The Independent Guide to Ad Tracking & Attribution
Understanding what your advertising data tells you, what it doesn’t, and how to make more informed decisions.
How to use this guide
This guide is designed for business owners, marketers, and anyone who needs to interpret advertising results without treating a dashboard as unquestionable truth. Read it in order, or use the checklist as a standalone decision aid. Terms and settings change: verify details against current platform documentation.
The guide does not endorse a particular tracking vendor. A better understanding of existing reports may be more useful than buying another tool.
1. What ad tracking can observe
Ad tracking connects identifiable interactions—such as an ad click, landing-page visit or form submission—with events recorded later. What can be connected depends on tags, permissions, identifiers, browser behavior, integrations and reporting rules.
Observation is not explanation. A recorded purchase after an ad click does not, by itself, show that the advertisement caused the purchase. A customer may already have intended to buy or may have encountered other marketing.
For every metric, ask: Was this directly recorded, inferred by a model, or assigned to a channel using an attribution rule? These are different kinds of information.
2. Clicks, sessions, leads, sales and revenue
An ad click counts an interaction with an advertisement. A session represents a visit under an analytics system’s session rules. A lead is a defined event, which might be a form submission rather than a qualified prospect. A sale should ideally reconcile to an actual order or payment record.
These totals need not match. A person may click twice and generate one visit, abandon the page before analytics loads, or complete a purchase days later. Systems may also filter invalid interactions differently.
Illustrative example, not WTE data: An ad platform shows 120 clicks, site analytics records 103 sessions, and the order database shows 7 completed purchases. This is not automatically a tracking failure. Start by comparing definitions and time ranges; then investigate missing events and duplication.
Revenue requires a clearly stated definition. Gross sales or gross order value reflect sales before relevant adjustments. Net revenue generally reflects applicable reductions such as returns, refunds and discounts, subject to accounting conventions. Profit is a separate measure that also considers relevant expenses. Decide which measure supports your decision.
3. Why two dashboards disagree
Common causes include different timezones, conversion windows, attribution models, event definitions, delayed reporting, identity matching, consent settings, browser restrictions and modeled conversions. Differences may be expected, but they should still be explainable enough for the decision at hand.
Example: A buyer clicks an ad on Monday, returns directly on Wednesday and purchases. One report may assign credit to Monday’s ad interaction; another may classify the Wednesday visit differently. Both may be applying their documented rules to the same order.
Before deciding that either report is “wrong,” align the conversion event, date range, timezone and reporting basis. If necessary, reconcile recorded orders against a payment or commerce system.
Further reading
4. Attribution is not proof of causation
Attribution assigns credit according to a method. It can help organize observed journeys, but it is not a controlled experiment. Tracking gaps from cross-device activity, consent decisions, offline exposure and missing identifiers can remain even when implementation is technically sound.
Incrementality asks a different question: what would have happened without the advertising? Carefully designed experiments, when feasible, may provide stronger evidence of causal lift than attribution reports alone. They also have assumptions and practical limits.
Decision prompt: If this dashboard were perfectly implemented, which important business questions would it still leave unanswered?
Further reading
5. A practical troubleshooting workflow
Step 1 — Define the question. Are you investigating missing purchases, conflicting attribution, or revenue that fails to reconcile? Do not troubleshoot everything at once.
Step 2 — Verify event definitions. Confirm whether each system counts form submissions, completed payments, qualified leads or another event. Note the counting method.
Step 3 — Inspect implementation. Test whether events fire as intended, whether they fire twice, and whether consent behavior changes collection.
Step 4 — Align comparisons. Use comparable dates, timezones, attribution windows and processing delays. Document exceptions.
Step 5 — Reconcile transactions. Where possible, compare purchase events with actual orders or payments, accounting for refunds and canceled orders.
Step 6 — Record uncertainty. Separate confirmed defects from expected differences and unresolved gaps. Prioritize fixes that could change a real decision.
Keep a short dated change log: settings adjusted, tests performed, results observed, and what remains unknown.
Further reading
6. Questions before buying software
New tracking software can sometimes improve event capture, reporting convenience or reconciliation. It cannot remove every privacy constraint, identity gap or attribution assumption. First check whether existing tools are misconfigured or whether the question itself requires a different method.
Ask prospective vendors: What data do you observe directly? What is estimated? Which integrations and permissions are required? How do you handle duplicate events, refunds, privacy choices and retention? What will implementation and maintenance cost? Can claims be tested against known transactions?
When not to buy: If you have not defined the reporting problem, cannot verify current event definitions, or have too little reliable conversion data to judge a new system, an audit of the current setup may be the better first step.
7. Printable evaluation worksheet
- ☐ We can state the business question in one sentence.
- ☐ We know what each conversion event represents.
- ☐ We have a system of record for completed transactions.
- ☐ We have documented reporting timezones and windows.
- ☐ We can distinguish observed events from modeled or attributed events.
- ☐ We have checked duplicate and missing event risks.
- ☐ We have considered consent, privacy and retention requirements.
- ☐ We can test a proposed tool with known transactions.
- ☐ We understand the full cost and ongoing maintenance.
- ☐ We have considered whether buying nothing is the best option.
Decision note: What uncertainty remains, and what evidence would change our decision?
8. Sources, limitations and further reading
These official documentation links are starting points, not an exhaustive technical or legal review. Specific platform behavior and regulations vary by product, region and date. This guide is not legal advice or a guarantee of tracking accuracy.
References and further reading
Download the free PDF edition of this guide (PDF).
No email address or registration required. The guide is freely available to read online or download as a PDF.
Explore more educational resources at WeTrackEverything.com
Editorial disclosure: WeTrackEverything.com may earn affiliate commissions through some links on its website, at no additional cost to you. This guide contains no affiliate links and recommends no tracking product for purchase.
This resource is educational and does not guarantee tracking accuracy.