Copy it.
Paste it.
Ask your data.

Take control of my GA4 tab. Tell me which property and date range is on screen, then list every report you can reach from here. Read only: don't change any setting, filter, or date range unless I ask.

Walk me through connecting GA4 to you with a Google service account, so you can read GA4 directly. Go one step at a time. Wait for me to say "done" before the next step, and do every terminal step yourself instead of asking me to type commands.

1. Check my computer: tell me if I'm on Mac or Windows, make sure uv is installed (install it if not), then install Google's GA4 server with: uv tool install --python 3.12 analytics-mcp
2. In the browser: help me create a Google Cloud project (or pick one I already have), and give me the exact links to turn on the "Google Analytics Data API" and "Google Analytics Admin API" for that project. Ask me for the project ID when I'm done.
3. In the browser: walk me through creating a service account in that project called ga4-reader with no extra roles, then creating a JSON key for it and downloading it.
4. Move the key I downloaded into a private folder: ~/.config/gcloud-keys/ga4-reader.json on a Mac (lock it so only I can read it), or %USERPROFILE%\.config\gcloud-keys\ga4-reader.json on Windows. Never print what's inside the key. Just read the service account's email address out of it and show me that email.
5. In GA4: walk me through Admin, Account access management, then adding that email as a Viewer. If I have more than one GA4 account, tell me to repeat it for each one.
6. Test it: list every GA4 property the service account can see, with its property ID and website, and match each one to my stores.
7. Connect it to Claude Code with: claude mcp add analytics-mcp --scope user -e "GOOGLE_APPLICATION_CREDENTIALS=<key path>" -e "GOOGLE_PROJECT_ID=<project ID>" -- analytics-mcp Then tell me to restart Claude and type /mcp to confirm it's connected.

If any step fails, tell me what the error means in plain English and how to fix it before moving on.

Read the ASC GA4 Standard v1.2 event list here: https://docs.google.com/spreadsheets/d/1Lrx7-qL3p17tMA7Dpox2q7yIAfmeIZ6W7NtkWBTU2j0/export?format=csv&gid=0

Then:
1. List every GA4 property I can access and which store each one belongs to.
2. For each property, pull event counts by event name for the last 90 days. Where the event_owner parameter is available, include it so we can see which vendor is sending each event.
3. Compare each store against the ASC 1.2 list. Mark every ASC event as Firing or Missing. If an event only makes sense for a tool the store may not have (parts forms, video calls, chat), mark it "No activity, confirm with vendor" instead of Missing.
4. Assign every missing event to the vendor responsible:
   - Website provider: page views, item and list page views, CTA, menu, media, special offer, form engagement, click to call, and sales form submissions
   - Service scheduler: service and service appointment submissions
   - Call tracking provider: voice engagement and voice submission events
   - Chat or texting provider: comm engagement and comm submission events
   - Digital retailing tool: retail process events
5. Show me one table: store, vendor, missing events, and the store's GA4 property ID.

Don't guess. If you can't tell whether an event is firing, say so. Stop after the table and check with me before writing anything else.

For each store and vendor with missing events, draft an email to that vendor.

Each email should include:
- The store name and website
- The GA4 property ID
- Each missing event by its exact name, with a one-line description from the ASC spec
- A request to implement ASC GA4 Standard v1.2, including the event_owner parameter
- A request for a go-live date within 30 days

Keep each email under 150 words, friendly, and specific. Group them by vendor in one document so I can copy each one. Draft only. Do not send anything.

Using GA4 for the last 90 days, find every page view that landed on a "page not found" page. GA4 doesn't record HTTP status codes, so find them by page title ("Page Not Found", "404", "Oops") and by the error_code parameter on asc_pageview if the site sends it. Tell me which method found what.

For each broken URL, show:
- Sessions and users
- Where the traffic came from: source / medium, campaign, and referring site
- What the URL was probably for: a sold vehicle's VDP, an old model page, an expired special, an old service page, or something else

Sort by sessions. Group the results by who caused the broken link:
- Our website (old internal links, deleted pages)
- Our ads (Google, Microsoft, Meta, or TikTok ads pointing at a dead URL)
- A third party (a listing site, the OEM site, a directory, an email or text campaign)

Stop after the table and check with me.

For each broken URL, recommend the best live page to 301 redirect it to:
- A sold vehicle goes to the search results page for that same make, model, and condition
- An old model or specials page goes to the current one
- Anything else goes to the closest matching page, never just the homepage

Give me the redirect list as a two-column table (old URL, new URL) the website provider can load directly. Then draft emails:
- To our website provider, with the redirect list attached
- To whoever runs our ads, listing each ad or campaign still sending traffic to a dead URL
- To each third party linking to a dead page, with the correct URL

Draft only. Do not send anything.

Using GA4 for the last 30 days, pull sessions by source / medium, campaign, and landing page. Classify every landing page as one of: VDP (one vehicle), SRP (search results or inventory list), model page, specials, service or scheduler, finance or trade, homepage, 404, or other. Use the URL patterns on my site, and tell me the patterns you used.

Then compare each campaign's intent to where it actually landed:
- Vehicle listing ads (Google VLA, Microsoft, and Meta or TikTok inventory and catalog ads) should land on a VDP
- Model or trim search campaigns should land on a model page or a filtered SRP
- Service campaigns should land on a service page or the scheduler
- Brand campaigns can land on the homepage

List every campaign where more than [10]% of sessions landed on the wrong page type. Show sessions, engagement rate, and lead rate for the right landing vs the wrong one.

For each mismatch, tell me the most likely cause:
- A VLA or catalog ad landing on an SRP or a 404: the vehicle sold and the inventory feed or ad catalog didn't update, or the feed's link template is wrong
- Inventory ads landing on the homepage: the catalog's URL field is broken
- Search or service campaigns landing on the wrong page: the ad's final URL is wrong

Estimate what those wrong landings cost: sessions on the wrong page times the difference in lead rate. Then draft one email per responsible party: our website provider, our inventory feed provider, and whoever manages each ad account. Draft only.

Before comparing anything, normalize names: ignore ® and ™, ignore case, treat drive type words in the trim (AWD, FWD, 4WD, 4x4) as part of drivetrain, not trim, and group model variants that are the same vehicle. Show me any names you merged.

Using the market inventory database, find the models and trims that competitors in my market stock in real volume and I have little or none of. Only include makes my store sells. Only count a trim if at least two different competitors carry it, so one dealer's opinion doesn't count as a trend.

If the scraper has at least two weeks of history, also check how fast each trim sells in my market: how many days units lasted on competitors' sites before disappearing. Rank by volume (and speed, if available), and give me a buy list for the used car manager and an order list for new.

Before comparing anything, normalize names: ignore ® and ™, ignore case, treat drive type words in the trim (AWD, FWD, 4WD, 4x4) as part of drivetrain, not trim, and group model variants that are the same vehicle. Show me any names you merged.

Compare my pricing to competitors for every model and trim we both have, same condition and same model year. Don't compare raw prices: packages swing MSRP by $15,000 on the same trim. Compare discount off MSRP instead, and convert the difference to dollars per unit.

Leave out competitors that don't publish MSRP, or whose price always equals MSRP (they probably only show MSRP), and tell me which ones you left out.

Give me two lists:
1. Discounted less than the market. These are my first price conversations.
2. Discounted more than the market. I may be leaving money on the table.

For each, show store, stock numbers, my discount, the market discount, the number of competitor units compared, and dollars per unit. If the scraper has two weeks of history, add days on lot. Flag any gap over 15% as "check equipment before repricing."

Look at every competitor price cut in the price history. For each competitor, tell me:
- How many days a unit usually sits before the first cut
- How big the first cut usually is, in dollars and percent
- Which models they cut most

Then tell me, for this week, which of their units are coming up on their usual cut day, and whether I have the same trim. That's who is about to undercut me.

Find competitor units that have been on their lot more than 90 days. Group by dealer and model. For each group, tell me:
- If I have the same trim, whether I'm priced above or below them
- If I don't have it, whether it's a dealer trade opportunity
- Whether this is a model the whole market is stuck on, and I should slow down buying it

Before comparing anything, normalize names: ignore ® and ™, ignore case, treat drive type words in the trim (AWD, FWD, 4WD, 4x4) as part of drivetrain, not trim, and group model variants that are the same vehicle. Show me any names you merged.

Using the market inventory database, find the trims where I have at least three units and my discount off MSRP is at or above the market's. If the scraper has two weeks of history, also require that the market sells them quickly. Those are the units my ads should push this week.

Then find trims where I discount well below the market or the market is flooded. Those are the ones to pause or lower bids on until the price changes. Give me both lists by store.

Check the scraper runs for the last 7 days. List any dealer that errored, found zero vehicles, or dropped more than 30% in unit count from one day to the next.

Then check data quality by dealer: missing MSRP, missing price, price above MSRP, and fields holding the wrong thing (like a transmission in the drivetrain field). Tell me which dealers' data I shouldn't trust for pricing yet, and why.

Compare my active inventory in the market database against GA4 item page views (asc_item_pageview, matched on VIN or stock number) for the last 14 days. List every unit on my lot with zero or almost zero views. For each, suggest why: missing photos, price above market, not syndicating, or low demand for that trim.

Find my units with the most GA4 item page views in the last 14 days and no form, call, or chat leads tied to them. Compare each one's price to the market. Tell me which look like pricing problems and which look like merchandising problems.

Rank the makes, models, and trims shoppers view most on my website over the last 30 days. Put that next to how many units I have and how many my competitors have. Where is demand high and my supply low?

Find VDP URLs that got paid traffic in GA4 in the last 14 days for vehicles that are no longer active in the market database for my store. Those are ads, feeds, or catalogs still spending money on cars I already sold. List them by campaign and ad platform.

Read the OhMyDash dashboard open in my Chrome tab. List every metric on the page and exactly how it's calculated. Flag any metric whose name might not mean what a GM thinks it means, and ask me the five questions you need answered to interpret it correctly.

Take three numbers from the OhMyDash dashboard in my Chrome tab and pull the same numbers straight from the data underneath it — GA4 if the dashboard is on my live data, or the matching CSV in demo-data/ if it's on the demo set — for the same date range. Where they don't match, explain why.

For each KPI on the OhMyDash dashboard, describe a realistic way the number could improve while the store gets less profitable. Then check the data for signs that's happening now.