The situation
A vertical SaaS platform serving a licensed profession, with a large installed base on its core product and a newer second product to sell into it. The brief carried a fixed monthly budget, a monthly qualified-lead target, and a sales cycle measured in days rather than weeks.
Dividing the first number by the second produced the constraint that designed everything else: a hard blended ceiling per qualified lead that no paid channel reaching that profession can hit. Click costs alone in that vertical can consume the entire allowance.
That single division rules out paid-first thinking before strategy is discussed. It is also the kind of arithmetic that is easy to skip, and skipping it is how a plan gets built that cannot arithmetically work.
The insight we built on
The platform’s own published research on its customer base showed a clear majority of small practices already using AI tools for professional work — and less than a third reporting revenue growth from it, against a far higher figure among large firms. A large majority had changed nothing about their pricing after adopting it.
The reading: this segment is not AI-hesitant. It is AI-active in the wrong tools — consumer chatbots with no confidentiality guarantee and no citation checking. The distance between everyone uses it and few profit from it is the warm addressable market for a professional-grade version.
So the pitch is never “try AI”. It is: stop doing unpaid, unprotected AI work in a tool that does not safeguard your clients or check its sources.
The most distinctive move: write the message twice
Practices in this profession split their work across two billing models, and most firms run both. The same time-saving tool means opposite things to each.
On hourly work, saving time reduces billable revenue, so efficiency is not a benefit. The pitch there is capacity: more client work, same hours.
On fixed-fee work, saved time is pure margin, so efficiency is the pitch directly.
Almost nobody segments on this. We made billing mix a first-class targeting signal, which was available in the core product’s own data and required no new tooling.
Segmentation on behaviour and economics, not firmographics
Three priority segments, each defined by a signal already present in the core product rather than by firm size or location: work-type and document volume for the research-heavy practices, billing mix for the fixed-fee practices, and hourly dominance for the third — with the message assigned to the segment rather than to the channel.
Programme design: the product fires the trigger
An always-on engine with monthly spikes, structured as a four-step journey. A felt-effort event occurs inside the core product — an open case file crossing a document threshold, repeated late-night sessions. An in-product card appears in context at that exact moment, never a banner and never a scheduled campaign. A lifecycle sequence follows. Escalation to a human play where it is warranted.
The principle is borrowed from how the best product-led businesses actually grow a second product: they do not run cross-sell campaigns, they instrument the product so the offer appears when the need becomes felt.
The copy carries the customer’s own number back to them — the actual document count on the file in front of them — a specific and checkable promise, and the confidentiality clause in the final sentence.
Where the money went, and why
Owned surfaces carry the majority of the volume at zero campaign cost. Zero is the honest number here: it runs on the stack already paid for, and the real cost is headcount and roadmap, not media. Every figure in the channel plan was built bottom-up from a movable assumption — addressable accounts, monthly trigger incidence, prompt engagement, hand-raise conversion — rather than reverse-engineered from the target.
The budget then buys only what cannot be built in-house: accredited education with filing fees and practitioner honorariums, customer-match retargeting capped at the point where a finite matched audience hits its frequency ceiling and further spend buys annoyance rather than leads, peer-proof content that assists more than it sources, and a professional-association pilot as a trust layer. A deliberate unallocated reserve funds the recovery of a downside month.
The arithmetic proof sits in the model: a paid-only plan lands at roughly two and a half times the ceiling. That is why owned surfaces must carry the majority, and it is a finding rather than a preference.
We also wrote down what we chose not to fund and why. Programmatic and IP-match display were skipped because the audience is a customer list — IP resolution adds cost without adding precision. Declining a channel is a decision with a named reason, not an omission.
Incrementality over attribution
A global holdout, untouched, sized so that a single month of data can detect a meaningful relative difference in qualified-lead rate. We would rather know whether the programme created demand than argue about which touch to credit for it.
The whole plan was stress-tested by simulation with every rate wobbled in both directions, so the downside case had a number and the reserve had a job.
The finding that reordered the priorities
Ranking the drivers by influence showed that the two handoff conversion rates dominate every channel variable in the model. The programme lives or dies in the handoff, not the media.
So the handoff got its own design rather than a footnote: a one-business-day response commitment agreed in writing before launch, because in a sales cycle measured in days a lead touched on day two is already competing with someone else’s day-one touch. And a context handoff, where the lead arrives with its trigger story attached — which event fired, on which work, and which message they saw — so the call continues a conversation instead of starting one.
Launch sequence
Instrument first: agree the qualified-lead definition and the routing commitment with sales in writing, build the segments from existing product data, baseline the dashboard, secure the in-product surface with the product team, brief the legal reviewer on claims. Then ramp through the quarter — in-product first, lifecycle on, education, paid layered, association pilot last.
Definitions before dollars. Instrumentation before spend.
The risk we led with rather than buried
A second product can substantially increase the total bill for the segment that spends least on software. We put price-stacking in the risk slide rather than hoping it would not come up, with consolidation economics as the argument, annual and bundle framing as a live test, and return-on-investment arithmetic in every asset.
Status
Delivered inside the last twelve months. We are not publishing outcome figures — the results are not in, and we do not publish numbers we cannot stand behind. The research cited above is the client’s own published industry study, quoted at the level it was published. No client figures, account data, product names or identifying details appear on this page.