Skip to main content
MethodMarketingMarketing analyticsStarter

Read your GA4 reports and find what is working

A plain-language reading of the standard reports in an existing Google Analytics 4 (GA4) property: where visitors come from, which channels, campaigns and pages lead to key events, and what changed between two periods, with the traps (scopes, Direct and Unassigned, untagged links, unfinished data, small numbers) checked before any conclusion.

Use this task 6 quick questions, then open it in ChatGPT, Claude or Gemini.

This is a practical, copy-paste process for you have ga4 on your website and want to know which channels bring enquiries, bookings or sales. Use it when you want a clear structure, then add your own details.

Verified on 2026-11-0532 sources cited

Quick start

1. Write your question in one line, for example: "Which channel brought the most bookings last month?"
2. In GA4, open Reports, then Acquisition, then Traffic acquisition. If you see "Generate leads" instead of Acquisition, Traffic acquisition is inside it; otherwise type "Traffic acquisition" in the search box at the top. It shows where every visit came from.
3. Set the dates to the last full month, then read the whole month before as its own date range (GA4’s Compare > "Previous period" counts back the same number of days, so it does not always match a calendar month). Do not trust the last 2 days: GA4 needs 24 to 48 hours to finish.
4. Under the Key events column heading, click "All events" and pick the one action that matters, such as your booking form. Key events are GA4’s name for important actions.
5. Judge each channel by its key events, not its visits. Under about 20 key events, treat it as a hint, not proof (a rule of thumb, not a Google rule).
6. If one platform shows up in several rows (instagram, Instagram, ig), add them up and ask for tidy tracking tags (UTM tags) on every ad.
7. Check the total against your booking diary, and do not change any setting in Admin.

Before you start

Full method

How to use this method

  1. 1. Copy or downloadCopy the prompt below, or download the method file and attach it to your chat.
  2. 2. Paste into any AIChatGPT, Claude, Gemini or Copilot. Replace each [PLACEHOLDER] with your details, or let the AI ask you.
  3. 3. Add your materialAttach your own files, data or brief so the AI works from your facts, not guesses.

The prompt

Paste it as-is. The AI will ask for anything still in [BRACKETS].

Download .md
# Read your GA4 reports and find what is working
Method by Deepstory Research: deepstoryresearch.com/method/ga4-reports-for-beginners

You are a patient web analyst who reads GA4 reports for small-business owners and explains them without jargon. Follow the method below step by step. Goal: A plain-language reading of the standard reports in an existing Google Analytics 4 (GA4) property: where visitors come from, which channels, campaigns and pages lead to key events, and what changed between two periods, with the traps (scopes, Direct and Unassigned, untagged links, unfinished data, small numbers) checked before any conclusion.

## My inputs
- Your question: [YOUR_QUESTION]  (example: Did Google Ads or Instagram bring more bookings last month?)
- Website and main action: [WEBSITE_AND_MAIN_ACTION]  (example: Physiotherapy practice website, booking form (key event book_appointment))
- Period to compare: [PERIOD_TO_COMPARE]  (example: October 2026 against September 2026)
- Channels you use: [CHANNELS_YOU_USE]  (example: Google Ads, Instagram ads and posts, Google Business Profile)
- Report data: [REPORT_DATA]  (example: Traffic acquisition export or screenshot for October 2026)
- Spend per channel: [SPEND_PER_CHANNEL]  (example: Google Ads GBP 900, Instagram ads GBP 440 (actual spend for the period))

Before you start: if any input above is still in [BRACKETS] or missing, ask me for it in one short numbered list and wait. Do not guess. If I attached files (data, briefs, Deepstory evidence files), treat them as the primary source.

## Steps to follow
1. Turn the user’s question into one report, one key event, one period and one comparison.
2. Check the key event exists, when it was marked, and that the period has finished processing.
3. Read Traffic acquisition (all visits) or User acquisition (first visits) with the right dimension, and add up rows split by inconsistent tags.
4. Rank channels by key events and key events per 100 sessions, and by cost per key event if spend is given.
5. Check Direct, Unassigned, (not set), paid rows with no ads run, thresholds and sampling before trusting the split.
6. Say whether the key event counts enquiries or bookings, compare with the booking diary or CRM, then write the answer with its limits and next checks.

## Deliver the output in this structure
- The question, report, key event and period used
- Channel table with key events, key events per 100 sessions and cost per key event
- What changed against the comparison period
- Data warnings (unfinished days, tags, Unassigned, small numbers)
- Answer in plain words and three next checks

## Quality rules
- Never rank channels by sessions alone.
- Never compare numbers from User acquisition with Traffic acquisition, or GA4 key events with Google Ads conversions, as if they were the same count.
- Do not change any Admin setting; list setting changes as recommendations for the property owner.
- Label every number as FACT (with source and date), ESTIMATE (show the calculation) or ASSUMPTION.
- If a Deepstory evidence file is attached, use its figures as the reference values and cite its source references; tag claims that are not from it as [UNVERIFIED].
- End with: the 3 assumptions that matter most, what would change the answer, and the single best next step.

What you need to give the AI

InputPlaceholderExample
Your question[YOUR_QUESTION]Did Google Ads or Instagram bring more bookings last month?
Website and main action[WEBSITE_AND_MAIN_ACTION]Physiotherapy practice website, booking form (key event book_appointment)
Period to compare[PERIOD_TO_COMPARE]October 2026 against September 2026
Channels you use[CHANNELS_YOU_USE]Google Ads, Instagram ads and posts, Google Business Profile
Report data[REPORT_DATA]Traffic acquisition export or screenshot for October 2026
Spend per channel[SPEND_PER_CHANNEL]Google Ads GBP 900, Instagram ads GBP 440 (actual spend for the period)

Use it when

  • You have GA4 on your website and want to know which channels bring enquiries, bookings or sales.
  • You want to know what changed this month compared with last month or last year.
  • GA4 numbers confuse you: Direct, Unassigned, (not set), or different totals in different reports.

Best with: Any chat AI with a CSV export or screenshots of the reports; a browser agent with Viewer access only.

What you get

  • The question, report, key event and period used
  • Channel table with key events, key events per 100 sessions and cost per key event
  • What changed against the comparison period
  • Data warnings (unfinished days, tags, Unassigned, small numbers)
  • Answer in plain words and three next checks

The AI will ask you first

Answer these before the AI advises you; each one changes what it recommends.

  1. 1

    What one question do you want answered? For example: where do visitors come from, which channel brings the most bookings, or what changed this month?

    Each question needs a different report: Traffic acquisition for all visits, User acquisition for first visits, Landing page and Pages and screens for pages.

    If unknown: Which channels brought the most key events last full month, compared with the month before.

  2. 2

    Which visitor action matters most (booking form, purchase, call or WhatsApp click), does it finish on your own website or on a separate booking or payment site (for example a booking widget or card payment page), is it marked as a key event in GA4, and since when?

    Key events are counted only from the day they are marked, GA4 only sees actions on the website, and a separate booking or payment site can take the credit for bookings as a referral.

    If unknown: Check Admin > Data display > Events for starred events. Ignore events GA4 may have starred automatically (session_start, view_item, view_item_list, view_search_results, add_to_cart, add_to_wishlist and begin_checkout after a Google Ads link with personalised ads; the qualified-lead and converted-lead events under the Generate leads objective) unless they match a real action. If no starred event matches a real enquiry or sale, stop and use the tracking setup method.

  3. 3

    Which period do you want, and what should it be compared with?

    Unfinished days and unequal periods make normal changes look like trends.

    If unknown: The last full calendar month against the month before, plus the same month last year if the property has a year of data.

  4. 4

    Which channels do you use (Google Ads, Instagram or Facebook ads and posts, email, Google Business Profile), and do your ad and social links carry UTM tags (the ?utm_source=... part of a link)?

    Untagged or inconsistently tagged links land in the wrong channel or split into several rows.

    If unknown: Assume Meta ads are untagged. Set the first column to Session source / medium and look for instagram and facebook rows (for example l.instagram.com / referral); in the channel view these sit under Organic Social, not Referral.

  5. 5

    What GA4 access do you have (Viewer, Analyst, Editor or Administrator), and is Google Ads linked to GA4?

    Reading needs only Viewer; Google Ads campaign names need a link with auto-tagging.

    If unknown: Assume Viewer access and no Google Ads link; check Admin > Product links if you can see Admin.

  6. 6

    How many real bookings, leads or sales did you record in the same period, from your diary, booking system or CRM?

    GA4 counts website actions, not confirmed bookings; the gap shows how far to trust the channel split.

    If unknown: Ask for the monthly total and, if the diary, booking sheet or CRM records a source for each booking, the count per source for the same dates; put those counts beside the GA4 channel table. If the stated monthly total and the sheet disagree, ask which one counts bookings before comparing. If nobody has a total, label the answer as GA4-only.

  7. 7

    Did anything change in the period: a new website, form or consent banner, new campaigns, a holiday such as Ramadan or Eid, or a price change?

    Most sudden jumps or drops come from tracking or calendar changes, not marketing.

    If unknown: Assume nothing changed, but check the line chart for annotations and the Home page AI overviews.

  8. 8

    How much did you spend on each paid channel in the period?

    Needed for cost per key event; GA4 does not hold Meta spend unless Campaign data import (called cost data import before November 2025) has been set up.

    If unknown: Use actual spend for the same dates from each ad account, not a monthly budget. Do not split a combined figure between channels. If spend per channel is not known, skip cost per key event and compare key events and key events per 100 sessions only.

  9. 9

    If you are giving me numbers rather than a CSV or screenshot: which dates did the date picker show, on what day did you read them, and did the Key events column show one event or "All events"?

    The date picker offers pre-set ranges such as Last 7 days and Last 12 months as well as custom dates, so "last month" may not be the calendar month; a read in the first days after the period may be unfinished; and "All events" mixes in events GA4 may have starred automatically.

    If unknown: Ask for a screenshot or CSV that shows the date range and the Key events selector; until then, label the figures "as typed by the user, dates not confirmed".

The procedure, step by step

Read this before you run it, so you can check the AI actually followed the method.

  1. 1

    Frame the question

    Write it as: report, key event, period, comparison. For example: Traffic acquisition, book_appointment, October 2026, against September 2026.

  2. 2

    Find the reports

    Reports in the left menu. Older properties and those set up with "Get baseline reports" show the Life cycle collection (Acquisition, Engagement, Monetization, Retention); properties set up with a business objective show the Business objectives collection (for example Generate leads, with User acquisition, Traffic acquisition and Lead acquisition). If a report is missing, an Editor can add it from Library. Menus move; search the report name if a path differs.

  3. 3

    Check the key event

    Admin > Data display > Events: is the action starred, and since when? If not, stop and use the tracking method.

  4. 4

    Check the data is ready

    Use finished days only. Set the date picker to the period. For month against month, read the previous calendar month as its own date range (or set it as the comparison if your date picker lets you pick the dates); do not rely on Previous period.

  5. 5

    Read Traffic acquisition

    First column Session source / medium (or Session default channel group for the overview). In the Key events column choose the one key event. Note sessions, engaged sessions, engagement rate, key events and session key event rate. Add up rows that are the same platform tagged differently. Key events per 100 sessions (key events / sessions x 100) is not the same as GA4’s Session key event rate (sessions with at least one key event / sessions); say which one you use.

  6. 6

    Read pages

    Engagement > Landing page for the first page of each visit, with Session source / medium as a secondary dimension; Engagement > Pages and screens for all pages viewed.

  7. 7

    Check the warnings

    Data quality icon (thresholds, sampling), Direct, Unassigned and (not set) shares, paid rows with sessions but no ads run or no spend, annotations on the line chart, and the Home page AI overviews (released as Generated insights). Treat AI overviews and the Analytics Advisor assistant as leads to check in the reports, not as answers.

  8. 8

    Compare with real results

    Ask for the diary or CRM total for the period, and the count per source if the diary records one; show the gap and put the per-source counts beside the channel table.

  9. 9

    Write the answer

    Fill the output template: the answer in one sentence, the table, the warnings and three next checks. Send ad-platform questions to the campaign review method and tracking fixes to the tracking method.

Platform settings(show)
PlatformPathSettingNote
Google Analytics 4Reports > Acquisition > Traffic acquisitionSession default channel group (default), Session source / medium, Session campaign; Key events columnSession-scoped: where each visit began.
Google Analytics 4Reports > Acquisition > User acquisitionFirst user default channel group, First user source / mediumUser-scoped: how new users first arrived.
Google Analytics 4Reports > Engagement > Landing pageLanding page with secondary dimension Session source / medium
Google Analytics 4Reports > Engagement > Pages and screensPage path and screen class; Views; Key events
Google Analytics 4Reports > Library > Life cycle collectionAcquisition, Engagement, Monetization and Retention topicsShown by default for properties created before 27 March 2023 or set up with "Get baseline reports"; otherwise the Business objectives collection.
Google Analytics 4Date picker (top right of a report) > ComparePrevious period, Previous period (match day of week), Previous yearChanging the date range is not saved for other users. Previous period is the same number of days just before the range, not the previous calendar month.
Google Analytics 4Report > Add comparison (top right)Compare subsets side by side, for example Session default channel group = Paid Social against Organic SocialApplying a comparison is temporary; saving one is an Admin change.
Google Analytics 4Admin > Data display > EventsKey events (starred); up to 30 on a standard propertyRead only for this task.
Google Analytics 4Admin > Data display > Events > Attribution settingsReporting attribution model and key event lookback windowRead only for this task; changing the model rewrites past reports.
Google Analytics 4Admin > Data Settings > Data Retention (wording on Google’s help page; if your Admin menu does not show Data Settings, type "Data retention" in the search box at the top)Event data retention: 2 or 14 months on standard propertiesRead only for this task; affects explorations and funnel reports only.
Google Analytics 4Admin > Product links > Search Console links / Google Ads links / Google Business Profile linksCheck which products are linkedRead only for this task; linking is for the owner. A Business Profile link (June 2026) adds calls, bookings, directions and website clicks from the profile in its own Reports collection; these metrics cover only a rolling 6 months, so do not use them for year-on-year comparisons.
Google Analytics 4Advertising > Conversion performanceGoogle Ads conversions created from key eventsUse for media performance; Google Ads conversions do not appear in Reports or Explore.
Google Analytics 4Home > AI overviewsSummary of the top changes since your last visitReleased as Generated insights (in detail reports April 2025, on Home February 2026); the help page now calls them AI overviews.

Worked example

The figures below are illustrative. Use your own numbers and facts.

Scenario: A physiotherapy practice owner in Manchester runs Google Ads and Instagram ads and asks on 4 November 2026: "Did Google Ads or Instagram bring more bookings last month?" GA4 was set up by a freelance developer; the booking form sends a key event called book_appointment when a patient confirms an appointment slot, marked in June 2026.

Input: Traffic acquisition, 1 to 31 October 2026, first column Session source / medium, Key events column set to book_appointment. Rows: google / cpc 1,240 sessions, 31 key events. instagram / paid_social 2,100 sessions, 14 key events. Instagram / Paid_Social (one ad typed with capitals) 160 sessions, 2 key events. l.instagram.com / referral (ASSUMPTION: mostly Instagram profile and story links; untagged ad clicks can also land here) 380 sessions, 6 key events. google / organic 18, (direct) / (none) 9, all other rows 4 key events. Total 84 key events. Spend for 1 to 31 October from each ad account: Google Ads GBP 900, Instagram ads GBP 440. Booking diary: 76 bookings from the website form in October; phone bookings not tracked.

Output excerpt: October can be read: it ended 3 full days ago (1 to 3 November), past Google’s 24 to 48 hours of processing and the rule-of-thumb 3rd of the month; a little late data can still arrive for up to 7 days, so very small differences may still move. Instagram ads are split across two rows because of the capitals; added together they are 2,100 + 160 = 2,260 sessions and 14 + 2 = 16 key events. Google Ads (google / cpc): 31 booking-form key events, 31 / 1,240 = 2.5 per 100 sessions, GBP 900 / 31 = GBP 29.03 per key event. Instagram ads: 16 key events, 16 / 2,260 = 0.71 per 100 sessions, GBP 440 / 16 = GBP 27.50 per key event. All Instagram including profile links: 16 + 6 = 22 key events from 2,260 + 380 = 2,640 sessions, 0.83 per 100 sessions. Paid rows with no spend: none; both paid rows match an ad account with spend. Answer: Google Ads brought more booking-form key events (31 against 16 from Instagram ads), but each cost about the same (GBP 29.03 against GBP 27.50). Instagram ads are under the rule-of-thumb 20 key event line, so treat the cost difference as a tie and check again after November. book_appointment counts confirmed bookings, so it is compared with the diary: GA4 shows 84 key events against 76 diary bookings, 8 more, about 10.5% high (8 / 76 = 10.5%). That is inside the rule-of-thumb 20% and far from double, so the split is usable; the extra 8 may be repeat form submissions or test bookings by the developer (ask the property owner which counting method the key event uses and whether internal traffic is filtered). Phone bookings are not in it. Next checks: lowercase the Instagram ad tags; ask the developer to track clicks on the phone number as a separate key event; judge Google Ads bidding in Advertising > Conversion performance or the campaign review, not here.

Check your result

Before you trust the AI's output, confirm each of these.

  • ☐ The answer names the report, the key event, the period, the comparison and the date the data was read.
  • ☐ The period ended at least 2 full days before the data was read, or figures typed in by the user are labelled "dates not confirmed".
  • ☐ Rows for the same platform were added together and the arithmetic (sums, per 100 sessions, cost per key event) is shown.
  • ☐ Rows under about 20 key events are labelled as a direction only.
  • ☐ Direct, Unassigned and (not set) shares and any threshold or sampling warning are stated.
  • ☐ The GA4 key event total is compared with the diary or CRM total, or the answer says it is GA4-only.
  • ☐ The answer says whether the key event counts enquiries or bookings, and paid rows with sessions but no ads run or no spend are listed as unexplained.
  • ☐ No Admin setting, link, saved comparison, annotation or dashboard was created or changed.
What you'll get(show template)
## GA4 reading: [QUESTION]

Data read on [DATE, or "typed by user, not confirmed"] | Report: [Traffic acquisition / User acquisition / Landing page] | Key event: [NAME, marked on DATE] | Period: [DATES] against [DATES]

### Answer
[One or two plain sentences.]

### Channel table
| Channel or source / medium (rows added together) | Sessions (share) | Key events (share) | Key events per 100 sessions | Spend | Cost per key event | Change against comparison |
|---|---|---|---|---|---|---|
| [row] | [n (%)] | [n (%)] | [n] | [amount or n/a] | [amount or n/a] | [+/- %, or "not read"] |

### Pages (if asked)
| Landing page | Sessions | Key events | Average engagement time per session (or Engagement rate if the report shows it) |
|---|---|---|---|
| [page] | [n] | [n] | [time or %] |

### Data warnings
| Check | Result |
|---|---|
| Period finished processing | [yes / no] |
| Direct, Unassigned and (not set) share of key events | [%] |
| Split or untagged rows | [list] |
| Rows under about 20 key events (direction only) | [list] |
| Thresholding or sampling icon | [none / details] |
| Paid rows with no spend or no ads run | [list, or none] |
| GA4 key events against diary or CRM | [n against n, gap %; key event counts enquiries / bookings] |

### Next checks
1. [Check]
2. [Check]
3. [Check]

Quality checks built into the prompt

  • • Never rank channels by sessions alone.
  • • Never compare numbers from User acquisition with Traffic acquisition, or GA4 key events with Google Ads conversions, as if they were the same count.
  • • Do not change any Admin setting; list setting changes as recommendations for the property owner.
  • • Every number labelled FACT, ESTIMATE or ASSUMPTION; claims outside attached evidence tagged [UNVERIFIED].

Decision rules

If/then rules the AI should apply. Heuristic rules are a practitioner's rule of thumb, not a platform rule.

IfThenType
The question is where visits come from, new and returning together.Open Reports > Acquisition > Traffic acquisition. Its dimensions start with "Session" and describe how each visit (session) began; the default rows are Session default channel group. Switch the first column to Session source / medium or Session campaign for detail, and pick the one key event that matters in the Key events column.Platform rule
The question is how new visitors first found you, or two reports give different numbers for the same source.Use Reports > Acquisition > User acquisition, whose dimensions start with "First user". Do not compare its numbers with Traffic acquisition: one is scoped to users, the other to sessions. Google’s example: someone who first came from Google and then typed your address 10 more times shows 11 sessions under Google in User acquisition; in Traffic acquisition the 10 return visits stay under Google if they fall inside the key event lookback window (90 days by default), otherwise they show as direct.Platform rule
You compare GA4 key events with conversions in Google Ads, or want to judge ad spend from GA4.Treat them as different counts. Key events are GA4’s important actions (GA4 called them conversions until 27 March 2024), shown in Reports and Explore, which Google designs for traffic acquisition and behaviour and does not recommend for measuring media performance. Google Ads conversions are created from key events, appear only in the Advertising section (Conversion performance report), and can use data-driven attribution. For ad budget decisions, use that report or the ad platform review.Platform rule
A key event was marked recently, or you are not sure which events are key events.Check Admin > Data display > Events; key events carry a star. Marking counts from that moment and does not change past data, and it can take up to 24 hours to show in standard reports. Compare only periods after the date it was marked. Some events are starred automatically: linking Google Ads with personalised ads (on by default) stars session_start, view_item, view_item_list, view_search_results, add_to_cart, add_to_wishlist and begin_checkout, and the Generate leads objective sets up the qualified-lead and converted-lead events. So never read the Key events column on "All events": always pick the one key event that matches a real enquiry or sale.Platform rule
You read channels such as Direct, Organic Search, Paid Search, Paid Social, Organic Social, Cross-network or Unassigned.Read them by Google’s fixed rules. Paid Social needs a social source AND a paid medium (one containing "cp", exactly "ppc" or "retargeting", or starting with "paid"); Organic Social needs a social source or a social medium, so Instagram ad links with no paid utm_medium cannot count as Paid Social. Cross-network is Google Ads Performance Max, Demand Gen, Smart and similar campaigns, so it is paid Google traffic too. Direct is source "(direct)" with medium "(none)" or "(not set)". Unassigned means no rule matched. Channel rules ignore capitals; source and medium rows do not. Traffic goes to the first channel it matches in the group’s order, Organic Social comes before Referral, and Google’s social source list includes l.instagram.com and m.facebook.com, so a row such as l.instagram.com / referral sits under Organic Social, not Referral.Platform rule
A Paid Search, Paid Social or Cross-network row has sessions, but the user says no ads ran or gives no spend for that channel.Do not credit the user’s ads. Switch the first column to Session source / medium and Session campaign to see what is behind the row, and ask whoever manages the ads or the website (for example the agency) whether ads ran in the period and how links are tagged. Google puts visits in these channels from Google Ads, but also from links tagged by hand: Paid Search and Paid Social take a search or social source with a paid medium (one containing "cp", exactly "ppc" or "retargeting", or starting with "paid"), and Cross-network takes any campaign name containing "cross-network". Report the row as "unexplained paid traffic" in the Data warnings until it is explained, and leave it out of cost per key event.Platform rule
You see an AI Assistant channel, or wonder where visits from ChatGPT or Google’s AI Overviews go (as of November 2026).The AI Assistant channel, added on 13 May 2026, covers visits from sources such as ChatGPT, Gemini, Copilot, Deepseek or Grok. Google sets the medium to "ai-assistant" and the campaign to "(ai-assistant)" when the referrer matches its list of AI assistants, and the channel takes medium "ai-assistant". Clicks from Google’s AI Overviews and AI Mode stay in Organic Search. Before that date these visits were classified by other rules, so a new AI Assistant row is not by itself growth.Platform rule
The AI Assistant row looks small, or Unassigned holds rows whose source is chatgpt.com or another AI assistant.Rule of thumb (Google’s pages do not describe this case directly; confirm in Session source / medium): some AI tools add their own tag to links, for example utm_source=chatgpt.com with no utm_medium. Google says missing UTM values show as (not set), Unassigned means no channel rule matched, and chatgpt.com is not on Google’s search or social source lists, so such a visit can show as chatgpt.com / (not set) under Unassigned instead of AI Assistant. Switch the first column to Session source / medium, add those rows to AI-assistant traffic for this answer and say so; they usually come from links tagged by the AI tool, not from the business’s own tags. Where your property shows the Source group dimension, Google says it groups sources like ChatGPT (OpenAI) and Perplexity.Heuristic
You see "(not set)" or a large Unassigned row.For a "(not set)" Session source / medium, which shows as Unassigned, check the causes Google lists: links with incomplete or wrong UTM tags, a missing session_start event (for example a Tag Manager trigger not set to Initialization), or consent banner set-up errors. For "(not set)" Google Ads campaign or ad group names in Advertising reports, check whether Google Ads is linked, auto-tagging is on, and every Google Ads account advertising the site is linked. Report the share of key events in each case and pass the fix to whoever manages the tags or the link.Platform rule
The booking, payment or calendar step runs on another website, or that tool’s domain shows up as a Referral source.Do not count that domain as a channel. Google gives third-party payment processors, where users return to your site after checking out on the other domain, as a common case of a referral you do not want, and offers two fixes: List unwanted referrals (Admin > Data collection and modification > Data streams > web stream > Configure tag settings) or cross-domain measurement, which measures one user and one session across both domains. Report that row’s key events as booking-tool returns and recommend either fix to the property owner; both are Admin changes, not part of this task.Platform rule
The same platform appears in several rows (for example instagram / paid_social, Instagram / Paid_Social and ig / social).Add those rows together for this answer, then recommend fixing the links. Google says UTM values are case sensitive (google and Google are different), that links should always carry utm_source, utm_medium and utm_campaign, with one source per platform, one medium per channel (for example paid_social) and lowercase. Where your property shows the Source group dimension (released June 2026), it groups variants such as instagram and ig into one Instagram value.Platform rule
You judge the quality of visits from a channel or page.An engaged session lasted 10 seconds or longer, had a key event, or had 2 or more page or screen views. Engagement rate is engaged sessions divided by sessions; bounce rate is the opposite. Traffic acquisition shows Engagement rate by default; in other reports an Editor may need to add it.Platform rule
Someone suggests changing the attribution model to change the Traffic acquisition numbers.It will not change them. Session and first-user dimensions always use paid and organic last click; the reporting attribution model (data-driven by default) only changes key event credit on event-scoped dimensions such as Source, Medium and Default channel group. First click, linear, time decay and position-based models were removed in November 2023. Changing the model rewrites past and future reports, so leave it to the property owner.Platform rule
Rows are missing, a report says thresholding was applied, or an exploration shows a sampling warning.Thresholds may hide data that includes demographics, and rows with search queries when there are too few users; you cannot switch them off, but a wider date range helps. Explorations and reports may be sampled when a query passes 10 million events on a standard property, and the data quality icon shows the share of data used. State either warning in the answer.Platform rule
An exploration or funnel shows no data for older months.That is data retention: standard properties keep user- and event-level data for 2 or 14 months, and this limits explorations and funnel reports, not standard reports. Use standard reports for year-on-year comparisons. Do not change the setting yourself; shortening it deletes data at the next monthly run.Platform rule
You need the Google searches that brought visitors, or Google Ads campaign names show as "(not set)".Search queries need a Search Console link; its reports hold up to 16 months, arrive 48 hours after Search Console collects them, and the Search Console collection stays unpublished until an Editor publishes it from Library. Google Ads campaign detail needs a Google Ads link with auto-tagging; data appears within 48 hours of linking. Linking is an Admin change: recommend it, do not do it.Platform rule
You compare two periods.Compare like with like: full months with full months (read each month as its own date range, or set the comparison dates yourself if your date picker offers that, because Compare > Previous period uses the same number of days just before the range, so October is compared with 31 August to 30 September), or full weeks with Compare > Previous period (match day of week). For seasonal businesses (Ramadan, Eid, summer, school holidays) also compare with the same month last year. When months have different numbers of days, compare per-day figures too.Heuristic
You decide which channel works best.Rank channels by the one key event that matters, then by key events per 100 sessions, then by cost per key event when the user gives spend. Never rank by sessions alone: a channel with many short visits can bring fewer bookings.Heuristic
A row has few key events in the period.Under about 20 key events in a row, call the result a direction, not a finding, and say that a difference of a handful could be chance. Widen the date range or wait another month before moving budget.Threshold: about 20 key events per row (rule of thumb)Heuristic
You read today, yesterday, or a month that ended in the last 2 days.Do not treat it as final. Google says processing can take 24 to 48 hours, standard reports are usually ready about 3:30 pm property time the next day, and some data arrives up to 7 days late. Rule of thumb: read a finished month on or after the 3rd of the next month, and say so in the answer.Heuristic
The question is about bookings, leads or sales.GA4 counts website actions, not confirmed bookings. First say what the key event counts: an enquiry (for example generate_lead, which Google’s recommended events define as a user who "submits a form or a request for information", or any contact form) or a completed booking or sale (Google lists close_convert_lead separately, for a lead that "became a converted lead (a customer)"). If nobody knows which it counts, say so and label the channel ranking unconfirmed. For a completed booking or sale, compare the month’s key event total with the booking diary or CRM. Rule of thumb: if GA4 is more than about 20% away, say so and treat the channel split as approximate; if GA4 is more than double or less than half the diary, do not rank channels until the person who set up tracking explains the difference. For an enquiry event, expect more key events than bookings: report the diary total beside it as the enquiry-to-booking check, not as a tracking error. Phone calls and WhatsApp chats are missing unless tracked. If the business has a Google Business Profile and many bookings come by phone or messages, recommend that the owner (Editor or Administrator in GA4, Owner or Manager on the profile) links it under Admin > Product links > Google Business Profile links; its calls, bookings, messages and website clicks then appear in their own Reports collection, for a rolling 6 months only. Linking is the owner’s step, not part of this task.Threshold: about 20% gap: split approximate; more than double or less than half: do not rank (rules of thumb)Heuristic

Failure modes

Common mistakes with this method, how to spot them, and how to fix them.

MistakeSymptomFix
Reading the Key events column on "All events" when GA4 has starred session_start or view_item automatically.Key events are close to sessions, the session key event rate is near 100%, and every channel looks like it converts.Pick the one key event that matches a real enquiry or sale in the Key events column; note the auto-starred events and ask the property owner whether to unstar them.
Counting a booking, payment or calendar tool on another domain (for example a booking widget or payment page) as a Referral channel.That domain looks like one of the best channels for key events, while ads and social look weaker than the diary suggests.Do not count it as a channel; report its key events as booking-tool returns with no source, and recommend that the property owner adds the domain to List unwanted referrals or sets up cross-domain measurement (an Admin change, not part of this task).
Counting test bookings and staff visits as real ones.Key events appear on days with no diary booking, or GA4 runs above the diary right after a website or form change.Ask whether the developer or staff submitted test bookings in the period and subtract them if known; recommend that the property owner sets up internal traffic filtering (an Admin change with a permanent effect, not part of this task).
Ranking channels by sessions or users.The channel with the most visits gets more budget, but bookings do not rise.Rank by the one key event, then key events per 100 sessions, then cost per key event.
Comparing User acquisition numbers with Traffic acquisition numbers.The same source shows different totals and the owner thinks GA4 is broken.Pick one report for the question: User acquisition for first visits, Traffic acquisition for all visits.
Instagram or Facebook ad links without a paid utm_medium, or tagged with mixed capitals.Paid Social looks tiny, Organic Social looks great (with source / medium rows such as l.instagram.com / referral inside it), and one campaign appears in several rows.Add the split rows together for now; recommend lowercase tags with utm_source, utm_medium (for example paid_social) and utm_campaign on every ad.
Treating GA4 key events and Google Ads conversions as the same number.Google Ads shows more or fewer conversions than GA4 and someone "fixes" settings.Explain they use different scopes and attribution; use Advertising > Conversion performance or the ad platform for ad decisions.
Reading yesterday or a month that has just ended as final.A drop that disappears a day later.Use finished days; read a month on or after the 3rd of the next month (a rule of thumb).
Reading a sudden jump in Direct, Unassigned or (not set) as a real change.A channel falls and Direct or Unassigned rises by a similar amount, often after a site, form or consent banner change.Check tags, the consent banner, links and Realtime for the key event; send fixes to the tracking method before concluding.
Drawing conclusions from a handful of key events.Budget moves after a 3 against 5 difference, and the next month reverses.Label rows under about 20 key events (a rule of thumb, not a Google rule) as a direction only; widen the date range.
Changing the attribution model, data retention or key events to make a report look right.Past reports change or data is deleted.Leave Admin settings alone; write recommendations for the property owner.
Comparing periods across the date a key event was marked or a channel rule changed.Key events jump from zero, or an AI Assistant row appears and looks like growth.Compare only periods after the change, and say so in the answer.
Treating chatgpt.com / (not set) rows in Unassigned as broken tracking.The AI Assistant row looks small while Unassigned holds rows whose source is chatgpt.com or another AI assistant.Rule of thumb: add those rows to AI-assistant traffic for this answer and say so; they usually come from links tagged by the AI tool, not from the business’s own tags.
Reading visits from unknown websites as real visitors.Referral rows from sites you do not recognise, with very short visits and no key events, or a Hostname (in Pages and screens) that is not your website.Leave them out of the answer and say so; recommend that the property owner considers a hostname data filter (available as an Include filter since September 2026; an Admin change with a permanent effect, not part of this task).

Agent mode: if an AI agent runs this unattended

May do without asking

  • • Open reports with Viewer access (Analyst only if the owner chose it), change the date range, the first column and the key event shown.
  • • Apply temporary comparisons and filters in a report without saving them.
  • • Download a report as a CSV file for the user and do the arithmetic.
  • • Read Admin pages (key events, attribution settings, data retention, product links) without changing them.

Must stop before (the user decides and does these)

  • • Changing any Admin setting: key events, attribution, data retention, channel groups, data filters or data streams.
  • • Creating, linking or unlinking Google Ads, Search Console, Business Profile or other product links.
  • • Saving or publishing comparisons, annotations, collections, reports or dashboards to the property.
  • • Uploading or importing data, such as Campaign data import (cost data) or a Meta, TikTok or other ad-platform connection, into GA4.
  • • Accepting any terms or data sharing settings.
  • • Changing ad budgets, bids or campaigns, or anything that changes spend or billing.
  • • Granting or removing user access, or sharing exports outside the business.
  • • Deleting anything.

Evidence for this method

This method runs on your own material: your files, brief or code. If your task needs market or economic numbers, attach a sourced context file from the evidence library alongside the prompt.

Browse evidence

Run next

Methods that chain well after this one.

Deepstory methods are process guidance for AI tools. Review AI output before relying on it; legal, tax, medical and investment decisions need a qualified professional.