How to tell if your small business marketing is working

Field notes · October 2, 2026 · 7 min read

You pay for marketing every month. A report comes back with impressions, clicks and a number called conversions. What it does not tell you is whether the business has customers it would not have had anyway. That is the only question worth the invoice, and most reports are not built to answer it.

Plenty of owners are stuck on the same thing. In the Federal Reserve's 2026 Report on Employer Firms, "reaching customers/growing sales" was the most common operational challenge, reported by 57% of small employer firms for the prior 12 months. The survey ran from September 3 to November 14, 2025 and drew 6,525 responses. It is a convenience sample, not a random one, and the Fed says so itself.

What follows is how we would check, in the order we would do it. It needs no new software, and it does not need us.

Start from the customer, not the click

A marketing report starts with the ad and counts forward: shown, clicked, form filled. Do it the other way around. Start with the people who paid you and walk back to how each one first found you.

Pull every new customer from the last 90 days out of whatever holds the truth about money: the invoicing or accounting software, the booking system, the job list. Not leads. People who paid. For each one, write down a single fact: how they first heard of you. Check the first email, the call log, the form entry, the note on the estimate. Where there is nothing, ask them. It is an ordinary question to ask a customer.

Then sort them into a short list: referral or word of mouth, returning customer, found you in search or on a map, clicked an ad, saw a sign or a truck, could not tell. Put revenue next to each line, not just a count. Keep "could not tell" as its own line. A large unknown pile is a finding in itself. It means the business keeps no record of where its customers come from, and no report can repair that after the fact.

Here is an illustration, not a real business. A remodeling company lists 40 new customers for the quarter. Nineteen came by referral. Nine were past customers back for another job. Six found the company in map results. Two came from ads. Four cannot be placed. The company pays for ads and for a monthly retainer. That table does not prove the ads failed. It does tell the owner where to look next, and it took an afternoon.

What the dashboard numbers actually count

None of this means the platform numbers are false. They count something narrower than they appear to.

Take calls from a Google Business Profile. Google's own help page defines that metric as "the number of times a customer clicked on the call button on your Business Profile." That is a tap on a button. It is not an answered call, a booked job or a new customer. An existing customer tapping to ask about tomorrow's appointment counts the same as a stranger.

Conversions in an ads account work the same way. Google's help page says: "You choose what you identify as valuable, like a purchase, sign-up, or phone call." Somebody chose what counts. If a short call or a visit to the contact page was set up as a conversion, the report will show a lot of conversions. Ask whoever runs your ads to list each conversion action and exactly what triggers it. It is a fair question and it has a short answer.

The record that settles the matter is yours, not the platform's. It is the list of who paid.

Attribution is not incrementality

Two words get used as if they meant the same thing. They do not, and the difference is where most of the confusion about marketing lives.

Attribution is bookkeeping. It says this sale can be traced to that ad, that call or that page. Google's documentation is plain about what it is: "Attribution models let you choose how much credit each ad interaction gets for your conversions." Credit is handed out by a rule. Under the long-standing "last click" rule, in Google's words, advertisers "give all the credit for a conversion to the last-clicked ad and corresponding keyword."

Incrementality is a bigger claim. It says this sale would not have happened without the ad. Attribution cannot tell you that, because the people most likely to click your ad are often the people who were already on their way to you.

The clearest published evidence comes from eBay. Economists Thomas Blake, Chris Nosko and Steven Tadelis ran what they describe as "a series of large scale field experiments done at eBay that were designed to measure the causal effectiveness of paid search ads." Their summary: "returns from paid search are a fraction of conventional non-experimental estimates." And at the extreme, "brand-keyword ads have no measurable short-term benefits." A brand keyword is a search for the company's own name. In plain terms, people who typed "eBay" were going to arrive at eBay. The ad they clicked on the way got the credit.

A second study points the same way. Researchers at Northwestern's Kellogg School and at Facebook compared the usual after-the-fact methods against true experiments, "using data from 15 US advertising experiments at Facebook comprising 500 million user-experiment observations and 1.6 billion ad impressions." Their conclusion was that "commonly used observational approaches based on the data usually available in the industry often fail to accurately measure the true effect of advertising."

Two cautions, so these are not read for more than they say. A small business is not eBay, and neither paper says advertising does not work. The eBay authors found that new and infrequent users were positively influenced by ads. The lesson that carries over is narrower: a traced sale is not a caused sale. If it took controlled experiments at that scale to tell the two apart, a monthly report cannot do it by counting clicks.

Three checks for this week

• The customer list. Described above: every new paying customer from the last 90 days, how each first found you, and the revenue beside each source. This is attribution done from your own records, and it ends at money.
• The name test. Ask for the search terms your ads were shown for, and add up the spend on searches that contain your own business name. Those people were already looking for you. In our opinion that spend is the first to question, not automatically the first to cut, since a competitor can advertise on your name too.
• A pause test, if you can afford one. Pick one channel or one service area. Write down what you expect to happen. Pause it for two to four weeks, then compare paying customers from your own records against the weeks before and the same weeks last year. It is rough. Seasons, weather and small numbers all muddy it. It is also the only one of the three that speaks to incrementality.

Then keep one standing habit. Ask every new customer how they found you, and write the answer in the same place the invoice lives.

What we think

This part is opinion, and you can decide whether you agree. Most small businesses that cannot tell whether marketing is working do not have a marketing problem they can see. They have a gap in the records between the lead and the invoice. Marketing lives in one system and money lives in another. Nothing joins them, so the question gets answered with activity: posts published, clicks, impressions.

An owner should be able to get the answer in customers and revenue, from the owner's own records. If a vendor cannot trace its work to people who paid, that is worth saying out loud, and it applies to us as much as to anyone.

It is also how we work. Amalgament builds systems under three names, and Scout is the one for new customers. Scout finds demand a business is not capturing and builds the path required to turn it into customers. We follow the customer, not the click. Where a client's systems allow, each result is traced from the system that created it to the revenue it produced, and we prefer read-only access to the systems of record so outcomes are verified and not estimated.

We hold ourselves to the same distinction as above. Tracing a customer to something we built is attribution. Saying that customer would not have come otherwise is incrementality, and we do not claim it casually.

None of the checks here require hiring anyone. Start with the list of who paid.