Why Is “Which Channel?” the Wrong Question on Its Own?

When a business owner asks “which advertising platform brings in the customers,” they are actually asking for something very human: a simple ranking. They want to know where to put the next unit of budget. The problem is that phrasing the question this way assumes there is one correct answer – even though marketing channels are not substitutes for one another, but work at different points in a process.

Think of a dental clinic offering premium implants. Google Search reaches the person who already knows exactly what they want and is comparing prices. Meta primarily reaches someone who may not even have been considering treatment yet, but suddenly stops at a before-and-after image. The email list brings back people who asked three months ago but then disappeared. Forcing these three into a single ranking would be like asking: what matters more in a car, the engine or the brakes?

The second problem is selection bias. Every channel brings people with different levels of intent, and those people become paying customers at different rates. Someone searching for a brand closes at a dramatically higher rate than a cold audience. If we look only at conversion volume, branded search will always win – even though the brand demand may have been created by the other channels. This is why we say that channels feed one another, and this contribution is completely invisible in most reports.

The third problem is the simplest and most uncomfortable one: most companies do not actually know which of the people coming from advertising eventually paid. One platform reports 42 conversions. Analytics reports 29. Accounting shows 11 new customers. All three numbers refer to the same month. Until these three numbers can be reconciled into one line, every answer to the channel question is, at best, an educated guess.

Why Should Platform Data Not Be Taken at Face Value?

Advertising-platform reports are not lies – they are interpretations. Every platform uses its own attribution model to decide what should receive credit for a purchase, and these models view the same journey differently. If you ignore this, the “best channel” will not be the most profitable one, but the one with the most aggressive attribution model.

Five Mechanisms That Inflate the Numbers

1. Last-click and similar default models. If the final click gets the credit, every channel that is present at the moment of decision – branded search, remarketing, email – becomes artificially overvalued. Upper-funnel work becomes invisible and is eventually cut.
2. View-through conversions. A purchase occurring within 1–7 days after an ad impression that was not clicked may still be counted. This mechanism regularly attributes purchases to advertising that might have happened anyway.
3. Modeled and estimated conversions. Because of missing consent data, systems increasingly use statistical models to fill in missing information. This is useful as a directional indicator, but it is not an audit-ready number.
4. Duplicate measurement. The same purchase may appear in the media platform, analytics and the webshop’s internal reporting. Adding all three reports together is a common and expensive mistake.
5. Different time zones, definitions and currencies. A significant part of the 8–15% differences seen at month-end is not fraud but semantics: the definition of a “lead” differs between media platforms, analytics and the CRM.

What Platform Data Actually Tells You

This does not mean platform data is useless. Quite the opposite: platform data is excellent for optimizing within the platform’s own system. If you are comparing two ad texts or examining which audience generates a better click-through rate, the platform’s own data measures exactly what it is designed to measure. The problem begins when we use it to make cross-channel decisions.

A useful mental rule worth remembering: platform data is operational; business data is strategic. An operational decision is: which creative should I switch off? A strategic decision is: what percentage of next quarter’s budget should go to search, what percentage to social media, and how much should remain for turning existing customers into repeat buyers? The platform’s own report will never be enough for the latter.

CRM Stages: This Is Where the Business Fact Is Created

If you can make only one structural change today, make it this one: measure channels not at the conversion level, but at the level of customer-journey stages. A conversion is an event on a website. A paying customer, however, is a state inside your business. Qualification steps need to exist between the two, and these must be recorded in the CRM.

A Working Stage Model

• Lead: provided contact details. We do not yet know anything meaningful about intent.
• MQL (qualified lead): matches the target profile – industry, size, geography and need.
• SQL (sales-ready): has a specific need, timing and decision-making authority, or at least two credible intent signals.
• Proposal: a formal proposal has been sent containing an amount and deadline.
• Won customer: signed contract or paid order.
• Repeat / upsell: second purchase or expansion. This is the level of real profitability.

Once this chain exists, the channel question suddenly becomes answerable, and most importantly: the answer often contradicts what the platforms claim. A typical pattern I see repeatedly is that social-media channels generate a high number of leads, but their SQL rate and closing rate are low. Search generates fewer leads, but a substantially higher proportion of them become contracts. If you look only at cost per lead, social media appears efficient. If you look at cost per won customer, the picture reverses.

The same applies on the revenue side. Two channels may generate the same number of customers, while the average contract value of one channel’s customers is twice that of the other. This difference comes from the nature of the channel: the level of intent is different, the decision-maker is different and the competitive situation is different. A conversion report will never reveal this because it has no knowledge of the contract value.

Practical implementation does not require a large system. A spreadsheet may be enough if it is maintained consistently: date of entry, source and campaign, stage, stage-change date and contract value. After three months of data, this table can give you a more accurate picture of channel value than any advertising-platform report. The important thing is not the tool, but that the fact of payment is connected to where the person came from.

The Three-Layer Measurement Model

Data-driven channel decisions are not about a single tool, but about connecting three layers. If any layer is missing, the decision will be incomplete. Let us look at what needs to be solved within each layer.

Layer 1 — Platform (Media)

This is where daily optimization happens: creatives, audiences, offers and bidding. At this layer, the platform’s own conversion metric is completely acceptable because the system is comparing performance inside its own environment. The task here is to maintain a consistent conversion definition: the same event should count as a conversion across every platform, and click-attribution windows should be documented. If the search campaign uses a 30-day click window while social uses 7 days, the two cannot be compared directly.

Layer 2 — Analytics (Behavior)

Analytics shows what happens on the website: where traffic comes from, where the process breaks down, and which events signal genuine intent. It is also worth defining micro-conversions here – using a price calculator, starting a quote request, watching 75% of a video – because these provide signals much earlier than the actual purchase. An important limitation of analytics is that cookies and consent requirements make the data incomplete: it undermeasures reality. Therefore, use it not as an absolute number, but as a tool for ratios and trends.

Layer 3 — CRM and Commercial Data (Business Reality)

This layer has the final word: how much revenue was generated, what the acquisition cost was, what the lifetime value is, and how long the investment takes to pay back. The technical minimum for connecting the three layers is: consistent use of UTM parameters, recording the source when the lead is created, and linking contractual data back to the source. After that, only one thing needs to happen regularly: reconcile the three layers monthly and document the differences.

The table reveals the key point: the two “high-reliability” layers are almost always missing from channel decisions. Yet these are exactly the layers that tell you how many of those 42 conversions became real business. Once you connect these two columns, arguments about the “better platform” largely disappear on their own, because you finally have a shared and difficult-to-dispute unit of measurement: profit.

Channel Scorecard: How to Rank Channels Properly

To compare channels, you need a common denominator. The following six metrics together are enough to evaluate a channel using mathematics rather than intuition. Important: all of them must be calculated for the same period and using the same definitions, otherwise the comparison is meaningless.

The scorecard becomes truly useful when you calculate it not for a single month, but over a rolling three-month period. In marketing, the short term can be misleading: after launch, a channel may perform more weakly for 4–6 weeks because the system is still learning and the audience does not yet know the brand. If you rank channels every month, you may end up always keeping the oldest channel running – not because it is the best, but because it is the most mature.