How to Read UTM Campaign Reports in GA4

Use the UTM Campaign Link Builder to prepare consistent links before distribution. Read their performance in the destination website’s analytics, not in the builder. This guide uses an invented repair-kit planning campaign to work through GA4 reports. Every number below is a teaching example, not a real customer’s result or a forecast of what a tagged link will achieve.

Open the session report and inspect its dimensions

To ask which links brought website sessions, begin with Traffic acquisition. Google’s documented route is Reports, then Acquisition, then Traffic acquisition. Navigation can be customized, so a missing report does not prove tracking failed. Ask someone with editing access to locate or restore it. Confirm the property, website stream, date range, and active filters before comparing rows.

Use the table’s dimension picker to select Session source / medium. Look for the word Session; First user source / medium answers a different question. Add Session campaign as a secondary dimension using the plus control when it is available, or switch to it and filter the campaign. If the needed dimension is absent from a customized report, request a suitable report rather than substituting a similarly named dimension without checking its scope.

Filter the fictional campaign to repair_kit_planning_2026q4 and record that exact spelling. Check the complete table, not only its first visible rows. Document any alternate campaign spellings before combining them. A broad text search can accidentally include an earlier campaign sharing the same words, so confirm what the filter includes.

Record the reporting period, timezone, dimension names, campaign filter, selected key event, and export date beside the numbers. This makes the observation reproducible. Do not change the date window to favor a platform after seeing its result. If a placement launched later, compare matched exposure periods or explain the timing difference.

Read a small fictional dataset before ranking sources

Suppose the kit-planning campaign leads to https://kits.example/planner?kit=starter. Its website records a successful planner submission as generate_lead, which the property has marked as a key event. For this exercise, key events use once-per-event counting. We also know how many distinct sessions contained that selected event. The following three rows are invented and already filtered to one campaign, one reporting period, and the same website measurement setup.

The instagram / social row has 240 sessions, 156 engaged sessions, 18 generate_lead key events, and 15 sessions containing generate_lead. The youtube / video row has 160 sessions, 112 engaged sessions, 16 key events, and 12 sessions containing the event. The linkedin / social row has 100 sessions, 55 engaged sessions, 6 key events, and 5 sessions containing the event. These counts describe website activity, not all viewers of the campaign content.

Across those rows, the example totals are 500 sessions, 323 engaged sessions, 40 key events, and 32 sessions with the selected key event. Instagram contributes 240 / 500 = 48% of sessions, YouTube contributes 32%, and LinkedIn contributes 20%. That establishes traffic volume within this filtered example. It does not establish each platform’s share of the whole audience, unique people, total sales, or business profit.

Define the selected event explicitly. Here, generate_lead means a successful planner submission, not a button click or attempted completion. If the event fires when someone merely opens the form, the calculation describes a different behavior. Verify the implementation before discussing quality. Correct arithmetic cannot repair an event that counts the wrong action or fires twice for one submission.

Calculate engagement and key event rates correctly

Engagement rate divides engaged sessions by sessions. In the example, Instagram is 156 / 240 = 65%, YouTube is 112 / 160 = 70%, and LinkedIn is 55 / 100 = 55%. The combined rate is 323 / 500 = 64.6%. Do not average the three displayed percentages without weighting them: the rows have different session counts. Use the underlying totals when calculating the campaign-wide rate.

For generate_lead, the session key event rate divides sessions containing that event by all sessions. Instagram is 15 / 240 = 6.25%; YouTube is 12 / 160 = 7.5%; LinkedIn is 5 / 100 = 5%. The combined rate is 32 / 500 = 6.4%. Confirm that your report or analysis uses the selected event rather than an all-key-events total that also includes unrelated actions.

Dividing 40 key events by 500 sessions gives 0.08 events per session, or eight events per hundred sessions. It does not mean 8% of sessions completed the action. Some sessions contain multiple submissions, which is why the example has 40 event occurrences but only 32 sessions with the event. GA4 also supports once-per-session key event counting, so inspect the configured method before assuming that repeated occurrences explain a particular difference.

YouTube has the highest selected-event session rate, while Instagram brings more sessions and more lead-containing sessions. Neither observation predicts what would happen if you moved all effort to YouTube: placement, audience, message, and landing experience differ. Keep social interaction analysis separate. The Engagement Rate Calculator addresses social engagement inputs, not GA4 engaged sessions or website key event rates.

Use content labels to ask a narrower question

A source row can contain several links with different jobs. In this fictional dataset, Instagram’s profile_checklist_a contributes 150 sessions, 90 engaged sessions, 6 key events, and 5 sessions with generate_lead. Its profile_comparison_b contributes 90 sessions, 66 engaged sessions, 12 key events, and 10 sessions with generate_lead. The two content rows add back to Instagram’s totals. Check that reconciliation before drawing a conclusion from a detailed breakdown.

The selected-event session rates are 5 / 150 = approximately 3.33% for checklist A and 10 / 90 = approximately 11.11% for comparison B. That is an observed difference of about 7.78 percentage points. The smaller row produced more lead-containing sessions despite receiving less traffic. This is a reason to investigate the comparison message, not proof that its wording caused the difference or that the same rate will persist.

In GA4, utm_content can populate Session manual ad content. The Manual report provides manual-tagging dimensions, including content, term, and campaign ID. Availability in your current navigation or dimension picker depends on the report setup and permissions. Use a suitable report or compatible exploration to examine the content values. The label ‘ad content’ does not mean your unpaid link has suddenly become paid advertising.

Suppose these profile destinations were used in different weeks. A seasonal interest change could explain part of the difference, and several posts may have pointed at each shared profile link. The tag identifies the distributed URL, not every post that influenced a visit. Compare launch dates and surrounding messages using your Content Calendar Generator plan or campaign register before choosing what to repeat. Record the next test as a hypothesis with a matched reporting window.

Separate first-user acquisition from key event attribution

Imagine a visitor first discovers the kit website through organic search. Several days later, they return through the tagged YouTube link and submit the planner. First user source / medium describes the original acquisition associated with that recognized user. Session source / medium describes the acquisition of the later session. Seeing the person under an original search source in one report and a campaign source in another is not automatically a tagging error.

User acquisition is useful when you ask how people were first acquired. Traffic acquisition is useful when you ask about sessions from new and returning visitors. A first-user campaign breakdown is therefore not a substitute for the session campaign breakdown used in the worked dataset. Also remember that recognizing a returning user depends on the property’s identity setup and available identifiers; a person using another browser may not appear as one continuous journey.

Event-scoped traffic-source dimensions, such as Source, Medium, and Campaign without the Session or First user prefix, support attribution of key-event credit. The reporting attribution model can distribute that credit across interactions and produce fractional values. Changing that model does not change the user-scoped or session-scoped acquisition dimensions. Keep an attribution analysis labeled separately from a count of lead-containing sessions; the two views describe different relationships to the action.

GA4 does not start a new session simply because it encounters another tagged campaign URL during an active session. Newly collected campaign information can relate to those events without replacing the existing session’s acquisition. This is one reason repeated test clicks may not create the session rows you expect. Do not put UTMs on internal navigation to force another campaign entry. Measure internal interactions separately and keep acquisition tags for the external distribution they describe.

Explain why platform clicks, sessions, and sales differ

Suppose platform reports show 900 clicks across the example campaign while GA4 shows 500 sessions. The difference is 400, but calling those 400 ‘lost visitors’ would invent a meaning the data does not support. Inspect what each platform calls a click, whether it measures the relevant external destination, and whether the periods align. A click total can cover repeat actions by one person or interactions that do not become a newly measured website session.

A visitor may click twice within one session, abandon the page before analytics runs, or use a browser where collection is limited. Another person may return later and create another session. Copied tagged links can produce visits without new clicks on the original platform. Consent and blocked scripts also affect collection. No universal adjustment converts clicks into expected sessions or proves the tags failed.

The example’s 40 generate_lead events are not 40 sales. Even a valid planner submission may be repeated, unqualified, or unrelated to a later purchase. Compare commercial outcomes with the appropriate order or lead records using a defined method. UTMs are descriptive campaign labels, not commission identifiers or proof that one platform caused a sale. A campaign row can support a useful observation while leaving business qualification and revenue reconciliation to other records.

Write totals with their units: 900 platform-reported clicks, 500 measured website sessions, and 32 sessions with a planner lead. Avoid presenting them as a single continuous funnel unless you can justify the matching definitions and population. Likewise, use the Follower Growth Calculator for account growth over a stated period, not as evidence that website sessions represent followers gained. Each measure answers a separate question about the campaign.

Troubleshoot the journey before rewriting the campaign

Start with the actual distribution location. Confirm that people can open a website destination there; a URL in ordinary caption text is not automatically clickable. Copy the real placement’s destination and compare its decoded tags with the register. A draft or manually typed address may differ from what people encounter. Record which path you tested.

Next inspect redirects with the website owner. Check a move from an old path to a new page, a language selector, and any intermediate landing service. The relevant question is whether campaign values reach the point where analytics collects them. A redirect can drop the query string. Conversely, a page may remove tags from the address bar after successful collection, so a clean-looking final address alone does not prove they were lost. Inspect collection timing as well.

Check the website measurement setup without trying to bypass consent. Is the correct analytics property receiving events? Does the consent state update when a visitor makes a choice? Is collection delayed, blocked, or configured differently on the landing page? Depending on implementation and eligibility, reports may include modeled data or leave measurement gaps. Read the report’s data-quality and modeling information; do not promise that modeling restores every visit or identifies the people who declined tracking.

Finally, inspect filters, spelling, date range, and reporting freshness. Repair_Kit and repair_kit can create distinct values, while a missing optional tag can legitimately leave a dimension unpopulated. Direct traffic or a ‘(not set)’ row can have several causes; neither is a diagnosis by itself. Test the suspected cause, document the finding, and correct future links or collection as appropriate. Rebuilding a URL does not guarantee a row appears or rewrite previously collected data.

Write an observation, a limitation, and one next action

A useful review note for the fictional dataset reads: ‘Instagram supplied 48% of the filtered campaign sessions. YouTube had the highest generate_lead session rate at 7.5%. Instagram comparison B had 10 lead-containing sessions from 90 sessions.’ This states what was counted without calling a platform the winner in every sense. Keep the underlying counts beside the rates so a reader can see the sample size and reproduce the calculations.

Add the limitation: ‘The variants ran in different weeks, the rows do not identify every influencing post, and planner leads are not verified sales.’ Choose one action, such as repeating the comparison angle with a registered link and matched window. Keep the destination and event definition stable where possible, so measurement changes are not mistaken for campaign improvements.

Export performance data from the destination analytics service and retain its report settings with the review. The builder’s CSV contains link fields, and its PDF is a link-review handoff, not a campaign-results report. Copying a URL does not install analytics, validate a remote page, shorten the address, or save a history. These boundaries make it easier to tell a link-preparation problem from a website-collection problem and an interpretation problem.

For the next release, open the Custom UTM Campaign Link Builder, import an existing address if useful, and review every managed field before generating the replacement. Blank optional fields remove their old managed values; unrelated query parameters and the fragment retain their meaning. Save the approved output in your own register. Consistent preparation and a short, reproducible review make later comparisons more useful without promising perfect attribution.

Questions about Read UTM Campaign Reports

Continue with another campaign guide

Explore a different link route or work on consistent naming and report interpretation. Return to the UTM Campaign Link Builder to apply the changes, or browse all free tools.