Marketing Attribution: Connect Spend to Sales

Marketing attribution is supposed to help explain which activity contributed to a result. It cannot produce a complete record of every conversation, recommendation or forgotten visit that influenced a buyer. Treating it as a perfect answer creates more confidence than the evidence deserves.

The practical goal is a dependable measurement process: consistent definitions, reliable records, clear reporting rules and a way to investigate the gaps. A business can make better decisions with an honest, partial picture than with a polished dashboard that hides its assumptions.

This guide explains how to connect campaign activity with leads, opportunities and sales. It includes a measurement contract, a simple reporting structure and a worked example showing why attribution totals should not be added together as though every platform describes a different customer.

Distinguish attribution from the sales record

The sales record answers what the business sold. Attribution estimates how credit for that result should be assigned across observed marketing interactions.

Those are different jobs. Your order system or finance-approved CRM should establish the commercial outcome. Marketing tools can provide views of the interactions associated with it.

Google Analytics describes attribution as assigning credit to interactions along the path to an important action. The model influences that allocation. It does not replace the underlying business transaction.

Keep three questions separate

1. What happened in the business?

2. Which interactions did our systems observe?

3. How are we assigning credit among those interactions?

This separation helps explain why a campaign can receive credit for a sale without proving that the sale would not have occurred without it.

To investigate that last question, you need evidence about incrementality, such as a suitable controlled experiment, rather than an attribution report alone.

Agree on a measurement contract

A measurement contract is a short internal agreement about what the report means. It is a practical operating document, not a new analytics product.

Define the terms that marketing, sales and finance will use. What counts as a lead? When does it become qualified? Does reported revenue include tax, shipping, refunds or cancelled orders?

Field Decision to agree Customer New customer, any buyer or a specific account typeLeadValid enquiry with sufficient information for follow-up Qualified lead Enquiry meeting the agreed fit criteria Revenue Gross, net, booked or collected valueReporting date Enquiry date, order date, close date or payment date Acquisition cost Media-only cost or wider sales and marketing costs Attribution view Platform, analytics or CRM allocation rules

Add the reporting timezone and currency. Small differences can produce repeated reconciliation disputes, particularly when platforms close their reporting day at different times.

A consistent definition is more useful than an impressive metric everyone interprets differently.

Capture campaign context without creating a naming mess

Use a shared naming convention for campaign links. Decide how you will record source, medium, campaign and any additional detail the team genuinely uses.

Google's campaign URL documentation explains how campaign parameters communicate information to analytics. Consistency matters: different spellings and capitalisation can split what should be one group.

For example, decide whether paid social uses one agreed medium label rather than allowing several people to invent their own. Keep campaign names descriptive enough to interpret months later.

Do not place personal information in campaign parameters. URLs can be copied, logged and shared in ways that make them unsuitable for customer details.

Keep a central naming register

Record the campaign name, channel, destination, launch date and owner. This can begin as a simple shared table. Its value comes from consistent use, not from adding more fields.

Review redirects and forms to confirm that relevant context survives long enough to reach the reporting systems that need it.

Connect the enquiry to the later outcome

For a service business, a form submission is the beginning of measurement rather than the end. The record needs a stable identifier and a way to preserve the source information available at creation.

Decide which fields belong in the CRM: original source, latest known source, landing page, campaign, creation date and relevant click identifiers where supported. Keep collection appropriate to your consent and data policies.

Then define who updates qualification, opportunity status, value and outcome. The system cannot connect marketing with sales if later stages remain blank.

Avoid overwriting the original source every time someone returns through another channel. Preserve the history needed for the questions you intend to answer.

Ask customers how they heard about you when that information helps. Their answer can reveal word of mouth, events or other influence that tracking missed. Keep it as a separate self-reported field rather than forcing it to agree with the software.

Different evidence can coexist without one being automatically wrong.

Expect platform totals to overlap

Consider a hypothetical month with ten confirmed customers. A search platform claims eight conversions and a social platform claims six. Adding those totals gives fourteen attributed conversions, not fourteen confirmed customers.

The same customer may fall within both platforms' crediting rules. Reporting windows, interaction types, identity matching and modelling can differ too.

Use a table that preserves those distinctions:

Reporting view Hypothetical count What it represents CRM10 Confirmed new customers under the business definition Search platform 8 Conversions credited under that platform's rules Social platform 6 Conversions credited under that platform's rules Analytics 7 Recorded outcomes under the analytics setup

These figures are illustrative. The correct response is reconciliation, not selecting the largest number.

Start with event definitions, date ranges, timezones and duplicate records. Then examine attribution windows, consent-related gaps and cross-device behaviour. Document what remains unexplained.

Do not force agreement by changing definitions until the numbers match. That produces a tidy report at the expense of useful measurement.

Use acquisition metrics with clear cost boundaries

Return on ad spend compares attributed revenue with advertising spend. It does not automatically account for product costs, service delivery, agency fees, sales salaries or refunds.

Customer acquisition cost is equally sensitive to what you include. A media-only figure and a fully loaded figure answer different questions.

For a hypothetical campaign spending $5,000 and producing ten new customers, media cost per customer is $500. If the agreed acquisition-cost definition includes a further $2,000 of relevant campaign and sales costs, the result becomes $700 per customer.

Neither calculation is useful without its label. Keep the formula and cost boundary visible in the report.

Review customer value with similar care. Revenue booked today may be collected later, and projected repeat purchases may never arrive. Use observed data where available and identify assumptions when modelling future value.

Budget decisions become more credible when the business can see which numbers are measured and which are estimates.

Compare cohorts with equal time to mature

A lead collected yesterday has had less time to become a customer than one collected two months ago. Comparing their current conversion rates without that context can unfairly favour older campaigns.

Group leads by creation period and follow them through the sales process. Review the percentage reaching qualification, opportunity and sale after a comparable interval.

This is particularly useful for services with a long sales cycle. A campaign that attracts strong enquiries may look weak in the first report if closed revenue is the only measure.

Use early signals carefully. Qualification can provide timely feedback, but its relationship with eventual sales needs checking. If the definition drifts, the early signal becomes less useful.

Keep a view of unresolved opportunities. A sales process with many stale records can make campaign performance appear uncertain long after the marketing work is complete.

Add controlled evidence when the decision warrants it

Attribution describes credited interactions. Incrementality asks what additional result the activity caused compared with what would have happened otherwise.

For a major budget decision, a suitable holdout or controlled experiment may provide stronger evidence than changing attribution models. The design depends on channel, scale, audience and the practical ability to maintain a comparison.

Microsoft's research on controlled experimentation supports testing changes rather than relying solely on assumptions. Applying that approach to marketing requires enough data and a design that fits the business.

Not every account can support a precise incrementality study. Do not substitute a casual before-and-after comparison and label it causal proof.

Use the best available evidence, explain its limitations and match the strength of the claim to the strength of the method.

Build a monthly report that supports action

A useful report can be short. Show business results, acquisition activity, pipeline movement and the decisions required.

Include a reconciliation section for material differences between systems. Note tracking changes and any periods with incomplete data. An unexplained blank is better addressed directly than disguised inside an average.

End the review with specific actions: repair a missing field, investigate a weak lead source, change a landing page or continue collecting evidence before increasing spend.

Assign an owner to each action and check it at the next review. Reporting becomes valuable when it changes the quality of the next decision.

Frequently asked questions

Why do Google Analytics and ad platforms show different results?

They can use different event definitions, attribution rules, date handling, identity methods and reporting windows. Consent and tracking gaps can contribute too. Reconcile the setup and compare each view with the business's confirmed sales records.

Which attribution model is best?

No model provides a complete picture for every business. Choose a view that suits the question, keep its assumptions visible and compare it with CRM outcomes. Use suitable experiments when you need stronger evidence about the additional effect of spending.

Can a CRM improve marketing attribution?

Yes, if it preserves source information and records later outcomes consistently. A CRM can connect enquiries with qualification, opportunities and sales. It cannot recover every unobserved interaction or compensate for stages that sales staff do not update.

Build a report the business can trust

Useful marketing attribution depends on disciplined records more than a single perfect dashboard. Agree on definitions, preserve campaign context, record commercial outcomes and explain discrepancies. That creates a basis for decisions that remains useful when a platform's preferred view differs from the company's results.

Explore The Coral Room's marketing services for CRM strategy, reporting and marketing systems that connect spend with the sales pipeline.