Ask your ad data. Get the answer, with the number.

An AI analyst that brings GA4, Google Ads, Meta Ads and Microsoft Clarity into one chat. It reads performance against your CPA and ROAS targets and tells you what to do, without changing anything in your accounts.

Analist
GA4Google AdsMetaClarity
Which channel was over target in the last 28 days?

Only Meta prospecting: CPA $538 against a $450 target, 14% worse than the previous 28 days. The most efficient channel is Google Ads brand search at $312.

CPA, last 28 days · line = target

Google Ads · Brand
$312
Meta · Retargeting
$356
Google Ads · Generic
$421
Meta · Prospecting
$538

Two prospecting ad sets have a frequency above 4. I'd move part of that budget to retargeting.

Ask a question…

Sample screen · figures are illustrative

Connects to

Features

Brings the number and tells you what it means

Your team types a question; the analyst pulls the data from the source, compares it with the previous period and answers briefly. With a table or a chart when it helps.

Every number from the source

Numbers are fetched from the platform when you ask, never recalled from memory. Every answer states its date range.

Judged against your targets

Write your CPA and ROAS targets per channel once, and performance is judged against them. If no target is set, it doesn't invent one.

As long as it needs to be

One sentence for a one-number question, a chart for a trend, a table for a campaign comparison.

How did daily CPA move over the last 14 days?

Google Ads is trending down, from $468 to $388; the last 7 days averaged 11% better than the week before. Meta is flat at around $400.

Daily CPA ($)

380420460Oct 1Oct 8Oct 14
Google AdsMeta

Notices when the numbers don't add up

Meta, Google Ads and GA4 each count conversions their own way. The analyst never adds them up, and it double-checks a surprising finding with a different cut before telling you.

No double counting

Decisions within a platform use that platform's data; cross-channel comparisons use GA4.

Surprises get a second look

Before saying "this channel made no sales", it looks again with a different cut, and tells you when the cause may be a tracking issue.

Diagnosis with behaviour data

Why isn't a landing page converting? Scroll depth, rage clicks and errors come from Clarity.

Did Meta really make zero sales last week?

No. It shows 0 by session source, but by first user source 23 sales came from Meta. This looks like a tracking issue, not a channel problem: the redirect on the payment page drops the session's source.

CutMeta sales
GA4 · session source0
GA4 · first user source23
Meta · 7-day click27
Soon

Reads organic search too

With Search Console connected, the analyst looks at paid channels and organic search together: which queries and pages are gaining or losing traffic, and which pages are missing from the index.

Query and page performance

Compares clicks, impressions, CTR and average position by query or page against the previous period.

Index status

Checks page by page whether your key pages are indexed and what Google saw on its last crawl. It also finds errors in your sitemaps.

Paid and organic side by side

When organic traffic drops, is paid making up for it, or are both falling? It puts both sources side by side for the same dates.

Why did organic clicks drop last month?

Organic clicks fell 18%. Most of the loss comes from two pages: /campaigns dropped out of the index, and /guides/getting-started slid from 4th to 8th place.

PageClicksPosition
/campaigns−412Not indexed
/guides/getting-started−1583.9 → 8.2
/pricing+365.0 → 4.6

On its last crawl Google saw a noindex tag on /campaigns. Remove the tag and request indexing in URL Inspection.

Reads your site's code as well

The answer to a tracking problem is often in the code. The analyst reads your site's code with read-only access and cites everything it finds by file and line.

Where does the event fire?

Which page sends the purchase event, under which condition and with which parameters: it's found in the code.

Is the click ID getting lost?

gclid and fbclid identify the ad a visitor came from. The analyst traces step by step whether they make it all the way to the sale.

Code versus data

The code shows what should happen. The analyst compares it with event counts in GA4 and tells you whether the two match.

Where does the purchase event fire?

On the payment confirmation page, once the booking reference arrives. Without a reference the event is never sent.

web/src/app/checkout/success/page.tsx:48
if (!booking?.reference) return;
gtag('event', 'purchase', {
  transaction_id: booking.reference,
  value: booking.total,
  currency: booking.currency,
});

Code read: web@main · 4f2c9e1 · Oct 12

Never changes anything in your accounts

The analyst recommends; your team applies. This isn't just an instruction to the AI, it's built into the software: tools that could change your campaigns are never given to the analyst.

No access to tools that make changes

Tools that create campaigns, update budgets or launch ads are disabled. If a platform adds a new tool, it isn't enabled automatically.

No more access than it needs

It connects to GA4 with read-only access. On Meta it only uses read tools that are explicitly allowed.

Gives you the steps

Ask it to "pause this ad set" and it tells you it can't, then writes out the steps.

Pause the "Summer – Broad" ad set.

I can't do that; I don't have permission to change your accounts.

Here's how to pause it:

  1. In Ads Manager, open the Ad sets tab.
  2. Turn off the toggle on the "Summer – Broad" row.
  3. To move its budget to "Summer – Retargeting", edit the campaign budget.

How it works

Setup happens once; after that, your team just asks.

1. Connect your accounts

Connect GA4, Google Ads, Meta Ads and Clarity with read-only access. Your site's code can be added too. If a source isn't connected, the analyst says so and keeps working with the rest.

2. Write down your targets

Write your CPA, ROAS and budget targets by channel and campaign type in a plain text file. You can add tracking quirks your team knows about as data notes.

3. Ask

Ask "How many visitors did we get yesterday?" or "Why did traffic drop, and what should we do?" The analyst answers in the language you ask in.

4 sources

Reach GA4, Google Ads, Meta Ads and Clarity data from a single chat.

Read-only

It has no access to any tool that could change your accounts.

Any language

It answers in the language you ask in. Metric abbreviations stay as they are.

Built for your team

Not one person's tool, but the whole marketing team's shared analyst.

Chat history

Every conversation is saved. When you edit a question or regenerate an answer, the old answer stays and the new one is added alongside it.

Sharing with the team

Share a conversation by link; teammates can read it but not change it. You can turn sharing off at any time.

File attachments

Attach a screenshot, PDF, Word, Excel or CSV file to a question. Figures in the file are checked against platform data where possible.

Model choice

Pick a fast or a deeper-thinking Claude model for each question. Simple questions are answered quickly; complex analyses get more time.

Sign in with your company email

Only company email addresses can sign in. Everyone sees only their own conversations.

Feedback

Rate any answer with 👍 or 👎 and add a comment. Ratings are recorded to track answer quality.

Frequently asked questions

Have another question?

Get in touch

Try it with your own data

We'll connect your accounts with read-only access and ask the questions your team asks every week, together.

Request a demo