A product funnel the whole company could finally trust

When I joined Lucen Software, the company steered by a single product metric almost no one trusted. Free-to-Paid conversion drove real decisions, yet leadership could not say how it was calculated, whether it measured the right thing, or why it moved. By the end, working closely with the CFO and the executive leadership team, we had defined a funnel together, a number the VP of Product, the CFO, and the CEO all recognized as the source of truth, and a way to reason about product conversion instead of second-guessing it.

Lucen was a product-led company built around a PowerPoint add-in, growing quickly, with data spread across systems that had each been built on their own. The metric had been inherited from engineering, with no business owner and no documentation. Fixing it was as much a problem of agreement as of data.

Key data points

• 5-stage funnel, measured end to end. Stage-by-stage conversion and an overall download-to-paid rate, in place of one opaque metric.
• One definition, resilient across systems. Telemetry, billing, and the CRM unified into one source of truth, defined by events rather than tools, so it survived a wave of migrations.

The Problem

The company measured product conversion with a single metric called Free-to-Paid. It had been designed by engineering and calculated from telemetry: it looked for the moment a user finished installing the software, then for the moment that same user's license turned from free to paid.

The trouble was a hidden setting. The paid license had to appear within a configured window after installation, whether fifteen minutes, one hour, or twenty-four, for the conversion to count. Change the window, and the reported rate moved sharply. Engineering understood the mechanics; the business did not.

When I interviewed stakeholders, the VP of Product, the CFO, and eventually the CEO and Marketing, the same sentence came back in different words: no one fully understood how the metric worked, and no one trusted it. Yet it was the only view of the funnel the company had. Whenever the rate dropped, the whole organization worried, but no one could diagnose why. That uncertainty left the business flying blind.

The Diagnostic

The instinct in the room was that the metric was broken. It was not exactly broken. It was undefined, and it measured the wrong things in the wrong way.

Two checks separate a trustworthy metric from a suspect one. Are you measuring the right thing, meaning the number is a fair stand-in for what you care about, and are you measuring the thing right, meaning it is captured cleanly rather than distorted by how it is collected. Free-to-Paid failed both. It counted a paid-license flag in telemetry rather than an actual payment, and it began from installation rather than any agreed moment of activation, so it was not clearly the right thing. And because a hidden time window decided whether a conversion counted, the same behavior produced different numbers depending on a setting, so it was not measured right either.

Underneath sat a deeper gap: no one had ever defined the funnel as a business object. When does conversion begin, at download or at install? Should a purchase be counted from telemetry, or from the money actually arriving? These had never been answered, so the metric meant something slightly different to everyone who used it. Even a clean definition would have struggled, because the customer was scattered across systems that did not agree on who anyone was: someone submitted a form with a personal email and paid with a corporate one, and the CRM, the telemetry, and the billing platform each held a fragment none of them stitched together.

So the real work was never going to be a better dashboard. It was to manufacture a shared definition, and to build the identity spine that could carry it. You cannot fix a number until the business agrees what it is supposed to mean.

The Approach

At first, we could not settle on a definition at all. Every attempt to pin down the funnel produced more debate, not less, and the work stalled. The turn came when the CFO, the VP of Engineering, and I stopped trying to define the perfect funnel and decided to treat the analytics like a software product instead of a report. There was nothing yet for a dashboard to be correct against, so we built what we called Version 0: a deliberately imperfect baseline stakeholders could argue with and ratify, then improve. Only once we changed course, from chasing a final answer to shipping a first one, did the project move.

Version 0 was validated against an external anchor, the count of new customer logos Finance already reported. The two were never expected to match exactly, but they told us whether the funnel behaved in a believable range. It set a principle I now carry into every engagement: telemetry explains behavior, but finance confirms reality.

The stage definitions were negotiated, not assumed, worked out with the CFO as executive sponsor so the funnel carried business meaning rather than only an engineering one. For example, users often attempted several installs, some of which failed, so we agreed that the first successful installation, not the first attempt, marked activation. From there the model tightened by increments, each shipped as its own ticket: Free-to-Paid was redefined to run from a real product user to a paying one, first-touch attribution from the CRM was folded in so conversion could be read by channel, and the funnel became a true cascading cohort, where a user only enters a stage after completing the one before.

One decision mattered more than any other, though it looked small at the time. We defined each stage by the business event it represented, not by the tool that happened to record it. A stage was a form submission, an install, a first open, a purchase, defined by when that event occurred rather than by which system produced it. That single choice is what later made the funnel durable.

The definition also transferred to problems it was never built for. The flagship product had a purchase to anchor on, but other products were still in beta, with no paid tier yet. Rather than leave them without a funnel, the template asked each team to name the event that stood in for a purchase. That is how they arrived at user activation, a defined set of events marking a genuinely active user, the closest adoption signal they had to revenue.

The Built

With the definition agreed, the build followed from it:

• A validated Version 0 funnel. The first business-owned baseline, benchmarked against Finance's new-logo count to confirm it ran in a credible range.
• A unified data model. Telemetry, billing, and the CRM ingested into the new lakehouse and reconciled into one analytical model.
• Identity resolution across systems. A tracking bridge tying a form-submission email to post-install telemetry, plus logic to reconcile personal and corporate emails into a single user.
• A five-stage conversion funnel. Form submission (from the CRM), start of install, completed install, and first open (from telemetry), then subscription purchase (from billing), measured as a cascading cohort.
• A full set of conversion metrics. Not one number but many: the conversion at each stage jump, and an overall rate from form submission through to paid subscription, which turned the funnel from a single figure into a tool for finding where users dropped.

One limitation was honest and stated up front: confidence in the final numbers was never literally 100%, because the underlying systems were genuinely complex. That remaining uncertainty, though, came from the complexity, not from disagreement about the definition, which was the part that had been broken before.

The Outcome

By the end, the company had stopped arguing about the number and started reasoning with it. The funnel had become a shared definition the leadership team recognized as the source of truth, and because it was defined by events rather than tools, it held up as the tools underneath it changed.

• One trusted definition of conversion. Leadership moved from an opaque engineering metric to a business-owned funnel it collectively recognized.
• Problems the team could act on, not just spot. Stage-by-stage conversion showed where users dropped, and because each stage carried different dependencies, a drop pointed to a specific owner in Marketing, Engineering, Product, or Finance. It turned the funnel from something to report into something to optimize.
• Resilient through a wave of migrations. As the CRM, the telemetry platform, and the billing system were each replaced, the funnel kept counting, combining old and new sources, because a stage was defined by when an event happened, not by which system recorded it.
• The blueprint for every product. The funnel became the template for how new products, including ones still in beta, defined and measured their own conversion.

What had been a metric nobody trusted became a definition the whole company could stand behind, and build on.