Your dashboard says 140 leads this month. How many of them paid you?

If you can't answer that in a minute, you don't have an attribution problem. You have a "what are we even measuring" problem, and attribution software won't fix it.

The pattern repeats. The marketing report counts one thing, the business runs on another, and nobody connects the two until somebody asks why the phone isn't ringing even though the report looks great.

This is the practical version for a small business. No model names you need to memorize. Just how to trace a dollar of revenue back to where the lead came from, and what usually breaks along the way.

What attribution actually is

Attribution answers one question: which marketing activity caused this customer to show up?

That's it. Everything else, the models, the windows, the dashboards, is machinery for answering that question with varying levels of confidence.

For a small business, the useful version has three links in a chain:

  1. Source: where the person came from (a Google ad, a map listing, a referral, a mailer).
  2. Lead: the call, form or message they sent.
  3. Revenue: the job they booked and the invoice they paid.

Most reporting stops at link two. Ad platforms report clicks and leads. Almost none of them see link three, because the invoice lives in your accounting software or your scheduling system, not in the ad account. The gap between link two and link three is where most marketing money gets misjudged.

Why lead counts lie

A lead is whatever the system is set up to count. That's the problem.

Depending on how tracking was built, a "lead" might include:

  • Spam form fills and bots
  • Wrong numbers and robocalls that lasted long enough to count as a call
  • Existing customers calling about a job in progress
  • Job seekers and vendors
  • Price shoppers who wanted a $200 service when you sell $8,000 ones
  • Someone who clicked "call" and hung up

So a campaign can look great on cost per lead and be a disaster on cost per job. I'll go deeper on that gap in cost per lead vs cost per job, but the headline is simple: if you pay an agency or a platform based on leads, they'll deliver leads. Whether those leads pay you is a separate question that someone on your side has to own.

Where tracking breaks in real businesses

The theory is tidy. Reality is messy, and in my experience the mess is usually mundane. Here are the breaks I run into again and again.

The tag isn't on every page

A site can show almost nothing in analytics for weeks. The easy assumption is that traffic is low. Often most pages simply aren't carrying the tracking tag. The site is fine. The measurement is broken, and every decision made off the numbers is a guess.

Check this first. Open your key pages, use a tag-checking tool or your analytics real-time view, and confirm the tracking fires on the landing pages, the thank-you pages and any form or booking widget.

The software can't connect a source to a job

On client work I ran into this more than once: the scheduling or practice-management software had no field for "where did this customer come from," or couldn't attach a source to an appointment. The ad account showed leads. The scheduling system showed appointments. Nothing joined them.

When that's the situation, you work with what you have. A manual field the front desk fills in on every new booking. A shared sheet that cross-references call tracking with who actually booked. It's clumsy and it works. A clumsy link between source and revenue beats a beautiful dashboard that stops at the lead.

Somebody else owns the tracking

I've inherited accounts where the call tracking numbers belonged to the previous agency, the website admin login was unknown and the analytics property sat under a former employee's email. You can't measure what you can't access.

Whatever you do, make sure your business owns its own accounts: ad accounts, analytics, tag manager, call tracking, domain and website admin. Agencies should be added as users, not hold the keys. If you're ever fired or you fire them, you keep the data.

Billing and account problems reset the data

A declined card that pauses ads for a week doesn't just cost you a week. It disrupts the platform's learning, and the comparison before and after gets muddy. When you read a trend, look for what changed in the account, not just in the market.

Blended numbers hide the answer

If you spend on three channels and report one blended cost per acquisition, the channel that's quietly losing money hides behind the one that's winning. I ran into a version of this when a third-party lead source was bundled into total spend and couldn't be separated, which inflated the blended number and made every channel look worse than it was. Break spend and results out by channel wherever you can.

A simple attribution setup that works

You don't need an expensive platform. You need a short list of things done consistently.

1. Give every channel a way to be identified

  • Phone calls: use call tracking numbers per channel (paid search, map listing, website, print). Number swapping on the website, or a distinct number per campaign.
  • Forms: add a hidden field that records the source. Most tools can capture campaign parameters from the link the visitor arrived on.
  • Links: tag your own campaign links so they show up in analytics with a clear name. Google's documentation on attribution explains how analytics assigns credit across channels.
  • Everything else: ask. "How did you hear about us?" on the phone and on the booking form. People's answers are noisy but they catch what tracking misses, like word of mouth and the neighbor's recommendation.

2. Record the source where the job lives

This is the step most businesses skip. When a lead becomes a customer, the source needs to follow them into your CRM or job system. If the job is created without a source field, the chain breaks right there.

Make source a required field on new jobs. If that's too much friction, make it required on the first booking only.

3. Close the loop on ad platforms

For paid search specifically, platforms can learn from what happens after the click, if you tell them. Google Ads supports importing offline conversions, which lets you send back which leads became booked or paid jobs so the system optimizes toward them instead of toward form fills. It takes some setup, and I'll cover it in a separate post on Google Ads offline conversions. The principle is what matters: teach the platform what a good lead is, using your revenue, not its own click counts.

4. Match the window to your sales cycle

If a roof replacement takes six weeks from first call to signed contract, a seven-day reporting window will undercount every campaign. If your jobs book the same day, a 90-day window will blur everything together. Pick a window that reflects how your customers actually buy.

5. Review monthly, with the money in the room

Once a month, put three columns side by side for each channel: spend, leads and paid revenue. Not "cost per lead" alone. Revenue per dollar spent, or at least cost per paid job.

You'll see things the platform reports never show. A channel with expensive leads that close at a high rate. A cheap channel that produces nothing but tire-kickers.

Attribution models in plain English

You'll hear about models. Here's all you need.

  • Last click: all credit goes to the final thing the customer touched before converting. Simple, but it overcredits brand searches and underrates everything that built awareness.
  • First click: all credit to the first touch. Good for seeing what starts relationships, bad for seeing what closes them.
  • Multi-touch and data-driven: credit is spread across touchpoints by a rule or an algorithm. More fair in theory. Needs a lot of data to be reliable, and a small business with a few hundred leads a month usually doesn't have it.

For a small business, I'd say: pick one model, don't switch it every month and don't treat any model as the truth. They're lenses. A model that tells you search and referrals both matter isn't broken. It's telling you customers use more than one door.

What matters more than the model is whether the data feeding it is clean, and whether it ends in revenue.

The part nobody wants to own: what happens after the lead

Attribution can tell you a lead came from a good source and still miss the real problem. The lead might have called and got voicemail. It might have sat in an inbox for two days. It might have reached a front desk that couldn't explain the next step.

A common version: the ads do their job and the conversion dies at the front desk. The calls come in. The first conversation loses them.

Marketing reports would never show that. They'd show "calls received" and call it a success. This is why I treat lead response and follow-up as part of the attribution conversation, not a separate one. I wrote about it in my lead response rule.

If marketing and operations don't share the same set of numbers, each side will quietly blame the other.

Who should own this

In a small business, usually nobody does, and that's the real reason it's broken. The agency thinks the owner has it. The owner thinks the agency has it. The office manager enters data and doesn't know why it matters.

Pick a person. It can be the owner, an operations manager or a trusted bookkeeper. Their job is small:

  • Confirm tracking works after any website, form or phone change.
  • Make sure the source field is filled in on new jobs.
  • Pull the monthly three-column view: spend, leads, paid revenue.
  • Ask the agency or platform one question every month: "Which of these leads paid us?"

Five questions you can ask your marketing vendor today, whoever it is:

  1. Which accounts do we own, and who has admin access?
  2. How are leads defined, and what's filtered out (spam, existing customers, wrong numbers)?
  3. Can you show cost per booked job, not just per lead?
  4. What is our conversion window, and why?
  5. When did tracking last get tested end to end?

A good partner welcomes these. If you get defensiveness, that's information.

When it's not worth building

Attribution has a cost. If you take twelve leads a month, a spreadsheet and a conversation with each new customer will tell you more than any tool. Get the basics right first: tracked phone numbers, a source field, a monthly review. Add sophistication when volume justifies it.

And if your numbers say a channel doesn't work, believe them. The point of measuring is to be willing to spend less somewhere.

Where to start this week

  1. Check that tracking fires on your key pages and forms.
  2. Confirm your business owns every account.
  3. Add a required source field to new jobs.
  4. Pull last month's leads for one channel and trace each one to a paid job, or to nothing. Doing it by hand for one channel will teach you more than a platform will.

If you want a second set of eyes on how your tracking is built, start with Di-Hy's free audit.

FAQ

What is marketing attribution in simple terms?

It's tracing a customer back to the marketing that brought them in. For a small business the useful version links a lead source to a paid job, not just to a click or a form fill.

Which attribution model should a small business use?

Pick one and keep it consistent. Last click is simple and fine as a starting point. What matters more is whether your data is clean and ends in revenue, because no model fixes bad inputs.

How do I know if my marketing is working?

Compare spend, leads and paid revenue by channel every month. If you can't connect leads to paid jobs, add call tracking, a source field on every job and a manual check of one channel's leads.

Why do my marketing reports say great results but my phone is quiet?

Reports usually count leads, and leads can include spam, wrong numbers and low-value inquiries. Check what is counted, then trace a sample of leads to paid jobs.


Sources: Google Analytics Help: attribution and attribution modeling, Google Ads Help: about offline conversion imports. Platform features change, so check the vendor's current documentation before you build.