Why Your Reports Lie About Where Leads Come From

Every month, the owner of a service business opens their CRM or ad dashboard and sees something like this: "Instagram — 40 leads, Google — 25, word of mouth — 15." Budget decisions get made on those numbers: add here, cut there. The picture looks solid — it's data, not guesswork.

The problem is that this picture isn't a fact. It's a reconstruction. And if you allocate budget by it, you can spend money in the wrong places for years without understanding why.

This doesn't mean analytics is useless. It means you have to use data while understanding how it's built — otherwise it misleads you more confidently than intuition does. Let's break down why reports distort the source of your leads, what that costs, and what to do about it if you don't have an analytics department.

Why "the source of a lead" is almost always an oversimplification

The word "source" in a report carries a hidden assumption: that a lead has a single source at all. In practice, a service business customer travels a path of several touchpoints — saw an ad, forgot, saw it again, asked friends, read reviews, visited the site, left, came back a week later and submitted a request. The "source" in the report isn't the reason for the inquiry — it's one of the touchpoints the system managed to record. Usually the last one. Here are six reasons it diverges from reality.

1. Credit goes to the last touch, not to what actually worked

Ad systems attribute a lead to a touchpoint near the end of the path. In its "last click" model, Google Ads gives all the conversion credit to the last-clicked ad. Google has now moved by default to a data-driven model that distributes credit across several touchpoints — but even that only sees what happens inside Google's own ecosystem. Meta counts a conversion in its own window, by its own rules.

Example: a person saw your Instagram ad for three weeks, never clicked, then remembered the clinic's name, typed it into Google, clicked through and booked. In the report, the source is Google. But the demand was created by Instagram; Google merely caught an intent that was already there. The report shows the performer, not the author.

2. Each platform counts itself — and they don't reconcile

Meta counts a lead in its attribution window, Google in its own. These systems don't reconcile data with each other. So the same booking can land in both the Meta report and the Google report — both platforms will honestly claim it.

Add up "leads by platform" and the sum comes out larger than the real number of leads in your CRM. This isn't a platform error — it's how independent counting works. But an owner who adds up numbers from two dashboards is counting leads that never existed.

3. Some leads never reach the report

Here's the opposite problem — undercounting. An ad dashboard sees a lead only if the signal reached it. And the signal gets lost:

• iOS and App Tracking Transparency. Apple requires apps to ask the user for permission to track their activity across other apps and websites. When the user declines, Meta loses some of the data about what they did after seeing the ad.
• Ad blockers prevent the browser pixel from firing.
• The cookie consent banner. If a visitor declines cookies, the browser-side tracker may not start at all.

Platforms fill these gaps with modeling — statistically reconstructing the missing conversions. Meta does this through Aggregated Event Measurement. Which means part of the numbers in the dashboard aren't measurement — they're estimates. That's a normal engineering technique, but "estimate" and "fact" are different things, and in a report they look identical.

4. Every platform uses a different ruler

The attribution window is the period during which a touch still counts. Meta's default window is 7 days for a click and 1 day for a view. Google has its own window and its own credit-distribution model. That means "25 leads from Google" and "40 leads from Instagram" were counted by different rules. Comparing them directly is like comparing two distances, one measured in meters and the other in paces.

5. The strongest channels hide

Word of mouth, outdoor advertising, a person who was recommended to you — these sources are either invisible to the systems entirely or disguised as something else. A customer who heard about you from a friend and typed your name into Google or 2GIS will show up in the report as "Google" or "direct." Branded search — when people search for you by name specifically — looks like a cheap and effective channel. But it doesn't create demand, it collects it. And what created the demand stayed invisible.

6. The human answer isn't reliable either

It seems like the question "How did you hear about us?" solves the problem — just ask directly. But a person names either the last thing they recall, or the most prominent thing, or simply "the internet." Word of mouth is systematically underreported in such answers, and conspicuous advertising is overreported. It's a useful signal, but not the truth: a customer's memory is just as approximate an instrument as an ad pixel.

What this costs in money

You could say: "so it's distorted — the numbers point roughly the right way anyway." The problem is that the distortion isn't random. It's systematically skewed in one direction. And that skew costs money.

In an article on marketing measurement in 2026, Angelina Eng, IAB's Vice President of Measurement (the IAB, Interactive Advertising Bureau, is an industry association that develops digital advertising standards), puts it this way: budget flows toward channels that are easy to measure, while channels that are hard to measure are underfunded — regardless of how much they actually influence the result.

The logic is simple and destructive. A channel that collects demand that's already there — branded search, retargeting (ads shown to people who already know you) — always looks cheap and effective in the report: leads seem to fall onto it by themselves. A channel that creates demand — display advertising, Instagram, building awareness — looks expensive: it works for the future, while the report looks at the past and at the last touch. The owner sees the numbers, cuts the "expensive" channel, and shifts budget to the "cheap" one.

An illustration (the numbers are hypothetical, for clarity). A clinic spends ₸600,000 a month: ₸400,000 on Instagram and ₸200,000 on branded search in Google. In the report, Google shows 60 leads, Instagram shows 30. The conclusion seems obvious: Google is twice as effective, move the budget there. A month and a half to two months later, branded search starts to dry up — fewer people are typing the clinic's name into search, because fewer people saw the Instagram ads. Total lead flow drops, even though no one touched the "effective" channel. This isn't a hypothetical paradox — it's a direct consequence of the report not distinguishing "created demand" from "collected demand."

What the industry recommends

The same IAB article also offers a framework for the solution. The core idea: stop treating the ad dashboard as the source of truth. In the article's own words, platform data should be treated as "directional, not definitive."

Instead of one report you trust for everything, it proposes a connected measurement system in which each tool has its own role:

• Marketing Mix Modeling (MMM) — a statistical model that estimates each channel's contribution to sales from historical data, without relying on tracking individual users. It answers the question "how to split the budget across channels overall."
• Incrementality tests — a test of causality. The question isn't "how many leads did a channel claim for itself" but "how many leads disappear if the channel is switched off." Only this shows whether the advertising actually works.
• Data-driven attribution — distributes credit across the sequence of touchpoints within what the system can see. Useful, but it only sees its own stretch of the path.
• Dashboard data — an operational signal for day-to-day optimization. Directional, not the truth.

Plus context that no dashboard contains: a promotions calendar, seasonality, competitors' moves, the overall market situation. A spike in leads might be the advertising's doing, or it might be a promotion you launched — without context you can't tell them apart.

The framework is sound. But it's written for large brands that have analytics and finance departments. Next — what of it is actually applicable to a service business.

What of this applies to a service business in Kazakhstan

Honestly: a full MMM is too early for a dental clinic or an auto service with a budget of ₸1–3 million a month. A model like that needs a lot of data and history; on small volumes it produces noise, not an answer. The good news is that most of the value can be obtained by simpler means. The principle is the same: better to be approximately right than precisely wrong.

Make your own CRM the source of truth, not the ad dashboards. A dashboard belongs to the platform and counts in the platform's favor. The CRM belongs to you. All leads are consolidated in one place, and the source is recorded automatically at the moment the lead is created — through UTM tags on ad links that the form saves into the record. This doesn't eliminate the distortions, but it gives you one honest reference point instead of three that contradict each other.

Count by deals and money, not by "conversions." A dashboard counts a "conversion" — a submitted form. What matters to the business isn't the form, but the customer who showed up and paid. A channel can bring in plenty of cheap leads that never reach a booking. So the number has to be carried through the whole funnel: lead → call → appointment → deal → money. A channel is judged by the money at the end, not by the leads at the start.

A simple "turn it off and watch" test instead of expensive experiments. Large brands run complex incrementality experiments. A small business has a simplified version available: switch a channel off for 2–4 weeks and watch not its dashboard, but the total flow of leads and deals in the CRM. If you switched off Instagram and total flow dropped 3–4 weeks later — the channel was working, even if the dashboard credited it with little. If nothing changed — that's something to think about. It's a crude instrument, but it measures causality itself, not the last touch.

Compare branches or cities against each other. If a business has several locations, you can switch a channel on in one city and not in another, then compare the result. The difference between them is the channel's contribution, cleaned of seasonality and the general background.

Ask the customer — but properly. Keep the question "How did you hear about us?", but treat it as one signal among several, not a verdict. It's better if the administrator clarifies it in conversation and cross-checks it against facts: was there a call, did the customer search for the name. Several weak signals together are more reliable than one that's trusted blindly.

Keep context next to the numbers. Maintain a simple calendar: when promotions ran, what season it was, what notable moves a competitor made. Then a dip or a spike in the report gets an explanation, instead of turning into a reason for panic or euphoria.

Where to start this week

You don't need to rebuild everything at once. A few steps you can realistically do in a week are enough:

1. Write down the numbers you actually make budget decisions by, and next to each one note where it comes from: an ad dashboard, the CRM, or memory.
2. Check that all your ad links carry UTM tags, and that the form and the CRM save the source at the moment of the lead.
3. Add up the leads from all your ad dashboards and compare the sum with the real number of leads in the CRM for the same period. The gap between them is the scale of the double counting and the losses.
4. Stop comparing channels by numbers from different dashboards as if they were one ruler — they use different counting rules.
5. Set up one report where the source of truth is the CRM, and the metric is the path to money, not to a submitted form.
6. Schedule one "turn it off and watch" test for the coming month, for a channel whose effectiveness you're not sure about.
7. Start keeping a context calendar — promotions, season, notable competitor moves.

There won't be a perfect picture — and that's fine

The main takeaway is unpleasant but freeing: an exact answer to "where did this lead come from" doesn't exist. The customer's path is too complex and the instruments too limited. Chasing perfect attribution means spending effort on the unattainable.

Something else is attainable — a measurement system you can trust enough to make decisions by. Not "the absolute truth," but an honest direction: which channel creates demand, which one collects ready demand, where the money is and where there's only the appearance of it. That's enough to stop cutting what works and feeding what's useless.

Alteora builds exactly this kind of measurement system for service businesses — consolidating data into a single point, carrying metrics through to money, and helping read them without illusions. With no promises of "guaranteed growth": we work from the premise that managed growth starts with honest numbers, not with a good-looking report.