Dan Harmeson and the customer who already told you

A customer can be perfectly pleasant while preparing to leave. The meetings continue. The account notes look respectable. Somewhere else, inside the software, the pattern of use is changing. Dan Harmeson has built a company around the possibility that the product is giving an earlier, more useful account of the relationship.

It is an awkward proposition for an industry that devotes so much energy to talking. Harmeson comes from that industry’s operating ranks: consulting, analytics, services, strategy and operations. He has worked inside the organizations responsible for making enterprise technology deliver something a customer will keep paying for. At QuadSci, which he co-founded with Sean Murray in August 2023, he wants the evidence of that value to reach the people making commercial decisions.

The premise has a compact slogan: “Telemetry is truth.” Telemetry is the trail left by software use, from a button clicked in an application to the services running behind it. Harmeson’s interest is in what those traces reveal about a customer’s direction. A subscription has a renewal date. A relationship has a history. The two deserve to be read together.

Before the forecast, the operating experience

Harmeson studied business administration at North Carolina State University. His professional route took him through strategy and transformation work at IBM, financial services consulting at Knowledgent, and then SAS. There, his responsibilities included emerging initiatives and the commercial strategy for the Internet of Things. The work placed applied analytics alongside the practical question of where an organization could use it.

There is a useful early marker in Munich. On February 20, 2019, Harmeson was on the German Data Science Days program representing SAS. His subject was intelligence for connected systems, spanning analytics at the edge and across the enterprise. Before QuadSci’s customer forecasts, he was already working around a related problem: how to turn the activity of a complicated system into information someone could use.

He later led advisory services in the Americas at MuleSoft, then strategy and operations for Elastic’s global Customer Success Group. These were roles close to implementation and execution. Software companies promise results; services and operations teams have to organize the work that follows. That background gives his present argument a particular emphasis. The point of collecting more information is what it allows a team to decide.

Harmeson and Murray had both worked at MuleSoft and Elastic. Their backgrounds brought them repeatedly to the same commercial difficulty: customer teams had conversations and records, while the product generated another body of evidence. Their startup emerged from the conviction that this behavioral record could support a more disciplined way of managing customer revenue.

Two CEOs, one stubborn question

QuadSci has a shared leadership arrangement. Harmeson and Murray are co-CEOs, and both sit on its board. Maggie Bartels leads the technology organization as CTO. The combination brings operating experience together with the work of extracting patterns from large datasets. It also places a fairly demanding translation job at the center of the company: make the analysis useful to someone who has a territory, a customer list and a limited number of hours.

The company’s two core products approach that job from different directions. Cohorts AI groups customers according to patterns in their actual product use. Growth AI connects behavioral evidence with predictions about revenue outcomes, including churn and expansion. Q-Chat adds a conversational way to interrogate the intelligence and work out what to do next. The products are aimed at the account, rather than merely the individual click.

A team photograph from July 2024 offers a less abstract view. Harmeson stands with data scientist Andrés Lomelí and Murray, a display of charts behind them. Lomelí described a working environment with room to experiment and build internal tools. The picture puts colleagues around the statistical vocabulary. Somebody has to construct the tools before anybody can ask them an interesting question.

For Harmeson, the recurring question is whether an account is receiving value in ways its commercial team can recognize. More activity alone does not settle that. A company can use one function intensely while abandoning another; different groups of customers can follow different paths. The appeal of the approach is the chance to connect those patterns with what later happened commercially, rather than treating every account as the same kind of customer.

Show the past before selling the future

Forecasting has an obvious sales difficulty: the decisive evidence is usually scheduled to arrive later. Harmeson’s discussion with Ray Rike on the June 2, 2026 episode of AI to ROI addresses how QuadSci builds trust before asking a customer to change its behavior. The company tests predictions against outcomes that have already occurred, including information the model did not train on.

That sequence matters. If a revenue team is going to redirect attention toward an account that seems comfortable, it needs a reason it can examine. A prediction becomes useful when people can understand enough of the evidence to make a decision. Harmeson’s emphasis on verifiability fits his operating background: an analytical result eventually reaches someone who must take responsibility for the next conversation.

“Telemetry is truth.”

Dan Harmeson’s recurring shorthand for the company’s premise

QuadSci reports predictive accuracy of around 90 percent for churn and growth at horizons of nine to eighteen months. Those are the company’s reported figures, tied to its deployments and published examples. For a customer team, the interesting implication is time. Earlier notice creates room to investigate, offer help or plan an expansion before a renewal becomes an end-of-quarter emergency.

The company’s work with Clari illustrates why the question extends beyond departures. Clari’s president, Kevin Knieriem, described finding growth signals while investigating churn. That pairing is central to Harmeson’s proposition. The same effort to understand a customer can reveal an opportunity to do more business together. A list of endangered contracts is only one possible result of paying closer attention.

A roadmap that leaves the building

Harmeson’s public account of a visit to Boomi’s Vancouver office brings the story back to a room full of users. He went with Karishma Bhambhani and Ryan Wendt for an enablement and product roadmap session with Boomi’s Lead Customer Success Architects. He described candid feedback and a team building its customer success practice around QuadSci.

The account is revealing because the feedback runs in both directions. QuadSci is built to help a software business learn from its customers’ behavior. Its own team also has to learn from people explaining how they work. A forecast may identify a change worth investigating; a working session can show whether the product makes that investigation easier. The people receiving the intelligence remain part of the design problem.

WATCH THE CONVERSATION ↗

His conversation with SBI Growth’s Nick Toman and Craig Riley similarly placed product analytics inside a broader discussion of retention. Joint research by SBI and QuadSci examined 160 billion telemetry data points across 9,100 accounts. The scale is part of the argument for machine learning: the useful pattern may be difficult to see by inspecting individual accounts or applying a single hand-written rule.

Eight million dollars, then another release

In February 2026, QuadSci announced an $8 million Series A led by Crosslink Capital, with Alumni Ventures, Correlation Ventures and individual investors participating. The company said the money would support product development, its commercial teams and enterprise growth. For Harmeson, the financing arrived less than three years after the company’s start.

By July 15, he was writing the announcement for Release 3.0. It introduced fully conversational Q-Chat and the QuadSci MCP Server, which makes the company’s behavioral intelligence available to connected AI systems. The release extended the way teams could reach the analysis. A user could ask a question in ordinary language, or bring the intelligence into another agent environment.

His concern in that announcement was whether AI investment changes revenue outcomes. Faster work and cleaner summaries have value, but Harmeson keeps returning to the existing customer base. His aspiration is to make customer intelligence part of how software companies operate. The ambition is easier to assess in those concrete terms: which account, which behavior, which decision, and how early?

His advice to people starting their careers follows a similar line. He pairs fluency in AI with a deep understanding of business processes. That combination connects the tool to the work it is supposed to improve. It also helps explain his own progression: the services leader and the operations executive are still present in the startup founder. A system can produce an answer without resolving the question of who should use it, how it fits into a process, or what changes afterward. Harmeson’s interest lies in getting that entire chain to work. The useful result reaches beyond the person building the model.

The burger at the end of the forecast

Ask Harmeson about New York and the vocabulary becomes pleasantly specific. He once lived in Stuytown. He enjoys Alphabet City, the East Village and the West Village. Minetta Tavern is a destination for a burger. After the cohorts, telemetry events and revenue forecasts, a menu is a reassuringly manageable dataset.

The detail gives the profile a human scale. Harmeson’s working life now includes a shared CEO role, customer discussions, hiring and successive product releases. The question connecting them is still the one that brought him and Murray to QuadSci: how can a software company understand its customers well enough to act earlier? His answer begins with the record those customers are already leaving, every time they use the product.