How Do I Know Which Marketing Is Actually Working?

How to know which marketing is working comes down to one unbroken chain: source, lead, qualified opportunity, sale, revenue, margin. Marketing is working when you can trace a customer back to the channel that produced them, know what that customer cost to win, and compare channels using the same revenue definition.

The harder question is what to do when your systems disagree, which they almost always do.

First, consider what that uncertainty is worth. A business doing $3M to $5M and putting 5 to 10 percent into acquisition is spending $150,000 to $500,000 a year. If half of that produces nothing, up to $250,000 walks out the door annually. You cannot tell us which half. That is the whole problem, and it is not a small one: Forrester’s 2026 global marketing survey found that 49 percent of B2C marketing decision-makers say analytics findings do not translate into action. And those are companies with analysts on staff.

Here is the version most owners recognize. Meta reports 42 conversions. Google reports 31. GA4 shows 55 form submissions. Your CRM contains 48 leads. Accounting shows 17 new customers and $41,000 collected. Which number should leadership use to decide where the next dollar goes?

In fact, every one of those systems is reporting accurately. They are just not reporting the same thing.

What Does It Mean to Know Whether Marketing Is Actually Working?

Marketing is working when it produces customers at a cost the business can afford, at a margin worth having, and you can prove which activity produced them. Three separate tests, and most businesses only pass the first one by accident.

Test one is traceability. Can you name the channel that produced a specific customer, not a specific lead? After all, leads are easy to count and expensive to misread.

Test two is cost. Do you know what it cost to win that customer, including the spend that did not convert? A channel is not cheap because its clicks are cheap.

Finally, test three is comparability. Can you put two channels side by side using the same definition of a lead, a sale and revenue? If Facebook counts a message reply as a conversion and your CRM counts a signed proposal, the comparison is meaningless before the math starts.

Notice what is absent from all three. Impressions, reach, engagement rate, follower growth and click-through rate are diagnostic. They explain why a channel performed the way it did. However, none of them answer whether the business made money.

What most reports show	What the business actually needs to know
Impressions and reach	How many customers arrived
Clicks and sessions	What each customer cost to acquire
Conversions per platform	Which system counted that conversion, and once or twice
Cost per lead	Cost per closed customer
Total leads	Revenue and gross margin by source

A useful way to think about it: activity metrics tell you what the marketing did. Meanwhile, business metrics tell you what the marketing produced. You need both, but only one of them belongs in a budget conversation.

Why Don’t Google, Meta, GA4 and Your CRM Show the Same Results?

They disagree because each one measures a different event, over a different time window, using a different rule for who gets credit. No single platform is lying. Instead, each is answering its own question honestly.

Ad platforms are the clearest example. Google Ads and Meta each report conversions they can connect to their own ads, inside their own attribution windows. In addition, both include view-through and click-through behavior by default. When one person sees a Meta ad on Monday, searches your brand on Wednesday and converts, both platforms can legitimately claim that same customer. Add the two reports together and you have counted one sale twice.

GA4 measures something else again. It tracks sessions and events on your site, applies its own attribution model, and cannot see anything that happens after someone picks up the phone. Google’s own attribution documentation is explicit that different models assign credit differently to the same set of touchpoints.

Your CRM measures a third thing: people your team actually recorded. However, its accuracy depends on whether a source field was required at entry, and whether anyone filled it in honestly at 4:45 on a Friday.

Finally, accounting measures a fourth thing. Money collected, and not money promised.

System	What it actually counts	Where it goes wrong
Google Ads	Conversions it can attribute to a Google click or impression	Claims sales other channels also influenced
Meta Ads	Conversions inside its click and view windows	Same double-count risk, different window
GA4	On-site sessions and events, by its own model	Blind to phone calls, offline sales and CRM outcomes
CRM	Records your team created and tagged	Only as good as the source field discipline
Accounting	Revenue actually received	No idea where any of it came from

None of them has to be wrong for all of them to disagree.

There is a practical consequence, therefore. Until you decide which system owns which fact, every budget meeting turns into an argument about whose report is real.

What Is Marketing Attribution?

Marketing attribution is knowing what actually brought a customer in. More precisely, it is the method you use to assign credit for a sale across the touchpoints that preceded it.

In practice, most buying paths have several. Someone finds you through a Google search, leaves, sees a Meta ad two weeks later, asks a neighbor, then calls the number on your Google Business Profile. Five touches, one customer, one revenue number to divide.

Attribution models differ in how they divide it.

Model	How credit is assigned	Best used for	Where it misleads
First touch	100 percent to the first interaction	Understanding what creates awareness	Overvalues top-of-funnel, ignores what closed
Last touch	100 percent to the final interaction	Simple, fast, easy to explain	Overvalues brand search and retargeting
Linear multi-touch	Split evenly across all touches	Seeing the full path	Treats a stray click like a sales call
Position based	Weighted to first and last	Balancing discovery and conversion	Weights are a guess unless tested
Data driven	Modeled from your own conversion patterns	Larger data volumes	Needs volume most businesses this size do not have

HubSpot’s attribution reporting documentation shows the same set of models applied inside a CRM rather than an ad platform, which is worth understanding, because a CRM can follow the path further than any ad platform can.

Here is what matters more than model selection. Pick one model, apply it consistently, and then know what it is biased toward. A last-touch model will always make branded search look like your best channel, because branded search is where people go once something else already convinced them. That is not a flaw to eliminate. It is a bias to remember when you read the report.

Attribution does not have to be perfect to be useful. It has to be consistent enough that a change in the numbers means a change in the business rather than a change in the settings.

How Do You Track a Lead From First Click to Closed Revenue?

You track it by making six stages connect to each other, in order, with a shared identifier: source, lead, qualified opportunity, sale, revenue, margin. The chain breaks at handoffs, so the handoffs are what you build.

1. Source. Every inbound path gets tagged before anyone clicks it. UTM parameters on paid and email links, call tracking numbers on offline placements, distinct landing pages where tagging is not possible. If a channel cannot be tagged, it cannot be compared later.

2. Lead. The source has to survive the form submission. That means the UTM values get captured into hidden fields and written to the CRM record, not just to the analytics session. Sessions expire. CRM records do not.

3. Qualified opportunity. Somebody decides whether the lead was real, using a definition written down in advance. Without that step, a channel producing 300 junk inquiries outranks a channel producing 40 serious buyers on every report you own.

4. Sale. The close gets recorded against the same record, with a date and an amount. If your sales happen by phone, this is where most tracking chains die, because nobody connects the closed job back to the lead that started it.

5. Revenue. Recognized revenue, from the system that actually owns that number. Usually accounting, sometimes an ERP or job-costing system, rarely the CRM.

6. Margin. Gross margin by source, because two channels can produce identical revenue at very different profitability.

Additionally, send the closed sale back to the ad platforms through offline conversion imports or server-side tracking. That is what lets Google and Meta optimize toward customers instead of form fills. Most businesses skip it, then wonder why their ad platforms keep buying the wrong leads.

In practice, two things break this chain more than anything else. Internal traffic that was never excluded from analytics, and conversion tracking that was never configured correctly in the first place. Someone connects an AI to that data and asks what is working. It comes back with traffic up, impressions up, here is your content strategy. The report is a beautiful story about traffic that will never buy anything, and there is no way to tell from inside the report.

Which Marketing KPIs Should a Business Owner Actually Track?

A metric becomes a KPI when it changes a decision. If a number goes up or down and nothing about how you run the business changes, it is a metric. Interesting, sometimes. Not a KPI.

For most businesses in the $2M to $5M range, seven numbers carry the weight:

1. Cost per qualified opportunity by channel. Not cost per lead. Qualification is where channel quality shows up, so this is the number that should move budget.
2. Close rate by source. Two sources with the same lead volume and a 20-point close-rate gap are not the same source.
3. Customer acquisition cost by channel. Total channel spend divided by customers that channel produced. Which costs belong in that number is where most calculations go sideways.
4. Revenue by source. Recognized revenue, tied to the source field.
5. Gross margin by source. The number that decides whether a channel is worth scaling.
6. Lead response time. Marketing spend evaporates in the gap between an inquiry arriving and someone answering it.
7. Payback period. How many months of margin it takes to recover acquisition cost. This is the number that determines whether you can afford to spend more.

Also notice what is not on the list. There is no follower count, no engagement rate, no impression share. Those belong in a channel review, not on the page leadership uses to allocate money.

One caution about benchmarks. An industry average close rate tells you very little, because your offer, price point, geography and sales process are not the industry’s. Your own trend line over four quarters is worth more than any benchmark you will find online.

How Do You Compare Marketing Channels Using CAC and Revenue?

Compare them by carrying both channels all the way to gross profit, then subtracting what each channel cost. Cost per lead comparisons routinely reverse once you do.

Two channels, one quarter, same business:

	Channel A	Channel B
Spend	$14,000	$12,000
Leads	200	400
Cost per lead	$70	$30
Close rate	25%	5%
Customers	50	20
Customer acquisition cost	$280	$600
Average revenue per customer	$2,400	$2,400
Gross margin	45%	45%
Gross profit produced	$54,000	$21,600
Contribution after channel spend	$40,000	$9,600

Channel B wins on every metric a lead report shows. In fact, it delivers twice the leads at less than half the cost. Yet Channel A produces roughly four times the contribution, and its customers cost less than half as much to acquire.

Now change one input. If Channel B’s customers were worth $6,000 each instead of $2,400, its gross profit would be $54,000 and the ranking would flip. That is precisely why average revenue and margin have to sit in the same table as cost per lead. Otherwise you are comparing the front of one funnel to the front of another and calling it a decision.

A word on what this comparison assumes. It assumes both channels define a lead the same way, that no customer is counted in both columns, and that the revenue figure comes from the same system for both. Break any of those three and the table produces a confident wrong answer, which is more dangerous than no answer.

When the numbers are this clear, the decision usually is not budget. It is what the losing channel is actually for. Channel B might be doing awareness work that Channel A converts, in which case cutting it moves the problem rather than solving it. That is a question a holdout test answers and a spreadsheet does not.

What Should You Do When Your Attribution Data Is Unreliable?

You stop trying to reconstruct the past and start building a forward baseline. This is the part nobody wants to hear, so here it is plainly.

I can’t tell you where last year’s customers came from. Nobody can, if it was never tracked. What I can do is make sure you never have that problem again.

Retroactive attribution on untracked data is guesswork wearing a suit. Any vendor who offers to rebuild two years of source data from an untagged CRM is selling you a model, not a measurement.

What does work, in order:

1. Exclude internal traffic. Your team, your office IP, your contractors. Until this is done, every engagement number you have is inflated by an unknown amount.
2. Fix conversion tracking properly. Server-side tracking through Meta CAPI and Google Enhanced Conversions, so conversions still register when browser tracking fails. This alone changes what the ad platforms optimize toward.
3. Make the source field required. One field, required at lead creation, with a short controlled list. Not free text. Free text produces “google,” “Google,” “GOOGLE ADS” and “web” as four separate sources by quarter’s end.
4. Ask every customer, and record the answer. “How did you hear about us” is imperfect self-reported data, but it is real data, and it catches word of mouth that no platform can see.
5. Set a 60 to 90 day baseline. Run the tracking clean, do not change budgets, and let the chain fill up. In effect, you are buying a starting line.
6. Test with holdouts. Turn one channel off in one geography for four weeks. Watch total leads and total revenue, not that channel’s own report. Incrementality is the only thing that tells you whether a channel is producing customers or just claiming them.
7. Agree on the system of record. Decide which system owns spend, which owns customer status, which owns recognized revenue, then stop arguing about it in meetings.

What good looks like at the end of that work: a customer can be traced back to a source, you know what that customer cost to acquire, one system owns each number, and two channels can be compared using the same business definitions. Nothing exotic. Most businesses this size simply never had anyone whose job it was to build it.

If the tracking is clean and channels still cannot be separated, the problem may not be in marketing at all.

FAQ’s

Not sure your marketing and revenue data actually connect? The Insight Engine traces how marketing, sales and operational systems work together so you can see what is producing revenue and what is only producing activity. See how the Insight Engine works.