To understand which advertising brings customers, connect spending, visits, enquiries and sales. The advertising account reports its part of the journey; web analytics shows website behaviour; CRM records enquiry handling and deal outcomes.
End-to-end analytics does not mean knowing everything about every interaction. Some data is lost or limited by user choices, devices and platform rules. A useful system honestly shows what is known, supports decisions and does not replace profit with click counts.
First agree what the metrics mean
For a service company, an enquiry is an accepted request, a qualified lead meets agreed criteria and a sale is a defined deal outcome. In a shop, an order, a paid order and a completed sale after returns also differ. Write down definitions before building a report.
If the advertising account counts a phone-link click as a conversion while management expects a paid service, the figures differ in meaning even before any technical errors. Intermediate actions help diagnosis but must be named accurately. Do not combine form openings, submissions and purchases into one “result” metric.
Also define which costs are included: media spend, content production, fees and team work. Media spend can be examined separately when evaluating a channel, but total customer acquisition cost requires an agreed set of expenses.
Build a measurement map
| Stage | Data | Where to check |
|---|---|---|
| Advertising | Spend, impressions, clicks | Advertising platform |
| Visit | Source, page, behaviour | Web analytics |
| Enquiry | ID, service, language, source | Website and CRM |
| Handling | Qualification, proposal, rejection | CRM |
| Outcome | Payment, revenue, returns | CRM and accounting |
Each handover needs a clear checkpoint. For example, the website assigns an enquiry ID and CRM stores it with the deal. Orders use a stable number to distinguish redelivery of an event from a new purchase. Do not use a person’s name as a technical identifier.
Introduce consistent UTM rules
UTM parameters describe the source, traffic type, campaign and ad variant. Agree letter case, separators and a naming dictionary. For example, a hypothetical link might use utm_source=newsletter, utm_medium=email and utm_campaign=autumn_services. This contains no personal data and does not prescribe one structure for every platform.
Google Analytics documentation explains the role of manual tags and other source information. For integrated advertising systems, also check auto-tagging and related settings. Do not change tags randomly after launch: this splits reports into incomparable names.
Which errors are most common?
A link may lose parameters during a redirect, lead to another language version or use different spellings of the same source. Tags on internal links can distort the interpretation of navigation, so use other events for internal journeys. Check the final address after all redirects, not only the link in the marketer’s spreadsheet.
Do not put email addresses, telephone numbers, names or personal document numbers in UTM parameters or URLs. Addresses enter logs and third-party systems. Use appropriate identifiers and controlled transfer for matching, taking processing rules into account.
Configure events around actual outcomes
An enquiry submission event should mean that the request was actually accepted. A button click is useful separately, but does not prove the form succeeded. For purchases, record the correct amount, currency and order identifier. Test page refreshes and repeat server notifications.
If the same event is sent from both browser and server, consistent deduplication is needed according to the chosen platform’s rules. Server-side delivery can improve the reliability of individual operations, but does not remove consent and data-processing requirements. Nor does it automatically make every report complete.
Test the journey from an advertising link to CRM. Reconcile the tags, event, identifier and status. Separately test rejection of optional cookies and form errors. The report should make clear which data is available in these scenarios and which is missing.
Why different systems report different numbers
Platforms use different attribution windows and models, count different interactions and receive different amounts of data. One may attribute a result to an ad view, another to the last known visit. Time zones, click dates and purchase dates also change daily totals.
GA4 distinguishes source scopes: the user’s first source, the session source and event attribution. This is described in the guidance on traffic-source dimension scopes. Comparing them as if they were the same metric is incorrect.
Refused consent, blockers and cross-device journeys can reduce the observed volume. However, refusing cookies does not automatically turn all such traffic into Direct. Depending on the configuration, an event may not be observed at all or may be processed differently. Do not infer the channel solely from a missing source.
Reconcile identical periods, events, currencies and return rules. Then investigate remaining differences. The aim is explainable data for decisions, not artificially making every number match.
Measure economics, not just website conversion
Cost per lead, CPL, is agreed advertising expenditure divided by the number of enquiries. Customer acquisition cost, CAC, uses the number of new customers and the selected full set of expenses. ROAS compares revenue attributed to advertising with ad spend, but does not show profit on its own.
A hypothetical example: spending €1,000 generates 50 enquiries, of which 10 become customers. CPL is €20, and customer cost based only on this media spend is €100. If content preparation and management cost another €500, acquisition cost using the broader expense set is €150. This is an arithmetic illustration, not a market benchmark.
Add quality measures: how many enquiries fit the service, what proportion can be contacted and how many proposals are accepted. Cheap enquiries can become expensive customers. Detailed calculations appear in the article on advertising budgets.
Allow for the delay between enquiry and sale
If decisions take several weeks, do not compare all current-month spending only with sales from new enquiries received that month. Some deals have not yet had time to close. Examine groups of enquiries by acquisition date and their subsequent outcomes.
A separate operational report shows current payments, while a cohort report shows what happened to customers acquired during a particular period. Both are useful but answer different questions. Do not switch off a source merely because its recent enquiries are still at the proposal stage.
Retain adjustments for returns and cancellations. Otherwise, a channel with many unpaid or returned orders will look better than it is. In B2B, distinguishing signed contracts from revenue actually received is also useful.
Separate attribution from causal impact
Attribution distributes an outcome among observed interactions. It does not prove that the sale would not have happened without advertising. This is especially important for branded search and retargeting: some people may already have intended to buy.
Controlled experiments, comparable groups or regions are sometimes used to estimate incremental impact. Such tests require sufficient volume and careful design. A simple before-and-after comparison can reflect seasonality, price changes or competitors’ promotions.
For a small business, the first priority is often fixing lost enquiries, establishing definitions and ensuring CRM is updated consistently. A complex model built on incomplete data creates an appearance of precision without improving decisions.
Make the report a tool for action
Check spending, event operation, the proportion of accepted enquiries and unusual changes weekly. Over longer periods, evaluate qualification, sales and economics. For each significant deviation, record a hypothesis and action: check the page, change the offer, fix the source or allow deals time to close.
The report should disclose limitations: which channels are tagged, how many enquiries lack a source and how completely loss reasons are recorded. Assign owners for the website, analytics and CRM data. The technical delivery process is explained in the article on connecting a website to CRM.
An example comparison of two advertising sources
Consider hypothetical data for a comparable, completed period. Source A spent €600 and generated 40 enquiries, of which four became customers. Source B spent €900, generated 30 enquiries and nine customers. The respective costs per enquiry are €15 and €30, while customer costs based on media spend alone are €150 and €100.
The first source looks better by CPL, and the second by customer cost. But the decision is not complete. Check revenue, contribution after variable costs, returns and the sales workload. If deals differ in size or service duration, equal customer numbers do not imply equal value.
Then check data quality. Perhaps managers in one department diligently close deals while another leaves them open. Or one source is new and its customers have not yet paid. Before reallocating money, ensure the difference reflects the business rather than record-keeping discipline.
If the advantage remains after checking, increase spending in a controlled way. Results can change at higher budgets: audiences, competition and available demand are not unlimited. Save the baseline metrics and define in advance which changes require revisiting the decision.
How to investigate a sudden conversion drop
First compare independent signals: CRM enquiries, accepted forms, orders and analytics events. If CRM shows normal activity but events disappeared after a website update, measurement is likely the issue. If actual enquiries also declined, check page availability, the form and traffic changes.
Do not change ads, audiences, the website and the event definition simultaneously. Otherwise, identifying what caused recovery or further decline will be difficult. Record the change date and investigate the most likely cause sequentially.
Also allow for reporting delays. Do not make a final decision based on an incomplete current day while the system is still updating data. For immediate response, directly check the critical journey; for performance assessment, use completed, comparable intervals.
This sequence distinguishes a technical incident from a commercial problem. One requires restoring events or delivery; the other requires reviewing the offer, traffic quality or handling. The downward graph alone cannot provide that answer.
Frequently asked questions about end-to-end analytics
Can we identify the source of every sale with absolute accuracy?
Usually not. Offline interactions, multiple devices and observation limits exist. It is better to retain unknown values and explain the method than artificially assign every sale to an advertising channel.
Do we need an expensive service from day one?
Not necessarily. Start with correct events, tagging rules and disciplined CRM use. An additional system is justified when it is clear which work it automates and which decisions it improves.
Which report should be the primary one?
For financial outcomes, rely on verified sales and accounting data. Advertising accounts manage placements, while website analytics diagnoses the journey. Their roles complement each other.