 days of a quarter to maximize upside while managing downside risk. That's a portfolio question, and it deserves a portfolio answer.
The revenue data most teams are sitting on is more valuable than it looks. The gap isn't in the data itself — it's in the questions being asked of it.
Check out our blog, join our community and subscribe to our YouTube Channel for more insights.
