Customer acquisition strategy

What does qualified mean, and who decides?

Qualified is a definition, not an adjective. Somebody has to write it down, and that somebody is whoever takes the first call. Marketing holds the pen, sales holds the veto, and the thing they produce is a page, not a feeling.

A definition that works has four parts, and each one has to be checkable by a person in under a minute.

Fit. Who they are, written as rules a machine could apply: a company size band, a country list, a role category, an industry exclusion list. If fit needs a judgment call, it is not written tightly enough yet.

Need. The problem they actually have, captured as their answer to one question on the form, in their words. Not inferred from the page they landed on. A free-text field with a real question in it ("what are you trying to fix?") will outperform three dropdowns, and it is the field that makes a lead readable six weeks later when somebody asks why it closed.

Timing. A window, stated. "This quarter" and "sometime next year" are different leads and should not share a number.

Access. Whether they can get the person who signs into a room. Not whether they are that person. Early-stage definitions that insist on the buyer's own job title throw away most of the good leads in companies over about 200 people, where the person with the problem is never the person with the signature.

Here is the test for whether your definition is real. Pull the last 20 leads. Give the list to the founder, the person who takes first calls, and whoever runs marketing, separately, with no discussion. Each marks every lead qualified or not. If the three of them agree on fewer than 17 of the 20, the definition is a mood, and every number built on it will be argued about. Rewrite the rule that caused each disagreement, then run the same test on 20 more.

Then put it where the work happens: one field on the record in the CRM, set by the person who takes the call, with a short list of reasons when the answer is no. Not a tab in a spreadsheet, and not a tag that only marketing can see. The reason codes are the part people skip and the part that pays: three months of "no budget this year" and "wrong country" tells you which campaigns to switch off without any attribution work at all.

The mistake that costs six weeks is doing this second. A team builds the reporting first, because reporting is the satisfying part, and the dashboard goes up. Then the first review happens, somebody says those aren't really qualified, and the whole exercise restarts, except now the numbers have a reputation. Write the definition in week one. The report is worth building the week after, and worth nothing before.

From my work: at a Series A B2B SaaS company, the month we lost was the month spent getting four people to agree on what counted as a good lead. I would now spend the first week only on that, in a document everyone can see, before touching a single campaign.

Which channels deserve a test, and which deserve a budget?

Two channels, run well, is the right answer for almost every company under about 50 people. The limit is not money. It is how many things one person can read carefully every week, and that number is two.

Before you choose, understand what a test costs, because a test that cannot produce a readable signal is not a test. It is a donation with a spreadsheet attached.

The platforms publish their own thresholds. Meta's ad sets leave the learning phase after roughly 50 optimization events within seven days of the last significant edit (Meta Business Help Center, About the learning phase, accessed Oct 2026). Google recommends judging a Target CPA campaign over the last 30 days with at least 30 conversions in them (Google Ads Help, About Target CPA bidding, accessed Oct 2026).

Turn that into a budget floor. If you optimize toward an event that costs $300, Meta needs about 50 of them in a week to settle, so the floor for that one ad set is around $15,000 a week. Most companies at this stage do not have it, which is why so many paid tests produce a cost per lead that swings by half every week and an argument about whether the channel works.

There are two honest ways out, and one dishonest one.

The first honest way is to optimize toward an event that happens five to ten times more often: a registration, a calculator completion, a specific page that only a serious reader reaches. The budget floor drops by the same factor. You are then testing whether the creative and the audience can produce interest, which is a real question, and you have to check the quality of what arrives by hand, every week, against the definition above.

The second honest way is to run the channel as a learning exercise on a fixed budget, with one written question, a date when you will answer it, and no expectation of a stable cost per acquisition. Written down in advance, that is a legitimate experiment. Undeclared, it becomes the campaign nobody will turn off.

The dishonest way is to split a small budget across six ad sets so that each one looks like a proper test and none of them leaves learning. This is the most common way paid acquisition fails at an early company, and it fails quietly, over a quarter.

For choosing which channels to shortlist at all, the most useful piece of thinking is still the bullseye framework in Weinberg and Mares's Traction. I have written up what holds up in it and what has dated. The short version: write down every channel so you cannot default to the one you already know, then screen the list on three questions. Can you reach these people there. Can you afford a test that produces a readable answer. Could this channel ever carry the volume you will need in two years. The third question is the one that gets skipped, and it is the one that kills channels after you have already built a team around them.

A channel graduates from test to budget when it has produced qualified leads at or under your ceiling for two periods in a row, each one long enough to contain a full sales cycle. Not one good month. Two periods, because the first one is usually the best audience you will ever get from that channel.

What should a lead cost?

Nobody can tell you, and the published averages are the proof.

Across thousands of Google Ads and Microsoft Ads accounts, the average cost per lead was $66.69, the average cost per click $5.42, and the average conversion rate 8.18%, with cost per lead running from $26.84 in arts and entertainment to $131.63 in legal services (LocaliQ, Search Advertising Benchmarks, updated June 1, 2026). A spread of five times between industries, inside one dataset, from one set of platforms. Those numbers are useful for calibrating what the auction feels like. They are useless as a target.

Your number comes from the other end of the business, and the chain is short enough to do on paper.

1. First-year revenue per customer.

2. Gross margin on it, so you are working in money the business keeps.

3. The payback period you can afford, which is a function of your bank balance and not your ambition.

4. CAC ceiling = annual gross profit per customer × (payback months ÷ 12).

5. Cost per qualified lead ceiling = CAC ceiling × the rate at which qualified leads become customers.

Worked through: a $12,000 contract at 80% gross margin leaves $9,600 a year. If you can carry a 12-month payback, your CAC ceiling is $9,600. If one in five qualified leads becomes a customer, your ceiling for a qualified lead is $1,920. And if one in three form fills clears the definition, your ceiling for a raw form fill is $640, which is roughly ten times the published average above.

That is the point of doing it. The benchmark says $66.69 and your ceiling says $640, so a $200 cost per lead is either comfortable or ruinous depending on a close rate that no benchmark knows. The method, the five mistakes it prevents, and the week it takes to wire up are in cost per qualified lead.

Two things to hold onto. The ceiling is a ceiling, not a target: paying it exactly means growth costs you everything it earns for a year. And the only comparison worth making is your own, channel by channel, on one definition, over time.

When is acquisition the wrong problem to solve?

Often. Three tests, in this order, and each one takes an afternoon.

The retention test. Draw the curve: of the people who arrived in a month, what share are still active one, two and three months later. If that curve is still falling at month three rather than flattening out, acquisition multiplies a leak. The arithmetic is unkind. A hundred new users a month against a curve that settles near nothing produces a base that does not grow however much you spend, and doubling the spend doubles the loss on the same month. How to read a retention curve is the first thing to do with a week, ahead of any channel work.

The activation test. Take the step from arrival to first real use of the product, and compare it with the same step elsewhere in your own funnel: a different traffic source, a different landing page, last quarter. If the paid traffic activates at a fraction of the rate the rest does, you are paying to buy people who never see the product work, and no change of channel fixes that. Activation is where that work lives.

The capacity test. Count how many qualified leads one person can genuinely work in a week, multiply by the number of people, and compare it with what you already produce. If leads are sitting for a day before anybody calls, the constraint is response time, not volume. The evidence on this is old and still the strongest number in the stage: in a study of over 15,000 leads and 100,000 call attempts across six companies, the odds of qualifying a lead fell by 21 times between a five-minute response and a thirty-minute one (Lead Response Management Study, James Oldroyd, MIT Sloan, with InsideSales.com, 2007). Spending more to generate leads that wait is the most expensive mistake on this page. The fix is in lead routing without a person in the middle.

There is a fourth test, about patience rather than problems. If your sales cycle is six months and you have been running a channel for eight weeks, you do not have a verdict on that channel. You have a cost per lead and no outcome yet. Decide in advance what you will judge at eight weeks, which is lead quality against the definition, and what you will not judge until month seven, which is whether any of it turned into revenue.

What good looks like

One written definition of qualified that three people mark the same way on a blind list of 20. Two channels, each with a cost per qualified lead under a ceiling you computed from margin rather than borrowed. One report that marketing and sales open in the same meeting without a preamble about whose number is right. And a new qualified lead reaching a human being in minutes.