How Canva took routable inbound from 11% to 68% by turning Clay into its GTM platform

Canva grew from 220 million to over 240 million monthly active users during this engagement. The B2B motion was scaling into that base, and around 3,300 marketing engaged leads were arriving every week from events, webinars, content downloads and product signups.

The growth team could not qualify that volume fast enough. Most leads landed in Salesforce without the fields needed to route them, so more than half were dropped before they reached Sales. The leads that did get worked were picked out of lists by hand.

Canva already had Clay. They were using it the way most companies do, as a one-time waterfall: upload a list, get emails back, move on. Nobody was using it as infrastructure.

Qualification and routing depend on a base set of firmographic fields. Miss one and the lead cannot be assigned to anyone. It is not rejected. It just sits there.

Only 11.4% of inbound leads arrived with the full set.

The reason is structural. Canva serves large consumer and business audiences at once, so the lines between the two funnels blur, and whether a given lead is pursuing a B2B use case is genuinely hard to detect. Canva also deliberately keeps web and product interactions frictionless, which means less profile information is gathered explicitly at the point of capture.

We built the resolution loop to handle the three ways a lead actually arrives: work email with a matching account, work email with no match, and personal email. The third case is the hard one, and at a PLG company it is most of the volume.

A waterfall of ten providers cascades behind it, ordered by hit rate and cost per successful match rather than by contract. Profile discovery runs first, work email resolution second, and the remaining providers fill the gaps.

Three decisions shaped the build. Account matching runs on contains rather than exact domain logic, so subdomains resolve to the right company. Vendor source tracking records which provider produced which field, so the team can audit its own data. Seniority maps against Canva's existing value set rather than being parsed from free text, so it matches what Salesforce already expects.

Routable leads went from 11.4% to 67.7%.

At 3,300 inbound leads a week, that is roughly 1,900 more every week that can now be scored and assigned. Close to 100,000 a year that previously went nowhere.

Canva asked how far the enrichment actually reached, so we ran a full audit on a batch of 34,196 registrations and analysed it field by field.

Company-level resolution held up. Domain came back at 100%, employee count at 87%, industry at 86%, country at 83%.

Person-level resolution came back at 31%.

Our first theory was that missing company names on the inbound webhook were starving the enrichment. The data did not support it. Only 0.3% of records carried a company name at all, so it could not be the cause. The real driver is that free-tier registrants use personal domains that no data provider has on file. Even leads matched to an existing Salesforce account only reached 35%.

Canva now knows where that ceiling sits and has stopped spending against it.

With resolution solved, the work moved up a level.

Entity resolution moved to the account level, so every play reads one definition of a company. Company classification consolidated into a single shared function across ten canonical values. A high-fidelity table resolved master name, city and description, so different plays stopped disagreeing about the same account.

On top of that, Canva now runs a live signal library across 63,000 companies, refreshed on a per segment cadence and activated through the same shared audience layer.

Enrichment became a shared asset rather than a per play rebuild, so the same question returns the same answer everywhere.

The same layer feeds paid media. Signal-qualified segments push to LinkedIn from the definition that feeds sales, so ad targeting and lead routing work from the same account list.

Personalized outbound to every marketing engaged lead globally was not a resourcing problem at this volume. It was not possible at all.

We built it with Relevance AI agents on top of the Clay foundation.

Canva measured it on edit rate rather than on opinion. Above 10%, the prompts get fixed. Below 10%, the copy ships and reviewers stop reading every message.

After three iterations, over 90% of generated copy shipped unedited. The review gate stayed open rather than being switched off, so quality drift shows up as a rising edit rate before it shows up in reply rates.

The engagement is now transitioning to Canva's internal team, with named owners on the enrichment tables and the ABM signal logic.

They own:

• Clay as a platform, with one account definition read by sales, ABM and paid
• Enrichment as a shared asset rather than a per-play cost
• A live signal library with documented triggers and downstream mapping
• A failure register and alerting runbook for the enrichment tables
• Written handover documentation for the enrichment and ABM signal logic

The edit rate matters most in the handover. It lets the team see whether the system is still working without needing to ask us.