What a B2B Lead Generation Agency Actually Buys

A B2B lead generation agency sells a decision, not a list. Clients could buy data themselves and most have tried. What they pay for is somebody who will decide which companies are worth contacting, write to them, and keep the sending working. The purchased inputs are commodities. The judgement layered over them is the service.

The disclosure belongs at the top, because it should change how you read everything below. We build Keaz Signals, which sells ecommerce leads with dated buying signals attached. One of the line items in this post is a thing we sell. We are not a neutral guide to your cost base, and there is no point pretending otherwise.

The phrase itself reads two ways, and the two readings want different articles. If you are a brand looking to hire an agency, this is not the one. If you run the agency and are deciding what belongs in your own delivery stack, this is written for you. The neighbouring case, an agency whose own buyers happen to be online stores and which is filling its own pipeline rather than a client's, is covered separately in agency lead generation when your buyers run stores. This post stays on the delivery side: what the service costs you to run.

That distinction matters commercially. An agency filling its own pipeline buys for one customer profile, which it chose. An agency running lead generation as a service buys for whatever profile the next client walks in with, which is a much harder purchasing problem and the reason so much of the stack ends up generic.

Five things get bought, and every agency buys all five whether or not it has named them. A source of companies. Contact data for the people inside them. Mailboxes and domains to send from. Software to run the sequence and catch the replies. And people, who are almost always the largest line by a wide margin.

1. The pool. Wherever the companies come from. Clients think this is what they are paying for; it is the cheapest line to replace.
2. The contacts. Bought separately from the pool and metered differently, which is why the two costs resist comparison.
3. The sending surface. Domains, mailboxes, warmup. Cheap per unit, and the only line whose failure is discovered weeks after the damage.
4. The sequencer. Schedules steps, throttles volume, threads replies. Mature, and largely interchangeable at the level most agencies use it.
5. The people. Whoever qualifies, writes, answers and books. Every vendor treats this as out of scope, and it is the line the other four exist to protect.

The data line gets mispriced because agencies compare it on the wrong axis. Records per month is the number on every pricing page, and it is the number that matters least. What decides whether the data earns its cost is how recently each row was true, and almost nothing sold by the record tells you that.

A record is accurate on the day it was collected and silently stops being accurate afterwards. Nothing in the file announces the moment it went stale. So an agency negotiating hard on price per record can win that negotiation and still be sending against a file whose age it cannot state, which is a worse position than paying more and knowing.

The bundling makes this harder to see. A product that finds companies, appends addresses and sends the email looks like one purchase and is three, priced as one, with one of the three usually much weaker than the others. Unbundling your own invoices is the fastest way to find out which job you are actually underbuying. We took the same knot apart at product level in what sales prospecting tools actually do.

The build-it-yourself instinct is reasonable here and worth costing properly rather than dismissing. Writing a scraper is a weekend; keeping one current is a standing job that nobody puts on the invoice, which we walked through in build a lead scraper, or buy the panel.

Sending infrastructure is the one line an agency cannot genuinely outsource, because the asset being spent is reputation and reputation attaches to whoever owns the domain. You can rent the software. You cannot rent the consequences of what goes out through it, and a vendor pooling your sending with strangers is deciding your risk on your behalf.

That is why the arrangement worth insisting on is boring: your domains, your mailboxes, your warmup schedule, your sequencer account. On our side, campaigns are created in the customer's own Instantly workspace and status and replies sync back, so the sending identity stays with the agency rather than with us. It is a deliberately unglamorous design and it is the part clients never ask about until something has already gone wrong.

For an agency running several clients this has a second consequence. Separate clients want separate sending identities, so the infrastructure line scales with the client count rather than with volume. That is unwelcome arithmetic and it is real: doubling your roster roughly doubles this line even if total sends do not move.

The sequencer sitting on top of all this is a solved category and a poor place to look for advantage. It runs the sequence well and is usually silent on who should have been in it, which is the argument we made in what a sales engagement platform does.

The sixth line item is timing: deciding which accounts get an email this week rather than eventually. Nobody sells it as a category, so it defaults to a person applying judgement, or to nothing at all. It is also the line that decides whether the other five were worth paying for.

The distinction underneath it is between a state and an event. A state is a durable fact: this company runs a particular platform, sits in a particular country, is a particular size. An event is a dated fact: it did something eleven days ago. A state can tell you a company is worth contacting one day. Only an event can tell you that today is the day.

This is why filter-based targeting wears out. If four agencies describe the same segment the same way, all four are looking at the same accounts on the same Monday, and the person receiving the fourth near-identical email is not being rude when they ignore it. A state field is not scarce, so it cannot be the differentiator.

Events also cannot be bought retroactively. They exist only if something was watching before they happened, which means the valuable asset is the history rather than the query. Whatever nobody recorded last month is gone. The longer version of that argument, aimed at bought data, is in what a sales intelligence platform knows, and what it cannot.

What we sell is that sixth line item, for one kind of account list only. The condition is simple and it is absolute: your clients have to be selling to online stores. If they are, we watch 4.1 million Shopify and WooCommerce stores across Europe and North America, re-scraped weekly, as of August 2026. If they are not, nothing below applies to you.

Because the weekly read is stored rather than overwritten, what comes out is a store plus what changed about it recently: new Meta ads running, active ad count rising, a product launch, an email marketing tool installed, social growth, newsletter activity, storefront and site changes. The current list is the buying signal catalogue, and the pool beneath it is described on the ecommerce leads database page.

Two of the five invoice lines collapse into that. Contacts arrive with the store, including founder addresses that are not publicly listed, so enrichment is not a separate purchase. Segments are built once with country, platform, follower and other conditions, with a live audience estimate as you narrow them, and then watched rather than re-exported. Access is capped per market, which is the mechanism that stops the segment you work being sold to three competing agencies the same week.

The writing step is optional. Our sales agent drafts per-store copy from the signal plus a knowledge base you write once, but the principle holds whether the drafting is done by software or by a junior with a template.

For most B2B lead generation agencies we are the wrong purchase, and the reason is structural rather than a matter of degree. An agency sells to whoever its clients sell to, and our pool covers one slice of that. If your roster is mixed, we solve part of one client and none of the rest, which is a poor shape for a stack decision.

• Your clients sell to anyone other than online stores. Then this is not a partial fit; there is nothing here for you.
• Stores on other platforms, or outside Europe and North America. Ours is Shopify and WooCommerce, in those two regions, and we will not present a thin slice as coverage.
• You need the data as a file. We do not export. Leads move into campaigns and stay there. If you hand lists to clients, or your CRM is the system of record, that rules us out and it should.
• You need LinkedIn. There are no LinkedIn signals and no LinkedIn sending here, and no plans to add either.

Two more. Registry and firmographic triggers, and hiring signals, are marked planned on our own site, and planned means not shipped today. And access is capped per market, so if the market is full you wait, which is a real constraint on an agency that needs to start a client on Monday.

An article on this subject is expected to end in benchmarks: a cost per meeting, a healthy margin, a reply rate to aim at. There will not be any, and the reason belongs in the text rather than in a footnote.

We are not publishing an agency margin benchmark or a cost per booked meeting, because we did not measure a population of agencies and nobody else writing about this has either. A number in that shape is an invention with a decimal point on it, and once it is in print it gets quoted back for years.

We are also not putting our price next to anyone else's per record or per credit. What each product meters is not the same thing, and a table that pretends otherwise would be arithmetic dressed up as a service to the reader.

And no reply rate is promised. One figure of our own, dated and caveated: running signal-led outreach on our own pipeline, as of August 2026, contacting a store in the week its event happened moved reply rate from about 3.9% to about 8.5%, with the message unchanged. That is one sender, one market, one offer, over a window we chose to look at. It is not a controlled trial and it is not a forecast for your clients. What survives the caveats is only that the message did not change, which makes timing a lever you can test on your own sending in a fortnight.

Price the stack against one client rather than against your whole roster, because a roster average hides the only thing you need to know: whether any single engagement pays for the tools it requires. Five questions settle it, and all five are answerable from your own numbers in an afternoon.

1. Write down who your clients actually sell to. If two clients want different worlds, you are buying two stacks, not one, and no vendor will tell you that.
2. Ask how old the data may be before it stops being worth sending against. If you cannot answer, that is the gap, not the price per record.
3. Count sending identities per client, not sends per month. That is what the infrastructure line really scales with.
4. Name who decides, each Monday, which accounts get written to. If the answer is a saved filter, nobody is deciding.
5. Ask whether the leads have to leave the system as a file. If yes, half the market is out, including us.

If that points at us, there are 1,000 free leads on signup, which is enough to find out whether a client's segment exists in the pool and whether a signal fires in it often enough to matter, before money changes hands. What a seat costs afterwards is on the pricing page, and how we sit against other shapes of tool, including where they win, is on the comparison pages.

Every figure in this post is our own, taken from our own pages and our own sending. Nothing here is quoted from a third party.

• Keaz Signals ecommerce leads database and the buying signal catalogue, retrieved 4 September 2026. Source for the 4.1 million Shopify and WooCommerce stores across Europe and North America, the weekly re-scrape, the signal list, the segment builder, the included contacts and the 1,000 free leads on signup.
• Keaz Signals sales agent, retrieved 4 September 2026. Source for the per-store drafting and the Instantly sending path.
• Keaz Signals compared honestly and the pricing page, retrieved 4 September 2026. Source for the export, LinkedIn, platform and market limits, and for the per-market access cap.
• Our own outreach, as of August 2026: reply rate of about 3.9% rising to about 8.5% when contact fell in the week of the event, message unchanged. One sender, one market, one offer; not a controlled trial. A moving number, and the date is part of it.