Reporting

Measure conversion lag before you judge the last few days

Works out, per campaign, how many days after the click conversions are recorded, estimates how incomplete the last few days still are and says whether a bidding decision should wait.

When to use it

When conversions look down over the last few days and you want to know whether late-arriving conversions explain it before you touch bids or budgets.

What you get

A lag curve per campaign, a raw versus lag-adjusted comparison of the last 7 days, a wait or investigate decision for each campaign and the number of days your account needs before a week can be judged.

Does it change anything in my account?

No. The assistant only reads your data and reports back.

Works with

Google Ads

Fill in before sending

[account name]

In my Google Ads account [account name] the cost per conversion for the last 7 days looks higher. Before I act, I want to know whether this is a real drop or just conversions that have not been recorded yet. This is a read-only review, do not change anything. 1. Know the conversion actions. List the actions with list_conversion_actions, including counting type and click-through lookback window (click_lookback_days). In the following steps only use actions that are included in the Conversions column (included_in_conversions). 2. Build the lag curve. With google_ads_query, pull the last 90 days per campaign segmented by segments.conversion_lag_bucket, with metrics.conversions and metrics.conversions_value. Give the date range with BETWEEN and explicit dates, because DURING LAST_90_DAYS is not a valid GAQL value. For each campaign, work out what share of conversions arrives within the first day, the first 3 days, the first 7 days and the first 14 days. 3. Estimate what is still missing. Pull the last 14 days day by day: clicks, cost, metrics.conversions (by click date) and metrics.conversions_by_conversion_date (by the date the conversion happened). Using the curve from step 2, estimate how much of each day's conversions is still to come. For example, if a campaign gets 60% of its conversions within 3 days, a day from 3 days ago currently shows about 60% of its final number. 4. Correct the comparison. Compare the last 7 days with the 7 days before in two ways: raw numbers and lag-adjusted estimates. Label every adjusted number as an "estimate". Decision rules: 1. If a campaign has fewer than 30 conversions in the last 90 days, its lag curve is not reliable. Do not estimate for it, write "not enough data" and show the raw number only. 2. If the adjusted estimate is within 15% of the previous week, attribute the drop to lag and do not recommend any bid, budget or target change. 3. If the adjusted estimate still shows a clear drop, lag does not explain it. Say so plainly and list the first three places to look next. 4. If a campaign needs more than 7 days to collect 90% of its conversions, do not recommend changing its target CPA or target ROAS based on the last 7 days. Format the output like this: a three-sentence summary first. Then one table per campaign row: campaign, conversions in the last 90 days, share in the first day, share in the first 7 days, days needed for 90% of conversions, last 7 days raw conversions, last 7 days estimated final conversions, raw and estimated cost per conversion, decision (wait, investigate, not enough data). End with one rule for the team: in this account, how many days to wait before judging a week.

1. Know the conversion actions. List the actions with list_conversion_actions, including counting type and click-through lookback window (click_lookback_days). In the following steps only use actions that are included in the Conversions column (included_in_conversions).
2. Build the lag curve. With google_ads_query, pull the last 90 days per campaign segmented by segments.conversion_lag_bucket, with metrics.conversions and metrics.conversions_value. Give the date range with BETWEEN and explicit dates, because DURING LAST_90_DAYS is not a valid GAQL value. For each campaign, work out what share of conversions arrives within the first day, the first 3 days, the first 7 days and the first 14 days.
3. Estimate what is still missing. Pull the last 14 days day by day: clicks, cost, metrics.conversions (by click date) and metrics.conversions_by_conversion_date (by the date the conversion happened). Using the curve from step 2, estimate how much of each day's conversions is still to come. For example, if a campaign gets 60% of its conversions within 3 days, a day from 3 days ago currently shows about 60% of its final number.
4. Correct the comparison. Compare the last 7 days with the 7 days before in two ways: raw numbers and lag-adjusted estimates. Label every adjusted number as an "estimate".

Decision rules:
1. If a campaign has fewer than 30 conversions in the last 90 days, its lag curve is not reliable. Do not estimate for it, write "not enough data" and show the raw number only.
2. If the adjusted estimate is within 15% of the previous week, attribute the drop to lag and do not recommend any bid, budget or target change.
3. If the adjusted estimate still shows a clear drop, lag does not explain it. Say so plainly and list the first three places to look next.
4. If a campaign needs more than 7 days to collect 90% of its conversions, do not recommend changing its target CPA or target ROAS based on the last 7 days.

Format the output like this: a three-sentence summary first. Then one table per campaign row: campaign, conversions in the last 90 days, share in the first day, share in the first 7 days, days needed for 90% of conversions, last 7 days raw conversions, last 7 days estimated final conversions, raw and estimated cost per conversion, decision (wait, investigate, not enough data). End with one rule for the team: in this account, how many days to wait before judging a week.

Open in Claude Open in ChatGPT Discuss in the forum

Example result

An example of the output this prompt produces. Its structure comes from the tools' real output; numbers and names are fictional.

Example output (based on a fictional account, not real customer data)

Summary. Raw cost per conversion for the last 7 days looks 31% higher than the week before. Adjusted for the lag curve, the gap on the Search campaigns shrinks to 6%, so no bid change is needed there. The Demand Gen campaign only had 11 conversions in 90 days, so it was not estimated.

Campaign	Conversions, 90 days	First day share	First 7 days share	Days to 90%	Last 7 days raw	Last 7 days estimate	Raw / estimated cost per conversion	Decision
Search Services	214	58%	86%	9	14	18.5 (estimate)	$41.20 / $31.20	Wait
Search Brand	96	81%	97%	4	9	9.6 (estimate)	$10.40 / $9.80	Wait
Demand Gen Video	11				1		$119.00 / n/a	Not enough data

Team rule: in this account, wait at least 9 days after a week ends before judging it.

More prompts

ReportingRead-only

Build a monthly scorecard of your brand's visibility in AI answers and compare it with last month

Measures the same question set in ChatGPT, Perplexity and Gemini every month, checks whether you are cited in Google's AI answers, and combines this with AI referrals in GA4 and brand searches in Search Console. States measurement noise and tool limits openly.

When to use itYou want to track your visibility in AI search every month on the same yardstick, not just once, and see whether your work is paying off.

Build this month's scorecard for the visibility of my brand [brand name] and my site [domain] in AI answers. Market: [country], language: [language code, e.g. en]. My competitors: [up to 3 competitor brand names]. My fixed question set: [5 to 6 questions that do not name my brand, phrased the way a real customer would ask]. Last month's scorecard: [paste last month's table, or write "first month"]. This is read-only.

Keep the question set identical every month, otherwise the months cannot be compared. If I gave no questions, propose 6 (2 of the "best [service] in [city]" type, 2 comparisons, 1 on price, 1 how-to), get my approval and say that the set will not change.

1. Credit check. Use seo_history to find queries run recently with these arguments. Call the reusable ones again with the same arguments, they return at no cost. Tell me how many new calls are needed.
2. Share in AI chats. Run seo_geo_tracker with brand=[brand name], brand_domain=[domain], the competitors and the question set, engines='chat_gpt,perplexity' and repeats=2. The tool runs at most 24 language model queries per call (questions × engines × repeats) and lowers the repeats itself when that limit is exceeded, so report prompts_used and calls from the reply to keep the months on the same scale. Take the brand mention rate, the number of answers citing my domain, the rate per engine, each competitor's rate and the most cited source domains. This tool does not return the answer text, only rates and source domains.
3. Google's AI answers. Run the 3 questions with the strongest commercial intent through seo_ai_overview: target=[domain], location=[country name in English, e.g. United Kingdom], language=[language code]. For each, record whether an AI answer appears, whether my domain is cited, and which domains are cited instead. Careful: the references list can come back empty while the sources sit only as links inside the ai_answer text; in that case do not trust target_cited, count the domains of the links in the text. ai_answer is cut at 1,200 characters, so write "not cited" as "not in the visible part", not as a certainty.
4. Overall mentions. Run seo_geo_citations with brand=[brand name] and take the mention counts in Google AI answers and ChatGPT. If my brand name is made of common words (for example two everyday words), this number can be inflated by unrelated text; say so and use this row for the trend, not as an exact figure.
5. Real visits. In my GA4 property [GA4 property], use ga4_report with dimension='source' to pull sessions and key events for the last 30 days and the previous 30 days. Add up chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai and any other source that clearly belongs to an AI assistant. Leave out paid traffic tagged with utm parameters (for example an ad source like chatgpt_ads), because that is not organic visibility.
6. Brand searches. In my Search Console property, use get_report with platform='search_console', breakdown='query', for the last 28 days and the previous 28 days, and add up impressions and clicks for queries that contain my brand name. Being mentioned in AI answers often lifts brand searches, so this row is an indirect signal.

Format the output like this. First a four-sentence summary. Then the scorecard table: metric, this month, last month, change, note. Rows: mention rate in ChatGPT, mention rate in Perplexity, answers citing my domain, each competitor's mention rate, questions where Google's AI answer cites me, seo_geo_citations mention counts, sessions and key events from AI assistants, brand search impressions. Then a per-question table: question, mentioned in ChatGPT, mentioned in Perplexity, cited by Google, most cited competitor source. Then the 10 most cited source domains and how many of them are my site, a competitor's site or a third party.

Decision rules: with two repeats, label any rate change below 10 points as "within noise" and do not read it as a real change. If this is the first month, leave the change column empty and record this month as the baseline. End with one recommendation for next month, and give the table in a form I can copy and paste back next month.

ReportingChanges after your approval

Monthly paid channel efficiency ranking with a budget split proposal

A ranking of every paid channel by cost per conversion, with spend share vs conversion share and a proposed budget split for next month.

When to use itWhen a new month is starting and you need to decide which paid channels deserve more or less budget.

Compare last calendar month across Google Ads, Meta, TikTok and Microsoft Ads.

Build one table with these columns for each channel:
- spend
- conversions
- cost per conversion
- conversion value and ROAS, where value is tracked
- share of total spend (%)
- share of total conversions (%)

Rank the channels by cost per conversion. My target cost per conversion is [target CPA].

Decision rules:
- If a channel's share of conversions is at least 10 points below its share of spend, mark it as a candidate to cut.
- If a channel's share of conversions is at least 10 points above its share of spend, mark it as a candidate to scale. For Google Search, only do this if impression share analysis shows the campaigns are losing impression share to budget.
- If a channel has fewer than 10 conversions, mark it "not enough data" and leave its budget unchanged.

Then propose next month's split in a table with current monthly spend, proposed spend and a one-line reason for each channel. Keep the total within [monthly budget].

Do not apply anything. If I approve, show a preview of each budget change and wait for my confirmation.

ReportingChanges after your approval

Monthly Google Ads client report with a shareable white-label link

Builds a one-page, jargon-free monthly report for a client with exact calendar-month figures versus the previous month, explains every big KPI move from campaign data and creates the white-label link only after you approve the draft.

When to use itWhen you have to send a client their monthly Google Ads results in plain language and explain what moved and why.

Prepare the monthly Google Ads report for [client account] for [month, e.g. 2026-09], compared with the previous calendar month.

Data:
- Use list_accounts to find the customer ID of [client account].
- Pull exact calendar-month figures with get_report (platform google_ads) using start_date and end_date for [month], then again for the previous month. Get both the account totals and the campaign-level rows.
- If you need more campaign detail, use performance_report at campaign level.
- KPIs: cost, clicks, conversions, conversion value, CPA, ROAS and CTR, each with the % change. If conversion value is not tracked in this account, leave ROAS out and say why in one line.

Write it for a client who is not a marketer:
1. A 3-sentence executive summary covering the result, the main driver and what comes next.
2. A KPI table with this month, last month and % change.
3. The top 3 campaigns by conversions. Also name the campaign with the highest CPA among those that took more than 10% of the month's total spend.
4. Three concrete next steps we will take next month, each tied to something in the data.

Flag every KPI that moved more than 15%. Explain each one with evidence from the campaign-level rows, such as a spend shift between campaigns, a CPA change in a specific campaign or a conversion drop. If the data does not explain the move, say so plainly and do not guess.

Keep it to one page with no jargon. Show me the draft first and stop.

After I approve the draft, create the shareable white-label link with client_report_link using brand_name [agency or client brand name]. Note that this tool only takes a rolling days window ending now, not a named month. So before creating it, tell me which exact dates the link will cover with the days value you plan to use, and whether that matches [month]. If it does not match (for example we are well past month end), tell me that clearly and let me choose: create the link anyway with its rolling window, or skip the link and I will send the approved draft myself. Then give me the URL.

ReportingRead-only

Agency portfolio triage: which client accounts need attention this week

Compares every connected Google Ads client account over the last 30 days versus the previous 30, adds month-to-date budget pacing, ranks accounts by risk and gives account managers one data-backed first check for the top 5. Read-only.

When to use itWhen you manage many Google Ads client accounts and need to know which ones to open first this week.

Triage all connected Google Ads client accounts. Compare the last 30 days with the previous 30 days.

How to get the numbers:
- Run portfolio_report with days=30, then again with days=60. For each account, derive the previous 30 days by subtracting the 30-day figures from the 60-day figures (cost, clicks, conversions, conversion value), then calculate CPA and ROAS for both periods yourself.
- Use list_accounts to get each account's customer ID, then run budget_pacing_alerts with threshold_pct=10 for each account. Pacing is a month-to-date projection to month-end, not the 30-day window, so label that column "pacing (this month)".
- If an account does not track conversion value, mark ROAS as n/a instead of treating it as below target.

Mark an account as at risk if it meets any of these conditions:
- CPA up more than 20%
- Conversions down more than 20%
- ROAS below [target ROAS, or list per-account targets]
- Any campaign flagged by budget pacing as over or under by more than 10% by month-end

Deliver one table with these columns: account, spend, conversions, CPA, ROAS, % change in CPA and conversions, pacing (this month), risk reasons. Sort it by the number of risk reasons, then by spend.

Below the table, list the 5 accounts our account managers should open first this week. For those 5 only, run conversion_tracking_health (days=30) and search_terms_report (days=30, sort by cost) to ground the advice. Give one concrete first check for each, for example "search terms in campaign X are spending without conversions" or "conversion action Y recorded nothing in 30 days", based on what the data shows. If neither report explains the account's problem, say what you would look at next instead of guessing.

Finish with the accounts that improved the most. These are good news for client calls.

This is read-only. Do not change anything in any account.