Zen Intent · Unit economics

What a lead costs, and what we charge

Two suppliers we already pay for, SerpAPI and Hunter, and one we are proposing to add, Apollo. What a lead costs to produce today, how the credit model turns that into revenue, where our current plans run out, and what Apollo would change.

Before we start

Seven words used throughout, in plain English.

One more, because it appears on a few slides. The pipeline is the automation that runs all these lookups in the right order for every lead, without anyone pressing a button. It is built in a tool called n8n.

Our model

We do not sell leads. We sell credits, and leads are what they buy.

The shape of it: recurring credit revenue, a cost of goods that steps up in tiers rather than rising per lead, and cash collected at request time rather than delivery. Acquisition is the free Welcome Gift, which hands over 42 leads worth of data before anyone pays.

One caution on that 95%. It counts revenue minus the data suppliers minus card fees, and nothing else. It excludes hosting, the AI pass, salaries, marketing and support, so it is a ceiling rather than take-home profit. There is a slide on this later.

The business model

We sell credits. A lead costs 35 of them.

1. Customer buys credits, as a monthly plan or a top upRevenue lands here
2. They request leads. We deduct the full cost the moment they ask, 35 credits a lead, or 60 if they tick intelligence up frontBalance drops immediately
3. We deliver fewer than asked. The shortfall is refunded, so they only pay for leads that arrivedCredits returned
4. The run fails. Every reserved credit goes backFull refund
5. They unlock intelligence later on a lead they already own. We deduct the 25 credit difference, not another 60Second bite

Deduction happens at request time, not delivery, and is held against the request as reserved credits. That is what protects us: we are paid before we spend anything with SerpAPI or Hunter, and we refund only what we failed to deliver. Free tier users and admins bypass the ledger entirely.

Top ups list at 10 cents a credit and each plan discounts them, from 10% off on Starter to 32% off on Enterprise. That ladder is the pull toward upgrading: a Starter customer topping up to Growth volume pays $368, while Growth itself is $347.

Free credits we issue

The Welcome Gift hands over $127 of product for $2.23.

The gift is granted once and flagged, so it cannot be claimed twice, and it lands in the same credit ledger as paid credits, which means the normal deduction and refund rules apply to it. On the books it is a 57 times ratio: we give away $127 of retail for $2.23 of real cost.

In acquisition terms that is cheap. One month of a Growth subscription pays for 155 gifted signups. The risk is not the cost, it is that 42 free leads may be enough to satisfy a casual user without ever converting them.

Everything on the invoice

Seven suppliers, and the biggest bill is the card processor.

Not on the list, and worth saying so: GoHighLevel has no credentials configured, so it costs nothing today. Meta CAPI, Google sign in and Slack alerts are all free tiers. Cache, queue and sessions all run on MariaDB, so there is no Redis line either.

Stripe is the largest single cost in the business, ahead of both data providers combined. Three of the seven have never been measured, and Gemini is the one most likely to matter.

What runs today

What we look up for every single lead.

The highlighted row is the one that drives the SerpAPI bill. Place details runs once per lead, not once per search, so it is twenty times the discovery cost. Gemini interprets what the other six fetched; it looks nothing up.

Per lead

Six cents of data inside a lead we sell for about three dollars.

Hunter credits assume 20% of leads take the intelligence unlock, which adds the person enrichment call. At 100% unlocks it is 3.5 credits and about seven cents a lead. Hunter charges nothing when it finds no email, so the real figure sits below this.

For ops

The upgrade staircase, in leads a month.

The highlighted row is where we are now, and it runs out at 3,700 leads a month, about 31 Growth customers. Hunter binds before SerpAPI does, at 9,990 of its 10,000 credits. Each lead burns 2.7 Hunter credits and 1.05 SerpAPI searches, so multiply this month's leads by those and check both dashboards. Above 9,250 leads Hunter has no published tier and needs a quote.

The catch on growth

Customer 32 costs us $150 the moment they sign.

Data cost does not rise per lead, it jumps in steps. At 3,744 leads a month Hunter moves from Growth to Scale, $149 to $299, and the jump lands in full the month it happens. Against $347 of new revenue that is 43% of the customer we just won, then it amortises away across the next several.

Two consequences. Any pricing change that adds volume should be timed with a tier headroom check, not launched blind. And the honest way to read the margin table is that margin dips every time we grow into a new tier, then recovers. It is not a warning sign, it is the shape of the cost base.

Unit economics

The intelligence unlock is the best line we sell.

A base lead costs us 5.3 cents. An intelligence lead costs 6.8 cents, because the only difference is one extra Hunter call. So the unlock charges 71% more and costs us 28% more, which is why its margin beats the base lead.

Every credit we sell is worth more than every credit costs, at every plan tier. The cheapest tier per credit, Enterprise at 6.7 cents, still keeps 97.7% of a base lead. There is no volume at which the data bill threatens the model.

Projection

Stripe takes more than SerpAPI and Hunter combined.

At our current ceiling of 31 customers, Stripe costs $321 against a data bill of $224, so payment processing is 1.4 times the thing this whole deck has been arguing about. Annual billing would cut it, since 2.9% plus 30 cents lands twelve times a year on monthly plans and once on annual.

Two costs are still missing. The free Welcome Gift hands over 1,500 credits, which is 42 leads and about $2.23 of data per signup that claims it, and Gemini charges per AI pass, which we have never measured. Neither changes the shape, but they belong in a real P and L.

Read this before quoting the 95%

95% is what the product keeps, not what the business keeps.

So the useful way to read 95% is as a ceiling. It says the data and the card fees are not what will stop this business being profitable. It does not say we keep 95 cents of every dollar, and nobody should quote it as if it did.

Where the money goes

Moving to bank debit saves more than we spend on data.

Bank debit is 0.8% capped at $5, against 2.9% plus 30 cents on cards, so on an annual charge the cap does almost all the work. The intuitive move is annual billing, but on its own it saves only $3.30, because the percentage still applies to the same total and all we drop is eleven 30 cent charges. The lever is the payment method, not the billing period. Combining both is what takes $124 a year down to $5.

Pricing

We could charge a third of today's price and still hold 90% margin.

We charge 56 times what a lead costs to produce, so cost tells us almost nothing about what to charge. Pricing here is a market and value decision, not a margin one. What cost does tell us is that discounting is close to free, and that any price we pick will still clear 90%.

The one thing that genuinely limits a price cut is not margin, it is the tier staircase from earlier. Cheaper prices win more volume, and volume is what actually costs us money.

Pricing levers, ranked

Four moves, in order of how much they return.

The first two are close to free money and need no pricing change at all. Bank debit alone returns more than our entire annual data bill. The intelligence unlock is already built and is the highest margin line we sell.

Repricing sits last deliberately. At 56 times markup we can afford it, but it is the only move on this list that cannot be undone quietly.

The pricing question

No, do not change the prices. They already work.

A price change is how you answer losing deals on price. We have not recorded that happening. Changing the one number we have no evidence about, while two free improvements sit untouched, is the wrong order.

Do these first instead. Offer payment by bank transfer, which returns $119 per customer a year and needs a settings change, not a pricing decision. Then push the intelligence upgrade, which is already built and is the most profitable thing we sell. Revisit price only if we start losing deals over it.

Side by side

What one lead carries now, and what it would carry with Apollo.

The first six rows do not move. Apollo has no access to Google Maps, and it cannot write the pain point analysis, so everything that makes our lead a local lead is unaffected. Three rows change, and only one of them is genuinely new.

The 6 in 10 is an assumption, not a measurement, and it drives the whole right hand column. If Hunter is really at 8 in 10 the gain shrinks to 259 contacts. If it is at 4 in 10 it grows to 777.

What Apollo would add

Seven of nine capabilities are already live.

The direct phone number used to be the argument for Apollo. Hunter's person enrichment already returns it, along with LinkedIn, title and seniority. That leaves two rows. Topic intent, which is thin for local trades, and searching for a person before we know their company, which is the one that matters and gets its own slides shortly.

But this table is about what each service is capable of, not how often it succeeds. Those are different questions, and the second one is where the real case for Apollo sits. That is the next slide.

The real case for Apollo

The question is not what Hunter can do. It is how often it succeeds.

The previous table shows Hunter is able to find an owner's email. It says nothing about how many times out of ten it actually does. Nobody has measured that. If Hunter finds six in ten, four in ten of our leads ship without a contact, and a second source would pick some of those up.

Read the direction carefully, because it is the opposite of intuition. The worse Hunter is, the more Apollo is worth. If Hunter is only finding three in ten, Apollo adds 24 points of coverage for ten cents a contact. If Hunter is already at nine in ten, Apollo adds three points and is hard to justify.

We do not know which row we are on. That single unmeasured number decides the whole proposal, and it is measurable from leads we have already delivered.

The bigger use for Apollo

Person first leads. Neither supplier we have can do this.

Everything so far assumed Apollo enriches a business we already found on Maps. The other way to use it is the opposite direction: search for people by job title, seniority, company size and location, and get the person and their company back together.

This is the strongest argument for Apollo in the whole deck, and it is not enrichment. It is a second product line beside the local business lead, with its own data shape and its own scoring.

What a person lead costs

Searching is free. The phone number is 8 times the email.

A person lead with just an email is cheaper than what we sell today, because one Apollo call replaces the Maps detail lookup and three Hunter calls. Add the phone number and it becomes five times dearer than a local lead, though still a 91% margin.

This also corrects an earlier slide. I costed Apollo filling missing contacts at $81 a month on the assumption of one credit each. If we want direct dials too it is $205 a month, because phones are 8 credits. Filling missing emails and adding direct dials are two different proposals with a 2.5 times difference in price.

What Apollo would cost

A seat before the first call, then pennies.

In plain terms: Apollo will not let us test whether their data is any good for our kind of business until after we have started paying.

Cost is not the obstacle. At 3,700 leads Apollo moves margin from about 98% to about 97%, which is nothing. The obstacle is that we would be paying for a second source of data we already collect.

What Apollo would let us sell

One new feature, and it needs 32 unlocks a month to pay for itself.

That is a low bar. One customer unlocking intent on a third of their leads would clear it. The catch is the third figure: topic intent is built for companies that generate trackable research behaviour, and a two truck plumbing firm generates almost none. If coverage is near zero the feature cannot be sold at all, and this is measurable only after we have paid for a seat.

Run your own numbers

Total monthly cost, both subscriptions included.

The proposal

One month of Apollo, with a decision rule agreed up front.

The whole risk is one month and between $81 and $205, depending on whether we reveal phone numbers as well as emails. Even the higher figure is under 2% of monthly revenue at 31 customers. Deciding this by argument costs more in meeting time than deciding it by measurement.

Recommendation

What we are asking for.

1. Pull Hunter's real fill rate from live leadsData we already have
2. Confirm the owner's email and profile link are actually being saved, not quietly discardedCode check
3. Plan the Hunter Scale upgrade for 3,700 leads a month$299, the real next spend
4. Only then trial Apollo on 25 real businessesOne month of Basic

Steps one to three cost nothing and are worth doing whether or not Apollo goes ahead. Step four is the paid trial, and the decision rule for keeping or dropping it is agreed before we start rather than argued afterwards.

If it goes ahead, one more thing needs an answer in writing first: Apollo licenses its data for internal use and restricts redistribution. We hand records to paying customers, so that needs clearing with Apollo before any build.

Assumptions

What is measured, and what still needs confirming.

Two of these decide the whole question: Hunter's real hit rate, which we can measure today at no cost, and Apollo's match rate on local trades, which we cannot measure without paying. That asymmetry is the argument for doing the free measurement first.