Your Attribution Problem Might Actually Be a Tracking Problem

Attribution gets blamed for a lot.

Meta says it generated 100 purchases. Google Analytics says 70. Shopify shows 85. A third-party attribution platform lands somewhere else entirely.

In B2B, the ad platform reports 40 leads, HubSpot shows 25, and the sales team says only five were worth following up with.

The immediate assumption is usually that the attribution model is wrong.

Sometimes it is. But a lot of the time, the real problem starts earlier: the tracking itself is incomplete, duplicated, mislabeled, or disconnected from the outcome the business actually cares about.

No attribution model can fix bad inputs.

Tracking and attribution are not the same thing

Tracking records what happened.

Attribution determines which marketing interaction gets credit for it.

If a purchase event fires twice, attribution doesn’t fix that. If Google Ads is optimizing toward account creations instead of paying customers, attribution doesn’t fix that either.

If every form submission is treated as a qualified lead, changing from last-click to data-driven attribution just redistributes credit for the wrong outcome.

This is why I generally separate the two questions:

1. Are we accurately tracking the actions that matter?

2. Once we trust those actions, how should credit be assigned?

Too many companies jump directly to the second question.

The platforms may be optimizing for the wrong thing

This is one of the most common issues I find in ad accounts.

A company believes its campaigns are optimizing for purchases or qualified leads. When we audit the setup, we find a much broader collection of actions marked as primary conversions:

• Purchases

• Add-to-carts

• Checkout starts

• Account creations

• Form submissions

• Button clicks

• Phone-number clicks

• Page views

• Engaged sessions

Those actions can all be useful to track. That doesn’t mean they should all be treated equally—or used for bidding.

If a campaign generated one purchase, five checkout starts, and ten add-to-carts, the account might report 16 conversions. The campaign looks much stronger than it really is, and the bidding algorithm is being told that all 16 outcomes are valuable.

They aren’t.

The same issue happens in B2B. A demo request from a decision-maker at a qualified company is not equivalent to a student downloading a report, a vendor filling out the contact form, or a job applicant reaching the wrong page.

If all of those actions are imported as the same conversion, the platform will optimize toward whichever type it can generate most cheaply. That is often not the type the sales team wants.

Lead generation needs lifecycle tracking

For most B2B companies, the initial lead should be the beginning of the measurement process—not the end.

The real journey might look something like:

New lead → Activated → Onboarding → Paid

Or:

Lead → MQL → SQL → Opportunity → Customer

If reporting stops at the first form submission, the marketing team can reduce its cost per lead while producing worse business results.

Imagine two campaigns:

• Campaign A generates 100 leads at $50 each, but only two become qualified opportunities.

• Campaign B generates 40 leads at $100 each, but ten become qualified opportunities.

If the ad platform only sees the initial lead, Campaign A looks like the obvious winner.

Once CRM outcomes are included, Campaign B is substantially more valuable.

This is why importing lifecycle stages from HubSpot, Salesforce, or another CRM matters. It allows us to evaluate cost per MQL, cost per opportunity, pipeline generated, and eventually revenue—not just the cheapest available form fill.

It also gives the bidding algorithms better information. Google and LinkedIn cannot optimize toward qualified leads if nobody sends qualified-lead data back to them.

E-commerce tracking can be just as misleading

E-commerce companies often assume their tracking is cleaner because a purchase is easier to define than a qualified lead.

It still goes wrong.

Common issues include:

• Purchase events firing more than once

• Missing purchase events

• Incorrect revenue values

• Taxes or shipping being included inconsistently

• Draft orders being mixed with online purchases

• Subscription renewals being counted as new customer acquisition

• Returning customers being mixed with first-time customers

• Add-to-cart or checkout events being set as primary conversions

• Browser and server-side events failing to deduplicate

• Product IDs not matching between the store, feed, and ad platforms

A campaign can show a strong return on ad spend while the business sees little improvement in total revenue or new-customer acquisition.

That doesn’t automatically mean the platform is lying. It might mean the account is measuring a broader or different outcome than the business thinks it is.

Disagreement between platforms is normal

Even with a clean tracking setup, Meta, Google Ads, GA4, Shopify, and third-party attribution tools will not match perfectly.

They use different:

• Attribution windows

• Identity methods

• Click and view definitions

• Conversion timestamps

• Cross-device modeling

• Consent assumptions

• Deduplication processes

• Rules for assigning direct traffic

Meta may credit a conversion after an ad view. GA4 may attribute that same order to direct or organic traffic. Google Ads may report the conversion on the date of the ad click, while Shopify records it on the purchase date.

Those differences are expected.

The goal should not be to force every platform to display the same number. The goal is to understand what each system measures and decide which source should answer each question.

For example:

• Shopify may be the source of truth for total orders and revenue.

• A CRM may be the source of truth for qualified leads, pipeline, and customers.

• GA4 may be useful for evaluating website behavior and broader acquisition paths.

• Ad platforms may be useful for bidding and platform-specific optimization.

• A third-party attribution tool may help compare channels using a consistent methodology.

There does not need to be one source of truth for every question. There needs to be a clearly defined source of truth for each decision.

How to tell when tracking—not attribution—is the problem

There are a few obvious warning signs:

• Reported conversions increase without a comparable change in revenue or qualified leads.

• A campaign suddenly improves immediately after a new conversion action is added.

• The platform reports far more leads than the CRM.

• Sales says lead quality is falling while cost per lead is improving.

• Purchase revenue in the ad platform exceeds store revenue.

• Conversion totals change dramatically when you switch reporting tools.

• One platform appears to take credit for nearly every sale.

• Campaigns are optimizing toward actions the team cannot clearly define.

• Nobody knows which conversions are primary and which are secondary.

• Nobody has recently tested the events themselves.

The bigger warning sign is when the reporting looks good but the business outcome does not.

That disconnect should trigger a tracking audit before a major strategy change.

The order of operations matters

When measurement is questionable, I use the following order:

1. Define the actual business outcomes

What are we trying to generate?

For e-commerce, that may be first-time customers, profitable orders, subscription starts, or contribution margin.

For B2B, it may be activated users, MQLs, opportunities, pipeline, or closed revenue.

“Conversions” is not a specific enough answer.

2. Inventory every tracked action

Review every event and conversion across:

• Google Ads

• GA4

• Google Tag Manager

• Meta

• LinkedIn

• The CRM

• The e-commerce platform

• Any third-party attribution tools

Document what triggers each action, where it is used, and whether it influences bidding.

3. Separate primary and secondary actions

Primary conversions should represent outcomes you are genuinely willing to have the platforms optimize toward.

Secondary conversions can still help with analysis and audience creation without being treated as equal to revenue or qualified leads.

A button click may be useful. A two-minute session may be useful. An account creation may be useful.

That does not automatically make any of them a primary business conversion.

4. Validate the implementation

Test the full journey.

Confirm that events fire when they should, do not fire when they shouldn’t, and pass the correct values and identifiers. Check thank-you pages, confirmation popups, embedded forms, phone clicks, checkout flows, browser events, server events, and CRM imports.

Never assume that an event is correct because its name looks correct in a report.

5. Connect marketing activity to downstream outcomes

For B2B, push lifecycle stages back into the advertising platforms.

For e-commerce, separate new and returning customers when possible and verify that transaction values align with the store.

The closer optimization gets to revenue, the more useful it becomes.

6. Then evaluate attribution

Once the underlying events are accurate and meaningful, attribution becomes a much more productive conversation.

Now we can compare attribution windows, new-customer return, blended performance, incrementality, and channel overlap without wondering whether the original conversion happened at all.

Better measurement usually changes the strategy

Cleaning up tracking doesn’t just make the dashboard more accurate. It changes how budget gets allocated.

A campaign that looked efficient may turn out to generate low-quality leads. A campaign with a high cost per lead may be the strongest source of pipeline. A retargeting campaign may be claiming revenue that would have happened anyway. A non-brand campaign may look inefficient in last-click reporting while consistently introducing new customers.

These aren’t small reporting differences. They affect bidding, creative, landing pages, channel mix, and how much the company can afford to spend.

That is why tracking should not be treated as a technical box to check before launching campaigns. It is part of the media strategy.

Start with the inputs

Attribution will never be perfect. Customer journeys are messy, privacy restrictions continue to grow, and no platform has a complete view of every interaction.

But imperfect attribution built on clean tracking is still useful.

Sophisticated attribution built on bad tracking is not.

Before changing attribution models, buying another analytics platform, or making a major budget decision, make sure the events feeding those systems are accurate, meaningful, and connected to the results the business actually cares about.

Your attribution problem might be real.

But I’d check the tracking first.