Acquisition channels that survive scrutiny

Every company we talk to has a channel it is supposed to be doing. Not doing well, not doing badly — supposed to be doing. Somebody in a meeting eighteen months ago said the words "we should really be on LinkedIn" or "we need to be doing SEO", nobody disagreed, a freelancer was hired for two days a month, and the thing has been quietly consuming money and calendar space ever since without anyone being able to say what it produced. Ask what it produced and you get a number for impressions.

Part of why this happens is that almost all channel advice is written by people who sell that channel. The agency's blog explains why the agency's channel is underrated. The conference talk about events is given by someone from the events company. This is not dishonesty; it is simply that people write about what they know and what they know is what they sell. But it means the literature has a systematic bias, and the bias is always in the same direction: every channel works, the only variable is execution.

Before arguing against that, it is worth conceding the strongest version of the opposite case, because channel scepticism has its own failure mode and it is worse. The sceptical operator who demands proof before spending anything ends up spending nothing, and companies that spend nothing on acquisition do not discover a cheaper route — they discover that growth was coming entirely from a founder's personal network and that the network is finite. There are also real companies whose entire trajectory came from one channel that looked absurd on paper, worked anyway, and could not have been justified in advance by any of the tests below. Channels are not fully knowable before you run them. The tests are not a substitute for trying things.

What the tests do is stop you from running four half-experiments forever. Our position is that a channel is only real for your company if you can answer three questions about it: does it have a plausible mechanism, is its payback period one you can survive, and what is its ceiling. Most channel conversations spend all their energy on the first, some on the second, and none at all on the third — which is unfortunate, because for a small or mid-sized company the ceiling is usually where the channel dies.

The mechanism has to be a sentence, not a category

A mechanism is the specific causal story by which money going in becomes a customer coming out. Not "we do content marketing." Something closer to: operations managers at logistics firms search for the phrase "customs broker onboarding checklist" when they are switching brokers, there are a few hundred such searches a month in our markets, we can rank for it because almost nothing good exists on that query, and someone reading that page is by definition mid-switch.

That sentence is testable. Every clause in it can be wrong, and you can find out which one. The category version — "content marketing" — cannot be wrong, which is exactly the problem. When a category underperforms you cannot tell whether the idea was flawed or the execution was, so the default response is more of it, for longer.

The mechanism question is also where most social channels quietly fail for B2B companies with considered purchases. It is entirely possible to build an audience on a platform and never find the sentence connecting that audience to a buying decision. Sometimes the sentence exists and is fine — a founder posting about the problem their software solves, reaching the people who have that problem, who then remember the name when the budget appears. Sometimes it does not exist and the audience is other people in your industry, admiring your posts and buying nothing, forever. Both look identical on a follower chart.

Payback is a cash-flow question, and it is where small companies get hurt

The second question is how long the money is out of the building. A channel with a twenty-two-month payback might be perfectly rational for a company that can fund twenty-two months of it and lethal for one that cannot. This is not a debate about whether the channel works. It works and you still die.

The gap between paid and organic acquisition is mostly a gap in this dimension rather than in efficiency. Paid search bills you this month and, if the mechanism is sound, produces enquiries this month; the cost per customer might be high but the loop is short enough that you can see it turn several times before committing serious money. Editorial content bills you in salary and months and produces nothing for two quarters, then produces compounding returns that no paid channel matches. We have gone through that trade-off properly in the comparison of SEO and paid search on a small budget, and the short version is that the right answer is set almost entirely by how much runway sits behind the decision, not by which channel is better.

The thing to be honest about is that payback periods are usually quoted from the wrong starting point. The clock does not start when the first customer arrives. It starts when you begin paying — including the two months of hiring, the agency's ramp, the three drafts nobody published. A twelve-month payback measured from first result is often an eighteen-month payback measured from first invoice, and eighteen months is a different decision.

The ceiling is the question nobody asks

Suppose the mechanism is sound and the payback is survivable. There is a third question, and it is the one that ends more channels than the other two combined: how big can this get before it stops mattering?

Every channel has a ceiling set by something outside your control. Search has a ceiling set by how many people type the query. Outbound has one set by how many companies exist that fit your profile. Referral has one set by how many customers you already have. Events have one set by how many people are in the room. You can execute perfectly and still hit it.

This matters far more at small scale than the literature admits, because the ceiling is absolute while your target is relative. A channel that tops out at a modest annual figure is a rounding error for a company with hundreds of employees and can be most of the growth plan for a company of twelve. That cuts both ways, and the direction people miss is the second one: a channel can be too small to be worth the organisational cost of running it, even when it is profitable. If a channel needs a person's ongoing attention and cannot ever produce more than a few percent of revenue, the person is the expensive part.

The mechanical way to estimate a ceiling is unglamorous and takes an afternoon. Count the addressable population — searches per month, companies in the segment, attendees at the three relevant events, existing customers who could refer. Apply a conversion rate you have actually observed somewhere, not one from a benchmark report. Multiply by your average contract value. The number will be wrong, but it will be wrong by a factor of two or three, and that is usually enough, because the channels that fail this test fail it by a factor of ten. If you cannot count the addressable population at all, you do not yet have a mechanism — you have a category.

A campaign has an end date; a channel does not

The word "channel" is doing a lot of unearned work in most companies. A channel is a repeatable route to customers that keeps producing while you attend to something else. A campaign is a finite push with a start, an end, and a result. Both are legitimate. Confusing them is not, because they are budgeted, staffed and judged differently.

Most of what small companies call channels are campaigns. The conference you sponsor once a year is a campaign. The outbound sprint the two founders ran in January is a campaign. The launch post that brought a week of traffic is a campaign. None of these become channels by being repeated annually; they become channels when there is a mechanism running in the background that does not depend on a specific person deciding to push.

The test we find useful: if the person currently doing it went on leave for six weeks, would enquiries continue? For search, yes — the pages stay ranked. For a mature referral programme, mostly yes. For founder-led outbound, absolutely not, and that is worth knowing before you build a revenue plan on it. Founder outbound can be excellent. It is a campaign that recurs, and its ceiling is the founder's calendar.

The practical consequence is about where the reinvestment goes. Campaign spend buys results now and nothing later. Channel spend buys an asset — a ranked page, a documented sequence, a partner relationship — that keeps working. Companies that only run campaigns spend the same money every year and start from zero every January, which feels like a marketing problem and is actually a definitional one.

Honest ceilings, channel by channel

What follows is the uncomfortable version. These are structural limits, not execution problems, and they apply to companies in roughly the ten-to-two-hundred-employee range.

Search has the best ceiling structure of any organic channel and the worst latency. The ceiling is genuinely large for anything with commercial search volume, and unlike most channels it grows as you add pages. What kills it at small scale is not the ceiling but the gap between spending and earning, plus a second-order problem: for many specialist B2B products the searches that indicate genuine buying intent are surprisingly few, and everything else is people researching for a report. The right question is not "how much search volume is there" but "how much of it is someone about to spend money".

Outbound has a hard, countable ceiling and this is its great virtue: you can know it in advance. If there are 900 companies in your segment, that is the number, and you will work through it faster than you expect. Outbound also has the steepest quality gradient of any channel — the difference between a well-researched sequence and a mail merge is not percentage points, it is whether the channel exists at all. What it demands is someone whose job this is. Two hours a week from a founder does not run outbound; it burns list.

Referral has the highest conversion rate and the most awkward ceiling, because the ceiling is a function of the customer base you already have. It cannot be the channel that gets you to your first hundred customers, and it is often the best channel once you have them. It is also the channel most likely to be already working without anyone noticing, which is a separate problem we have written about in the referral channel nobody measures. Before building a referral programme, find out what the existing rate is. Companies are routinely surprised, in both directions.

Events have a ceiling you can count on your fingers, and it is the most honest number in this piece: the number of relevant people in the room, times the fraction you can actually speak to, times the fraction with an active need. For most industries that is a small number per event and there are perhaps four events worth attending. The reason events survive anyway is that the conversations are qualitatively different — an hour at a stand does something that no amount of email does — and the reason they are so often a waste is that the cost is fixed and large while the ceiling is fixed and small. Events reward companies with high contract values and punish everyone else.

Marketplaces and platforms — app directories, procurement platforms, integration listings, partner catalogues — deserve more attention than they get from small companies, because the demand is pre-qualified and the acquisition cost is close to zero once the listing exists. The ceiling is set entirely by someone else's traffic and someone else's ranking algorithm, which means it can move without warning and you have no recourse. Treat it as real revenue with a tenancy risk rather than as a channel you control.

Paid social is the channel where the mechanism question does the most damage. It is excellent at generating volume and indifferent to whether the volume means anything, which is a dangerous combination for a company whose sales cycle is long enough that the feedback loop never closes. It works when the purchase is impulsive enough that the click and the decision are close together, or when you have enough volume that you can actually measure downstream. A company doing a few dozen deals a year has neither.

Channel	Ceiling set by	Latency	Survives the owner leaving
Search	Commercial query volume in the niche	Two quarters or more	Yes
Outbound	Size of the addressable list	Weeks	Only with a dedicated owner
Referral	Size and satisfaction of the customer base	Immediate, then slow	Mostly
Events	People in the room, times events per year	One event cycle	No
Marketplaces	Someone else's traffic and ranking	Weeks to months	Yes, until the platform changes
Paid social	Budget, then relevance decay	Days	Yes, expensively

The pattern in the right-hand columns is the useful one. Channels that survive the owner leaving are the ones worth calling channels. The others are campaigns with good attendance.

What works at 200 people frequently cannot work at 12

A large amount of channel advice is technically accurate and structurally inapplicable, because it assumes fixed costs the larger company has already paid.

Content marketing at a company with an established domain means publishing and ranking. At a new company it means publishing and waiting, possibly for a year, because the same article on a domain with history and a domain without it are not the same asset. Anyone who tells you content is cheap is measuring the marginal article at a company that already did the expensive part. We have looked at whether that investment still pays back under current search conditions in the piece on content marketing's return, and the answer is conditional in ways that mostly turn on how long you can wait.

Outbound at a larger company means a sequence lands and a salesperson works the reply within the hour. At a twelve-person company the reply arrives while the person who sent it is doing something else, and response speed is most of what determines whether an outbound reply becomes a meeting. The channel does not fail because the copy is worse. It fails in the gap between the reply and the follow-up.

Paid acquisition at scale means a team reads the data weekly and reallocates. At small scale it means a monthly report nobody has the volume to interpret, because the difference between two campaigns at forty conversions a month is noise, and acting on noise is worse than not acting.

There is a whole class of channels — partner programmes, category advertising, sponsorships, anything that works by being recognised — where the mechanism depends on brand recognition the small company does not have and cannot buy at the same price. We have set out which ones those are and what the substitutes look like in the channels that only work at scale. The practical rule is to be suspicious of any tactic whose success story comes from a company more than ten times your size, and to ask specifically which fixed cost they had already paid.

Four channels run badly is worse than nothing

The most expensive mistake we see is not choosing the wrong channel. It is refusing to choose.

A small company running four channels is not diversified. It is running four channels at the level of attention four channels allow, which for a team without a dedicated marketing function means each one gets the worst version of itself: outbound with a list nobody researched, content published on the weeks somebody had time, paid campaigns nobody optimises, events attended without follow-up. Every one of those underperforms its potential by enough that the mechanism never becomes visible. And because none of them is clearly failing, none of them gets cut.

The compounding cost is informational. Four weak channels produce four weak signals, and a weak signal cannot tell you whether the channel is bad or your execution is. You end up a year later with the same four channels, the same ambiguity, and a marketing budget that has been spent teaching you nothing. One channel run properly produces a clear answer either way, and a clear negative answer is worth considerably more than four ambiguous ones — it is the only thing that lets you move on.

There is also a threshold effect that argues against spreading. Most channels have a minimum viable intensity below which they produce approximately zero rather than a proportionally smaller amount. Publishing one article a month for a year does not produce a twelfth of the result of publishing twelve; it produces nothing, because it never accumulates enough to rank. Attending an event without booking meetings beforehand does not produce a fraction of a staffed presence. Below the threshold, spend is not slow progress, it is waste, and half of what looks like diversification is four sub-threshold efforts.

How to decide within a quarter

The recommendation, then, and the reasoning that gets there.

Pick one channel — the one where you can write the mechanism sentence with the fewest guesses, and where the ceiling you estimated is at least two or three times the revenue you need from it. Two or three times, because your estimate is optimistic and because a channel you have to run at full saturation to hit plan is a channel with no room in it. Fund it above the threshold for one full sales cycle plus the time to fix the first round of mistakes. Do not add a second until the first is either near its ceiling or clearly answered.

Then decide in advance what you will look at, because the metric that tells you something is almost never the one the channel reports. Impressions, followers, sessions and open rates all move independently of whether anyone bought something. What tells you the mechanism is alive is the shape of the conversations arriving: whether the enquiries come from the segment you aimed at, whether they mention the specific problem you wrote about, whether they arrive already knowing what you do. Ten of those beats four hundred of anything else, and you can read them in week three rather than month nine.

Keep one thing running that you know already works — usually referral, usually informally — and do not disturb it while you experiment. The point of the discipline is not austerity. It is that a company which tests one thing properly per quarter knows four real things about its market by the end of the year, and a company running everything at once knows what it knew in January.