The lead you bought was also sold to three other companies

The notification arrives on the van's dashboard mount at twenty past ten: a new lead, a water heater replacement, two streets from the job the crew is finishing. It cost fifty-two dollars. By the time the plumber has washed his hands and called the number, the homeowner has already spoken to two other companies, one of which is on its way. He leaves a voicemail. The fifty-two dollars is gone either way.

That scene, repeated a few hundred times a year, is the economics of the home-service lead marketplace. Angi, HomeAdvisor, Thumbtack and their smaller relatives do something genuinely useful: they collect people who need a plumber, a roofer or a cleaner and put them in front of businesses that want the work. What they charge for is the introduction. Whether the introduction becomes a job is left, entirely, to the business that paid for it.

A marketplace sells you a chance to compete for a job, and prices it like a customer. The difference between those two things is where the money goes.

What a lead marketplace actually sells

A home-service lead marketplace collects service requests from homeowners and charges the providers it matches them with, usually per request, whether or not the provider wins the job. Many requests are sent to more than one provider. The product is access to a homeowner at a moment of need, not a booked job.

It helps to be precise about this, because the language on both sides blurs it. Providers talk about buying leads as if they were buying customers. Platforms talk about connecting you with homeowners, which is accurate, and about the jobs their pros win, which is also accurate on average and tells you nothing about your account.

The mechanics vary between platforms and change often, but the shape is consistent:

• A homeowner describes a job: the service, the location, sometimes a timeframe and a budget.
• The platform matches the request to providers whose categories, service area and settings fit.
• Each matched provider is charged, typically a fee that varies with the size and type of job, the market and how much competition there is for it.
• The provider contacts the homeowner, or the homeowner contacts the provider, and a quote may or may not follow.
• If a job happens, the money for it is paid to the provider and recorded in the provider's own systems, not the platform's.

That last step is the one that matters for measurement. The platform knows the lead. It knows the fee. It may know whether you replied, and on some platforms whether the homeowner marked you as hired. It does not know what you invoiced, whether the invoice was paid, or whether the same household came back in March for a second job. You know all of those things, in a different system, under a different identifier.

The arithmetic of a shared lead

When one request is sold to several providers, each provider's chance of winning it is roughly their share of the competition, adjusted for how fast and how well they respond. A lead shared with three others is, for an average responder, about a quarter of a direct enquiry — before any question of quality.

Shared leads are not a trick. The homeowner asked for quotes and is entitled to compare them. But a provider who budgets as if each lead were a prospect of their own will be wrong by a factor equal to the number of competitors on it, and nothing in the lead report corrects for that.

The second row is the one to dwell on. Being average on a shared lead is expensive. Being first is not a small edge; it is most of the result. The research on how quickly enquiries go cold is old now, and has never been seriously contradicted.

On a shared lead, speed of response is not an operational detail. It is the price of the lead, divided by your chance of winning it.

Not every lead is the request it appears to be

A lead can fail before anyone competes for it: a number that does not connect, a job outside your service area or trade, a homeowner who was only researching, or a request that did not come from the platform's own site at all. The only reliable record of how often that happens is your own call log and customer file.

Platforms have every reason to send good leads, and most leads are what they say they are. But the incentive runs one way on the margin. A borderline request — the wrong postcode, a vague job description, a homeowner who ticked the box for "just planning" — is revenue if it is sent and nothing if it is withheld. The provider pays the cost of the borderline cases, and without a record of how each lead actually went, nobody notices how many there are.

None of this requires assuming bad faith. It requires keeping a reason against every lead that went nowhere, because "it didn't work out" and "the number was disconnected" are different findings with different remedies. The first is competition. The second is a credit you should have claimed.

• Unreachable: the number is disconnected, wrong or never answered after several attempts on different days.
• Mismatched: the job is a service you do not offer, or the address is outside the area you set.
• Not ready: the homeowner says they were researching, or have already hired someone.
• Duplicate: the same household appears more than once, from the same platform or another.
• Already yours: the person is in your customer or lead records from before the marketplace introduced them.

Tag each unconverted lead with one of those, and after a quarter the pattern is usually obvious. A category where half the leads are mismatched is a settings problem or a platform problem. A category where most leads were reached and quoted but lost is a pricing or speed problem. They look identical in a cost-per-lead report.

The platforms change the product underneath you

Marketplaces adjust how requests are matched, how many providers receive each one and what counts as a lead, often in response to their own economics. A provider's results can move for reasons that have nothing to do with the provider, and only a month-by-month record of cost per paid job will show it.

This is the part of marketplace spend that surprises people who have run it for years. The channel is not a fixed thing. The rules for who gets a request, and how, are product decisions made by a company with its own targets, and they change.

If your cost per paid job from a marketplace moves sharply in a month when your own prices, crews and response times did not, the platform probably changed something. That is worth knowing, and worth asking about. You can only notice it if you were measuring the outcome in the first place.

The lead who was already yours

A marketplace charges for introducing a homeowner even when that homeowner is already a customer or an open lead of yours. Matching marketplace leads against your existing records on phone and email shows how much of the spend paid for introductions you did not need.

Homeowners do not use one route at a time. The woman who had you service her boiler two winters ago may not have your number to hand when the kitchen tap starts leaking; she opens an app, describes the job and, if you are on that platform in her area, you may well be one of the providers she is matched with. You pay to be introduced to your own customer.

The same happens within a single purchase. A homeowner searches, clicks your Google ad, looks at your site, does not call; then fills in a marketplace request that evening and receives your number among three others. You paid for the click and for the lead. Which of them earned the job is a matter of opinion; that you paid twice is a matter of record.

Neither case is a scandal. Both are measurable, and together they are often a larger share of marketplace spend than providers expect. The check is the same one that applies to affiliate commissions: for each marketplace lead, was this phone number or email already somewhere in your customer file or your lead history, and if so, since when?

The cheapest lead a marketplace sells you is one for a customer you already had. It is also the one you should not be buying.

Pricing a marketplace by the jobs it produced

Cost per paid job for a marketplace is total spend on the platform, including subscriptions and fees, divided by the number of marketplace leads that became invoiced and paid jobs. It is found by matching the platform's lead list against your job records on normalised phone number and email.

The method needs two lists and some patience with phone formats. It does not need an integration, an API key or a developer, and it works on months that are already over.

1. Collect every lead you were charged for in the period: from the platform's lead history or export where it offers one, otherwise from the leads screen or billing statement. Keep the date, the category and the homeowner's phone number and email.
2. Export the jobs you invoiced from your field-service or accounting system for the same period plus the following two or three months, because quotes take time: customer phone and email, invoice date and amount, and whether it was paid.
3. Normalise every phone number to one format — country code, no spaces or punctuation — and lower-case every email. Most failed matches are formatting, not absence.
4. Match leads to jobs on phone or email. Treat an exact match on either as a win; set weaker matches, such as the same surname at the same street, aside for a person to check.
5. Add up everything paid to the platform in the period: lead fees net of any credits, plus membership, subscription or advertising charges.
6. Divide that total by the number of leads that became paid jobs. Then do it again for each category and, if the volume allows, for each area.

Here is what the result tends to look like. The figures below are illustrative, constructed to show the shape rather than taken from any account, but the pattern — one category carrying the others — is the common one.

Three things stand out that a cost-per-lead report could never show. Leak and drain leads are the cheapest per lead and the cheapest per job, yet they bring in the least money per job, which makes them a volume business that depends on how fast the crew can turn them. Water heater leads cost more each and are the best buy on the page. And the single remodel that closed looks brilliant in isolation and is a poor use of money across the quarter, because one job in twenty-one is not a channel, it is a coincidence.

CloseRev is built to do this reconciliation. It matches a lead or call export against your closed jobs on normalised phone and email, counts only high-confidence matches automatically and shows both records for every match, so a marketplace's lead history can go in as a source beside your ads and call tracking. Jobs it cannot tie to any source stay in Direct / Unknown rather than being credited to whichever channel is closest.

What to do with the answer

Once each category has a cost per paid job, the decisions are concrete: stop buying categories and areas that cost more per job than the job is worth, cap the price you will pay per lead in the rest, fix response time where competition is decided by speed, and dispute the leads that failed for reasons the platform's policy covers.

Two of those changes cost nothing. Response speed is the largest lever on shared leads and it is entirely yours; a business that routes marketplace notifications to whoever is free to call within five minutes, rather than to the owner who will get to it after the job, will see its cost per paid job fall without buying a single extra lead. And disputes, where the platform allows them, recover money that is simply owed. Both are much easier to do well when the reason for every failed lead is written down.

The comparison that finally settles the budget is across channels, not within one. A marketplace at $360 per paid job is expensive or cheap only relative to what a paid job costs through Local Services Ads, paid search, your Business Profile or direct mail — measured the same way, against the same invoices. The marketplace's own dashboard cannot make that comparison, and neither can any single platform's. Your job records can.

What the reconciliation cannot settle

Matching leads to jobs shows which marketplace introductions became paid work and what each cost. It cannot show what the marketplace's profile, reviews and visibility contributed to jobs that arrived by other routes, and it cannot show what would have happened had you not been on the platform at all.

This matters most for businesses with a strong marketplace profile. Homeowners read reviews on these platforms and then search for the business by name, or save the number and call later. That work arrives as a direct call or a branded search, is recorded as such, and the marketplace gets no credit for it in any report — yours or theirs. It is real, and it is not measurable from records alone.

The honest way to estimate it is to change something and watch: pause the lead spend in one area for a month while keeping the profile live, or pause the profile, and see what happens to direct and branded-search enquiries there compared with an area left alone. It is a rough experiment, and seasonality and weather will muddy it. It is still better evidence than an assumption in either direction.

What the reconciliation gives you, without any experiment, is the part that is not a matter of opinion. For every lead you paid for, you will know whether it became a paid job, what that job was worth, and — for the ones that did not — why not. That is enough to stop paying for the categories that do not work, to claim credits you were owed, and to go into the next pricing conversation with the platform holding your own numbers rather than theirs.

A marketplace can tell you what a lead cost. Only your invoices can tell you what a job cost, and the job is the only thing you were ever trying to buy.

Questions people actually ask

Are Angi and Thumbtack leads worth paying for?

For some businesses, in some categories and postcodes, clearly yes; for others they lose money on every job. The platforms cannot tell you which, because they only see the lead. You can, by matching the leads you were charged for against the jobs you invoiced, on the customer's phone number and email, and dividing what you spent by the number of paid jobs. That figure, compared with the same figure for your other channels, is the answer.

What is a shared lead?

A request from one homeowner that is sent to, and charged to, several service providers at once. Each provider pays for the introduction; usually only one gets the job. It is not dishonest — the homeowner asked for quotes — but it means the price of a lead is not the price of a customer, and the gap depends on how many competitors received the same request and how quickly you respond compared with them.

How do I work out cost per job from marketplace leads?

Export or copy the list of leads you were charged for, with each lead's phone number and email and the date. Export the jobs you invoiced over the same period and a few months after, with the customer's phone and email and the invoice amount. Normalise both sets of phone numbers to one format, match them, and count the leads that became paid jobs. Divide total marketplace spend by that count. Do it by category if you can, because the answer usually differs sharply between them.

Should I dispute bad leads with the platform?

Yes, where the platform's policy allows it, and with evidence. The strongest evidence is from your own records: a number that was disconnected when you called it, a request for a service you do not offer, or a customer already in your system from months earlier. Keep the reason with each lead you dispute, because a pattern of the same fault in one category is worth raising in its own right.

Why would a marketplace lead already be my customer?

Because homeowners use several routes at once. Somebody who has had you out before, or who called your own number last month, may still fill in a marketplace form when a new job comes up, and you are charged for the introduction to a person you already knew. Matching the lead's phone and email against your existing customer and lead records shows how often that happens.

Does a marketplace profile bring in work that its lead report does not show?

Often, and it is hard to measure. Homeowners read reviews on a marketplace and then search for you by name or call your own number, so the job arrives as a direct call. You cannot prove that link from the marketplace's report or from your own. What you can do is watch whether direct and branded-search enquiries fall when a marketplace profile is paused, which is the nearest thing to an experiment most small businesses can run.

Found this useful?

• Share on X
• Share on LinkedIn
• Share on Reddit
• Share on Email