A headless access graph: define your ICP, get back everyone who has access to it — ranked, by partner type, with receipts.
Every existing B2B database answers "who IS my ICP." This answers "who has ACCESS to my ICP."
That's not a firmographic lookup — it's a distribution graph. A partner leader (or a founder, or Hundo) describes their ideal customer, and the API returns every entity whose audience overlaps that ICP: complementary tools, agencies, communities, media, marketplaces — grouped by partner type, ranked by how strong and how warm the access is. Nobody sells this cleanly. AffiliateFinder-type tools force everyone through their UI; we expose the primitive itself, and Hundo becomes consumer #1.
"Who has access to my ICP" collapses into set math, so it's engineering, not magic:
T — pulled from rented firmographics. (We rent the company substrate; we never build it.)A(partner).T lives inside A(partner), weighted by how strong the access mechanism is.The entire product reduces to one hard problem: for any entity, estimate the set of companies it has access to. Nail that and the query is intersect-and-rank.
overlap = how much of your ICP they reach · mechanism = how directly/credibly they hand you a customer (a warm advisor intro beats a banner) · fit = non-competitive, aligned incentive to refer.
| Partner type | How we estimate their audience A(P) | Precision |
|---|---|---|
| Complementary SaaS | Reverse tech-stack index — every company running tool P (BuiltWith/Wappalyzer). Exact, named companies. | Highest |
| Integration / marketplace | Published — scrape app marketplaces (Shopify, HubSpot, Salesforce…). Literal access maps. | Highest |
| Agencies / consultants | Client lists — case studies, "we work with," Clutch / partner directories. | Medium |
| Communities | Members / topic — inferred from who participates and what stack they discuss. | Fuzzy |
| Newsletters / media | Audience estimate by topic + subscriber demographics. | Fuzziest |
Start at the top and only the top. Tech-stack overlap gives an exact, verifiable set of named companies — a "holy shit" result with zero fuzziness and no trust problem, because you can show the receipts.
ICP (from our own scoring model & playbook): a B2B founder doing ~$500K–$10M who knows partnerships should be a channel, has nobody who owns it, and sits in one of three states — greenfield (no program), tool-no-operator (installed an affiliate tool, nobody runs it), or stalled/dying. Signal-defined and finite.
Run "who has access to that buyer" and it resolves unusually sharply:
Rewardful, Tolt, FirstPromoter, Refersion, Tapfiliate, Dub, PartnerStack.
Almost absurdly perfect: their entire customer base is our "tool, no operator" ICP — by definition. Rewardful's ~9,100 live sites are the enumerable market. The overlap isn't high, it's near-total.
And the incentive is aligned, not competitive: a Rewardful program that's a graveyard is churn for Rewardful. Hundo makes their tool sticky. They have a retention reason to hand us their under-activated accounts. Mechanism: co-sell / integration referral.
Fractional CMO/RevOps & growth agencies serving $500K–$10M B2B SaaS, and HubSpot Solutions Partners. When a client's paid CAC climbs (our exact "why now"), they get asked "what else?" Mechanism: referral / white-label. Medium precision — named via directories, not full client rosters.
Key nuance: partnership communities (Partnership Leaders, Nearbound, Pavilion) only reach our stalled buyers — people who already care about partnerships. Our greenfield founders (highest-intent slice) aren't there yet. They're in founder / bootstrapped-SaaS spaces — MicroConf, SaaStock, indie SaaS communities, the Medellin AI meetup. Different access surface entirely.
The ICP's GTM stack — Clay, Apollo, Instantly, Stripe, HubSpot. Huge reach, weak precision (not every Apollo user has a channel gap). This is where the co-occurrence math earns its keep by filtering down.
Our Dub/Tolt/Rewardful signal campaigns already target these exact tool vendors' customers. The hand-built access graph independently landed on the same place. That's a validation event: the product, once built, would automatically reproduce the best GTM instinct we reached by hand — the "it found what I already know is right" test, passing on our own data.
We should be building actual distribution partnerships with Rewardful, Tolt, Dub, and FirstPromoter now. Their churning "graveyard program" accounts are the perfect Hundo wedge, and saving those accounts is the vendor's retention win. A co-sell partnership pitchable this month. The product systematizes it later; the partnership is worth doing regardless.
Not the full API. The tech-stack co-occurrence engine for one vertical, exposed as a minimal API, with us as user zero. One partner type (complementary SaaS), done with exact evidence, beats five done fuzzily. Prove "found 20 real partners with access to the 100.partners ICP, with receipts." Then widen partner types → let Hundo drive it → add the signal layer.
T (rented firmographics).The moat isn't the firmographics (commodity) — it's the partner-signal layer, the ranker, and the evidence loop: every partnership Hundo actually lands writes back a verified access edge. Over time the graph stops being inferred and becomes validated-by-outcome. AffiliateFinder can't copy that; they don't run the activation layer. We do.