Data to AI™ / Gemini/Google Ad Manager
Ad ops engineers and publisher developers pull network reporting into the terminal or editor, check trafficking against code and commit the output.
Better answers, fewer tokens.
Gemini CLI
Which ad units earned the most last week, and where did inventory go unfilled?
video_preroll earned the most, 6,840 from 380,000 impressions, an eCPM of 18.00. article_sidebar served the most impressions but left 412,000 unfilled.
| Ad Unit Name | Total Impressions | Total Line Item Level All Revenue | Total Inventory Level Unfilled Impressions |
|---|---|---|---|
| video_preroll Google Ad Manager | 380,000 | 6,840 | 21,000 |
| article_sidebar Google Ad Manager | 2,310,000 | 5,544 | 412,000 |
| homepage_leaderboard Google Ad Manager | 1,240,000 | 4,960 | 86,000 |
Last week is re-read on every scheduled run, so its rows are replaced, not added twice.
Google Ad Manager is Google's ad serving platform for publishers and advertisers. Dataddo extracts one flexible report dataset per network: choose from 101 standard metrics, 100 dimensions and line item attributes.
Through dataddo, find line items whose Line Item Start Date Time is after their Line Item End Date Time or whose Line Item Cost Type is not CPM, and save them to checks/line-items.csv.
An ad operations engineer runs this in the terminal each afternoon before new campaigns start and attaches the file to the trafficking ticket.
Read src/ads/slots.ts, then query Ad Unit Name with Total Impressions for the last 7 days and tell me which slots in the file never served and which ad units are not in the file.
A front-end developer uses Gemini Code Assist agent mode in VS Code while refactoring templates, so no live slot is dropped by mistake.
Get Total Line Item Level All Revenue and Ad Exchange Estimated Revenue by Country Name for each of the last 12 months and write data/gam-revenue-by-country.csv.
A monetization analyst refreshes the forecasting notebook input in one command and commits the CSV beside it, so the run can be reproduced.
Data to AI™ is priced per seat, like the AI assistants your team already uses, with ~25% off on annual billing. Need tailored contracts or volume pricing? Talk to us about Enterprise.
Serve governed business data to Claude, ChatGPT and other AI tools over MCP.
$10.49/month
per seat
3 seats minimum
Flexibly deploy Dataddo in any cloud or hybrid environment.
Custom
We have a payment model that works for you
| Without Dataddo |
With Dataddo
|
Outcome for you | |
|---|---|---|---|
| Reports as files in a repository | Someone downloads a report from the Ad Manager interface and copies it into the project folder, where it quietly goes out of date. | Gemini CLI queries the scheduled Dataddo flow and writes the CSV or JSON straight into the working tree, so the file can be regenerated with the same prompt. | Reproducible report files |
| Control over lookups | A script with a stored API credential pulls whatever it was coded to pull, and nobody reviews the calls it makes against the network. | Gemini CLI asks you to confirm each Dataddo tool call, and you can pick Always allow only for the tools you trust, such as describing fields. | Each call approved by you |
| Trafficking mistakes | Swapped flight dates or the wrong cost type are found when a line item fails to deliver or bills incorrectly, usually days after launch. | Line item attributes sit in the same Ad Manager dataset as delivery, so a terminal check lists suspect line items before they go live. | Errors caught before launch |
| Token consumption | Gemini spends tokens exploring raw Google Ad Manager columns, sampling rows and retrying until it understands the data. | Field definitions and relations are supplied up front, so the model skips the exploration and goes straight to the answer. | Fewer tokens per answer |
| Answer quality | Prone to wrong joins, invented fields and numbers that are hard to check. | Answers grounded in defined fields. In the Dataddo benchmark, 78.5% of questions were answered correctly, against 65.5% with plain CSV files. | Better answers |
The Google Ad Manager datasets marketing teams use most, with their real field names. Gemini queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask Gemini
|
|---|---|---|
| Delivery |
|
Export Total Clicks and Total CTR by day for the last 30 days as a CSV I can load into a notebook. |
| Revenue |
|
Write daily Ad Exchange Estimated Revenue for the last 90 days to a CSV file in this repo. |
| Inventory and fill |
|
On which days in the last 30 did Total Code Served Count rise while Ad Exchange Coverage fell? |
| Viewability |
|
List the weeks this year when Total Active View Viewable Impressions Rate was below 70% as a table. |
| Orders and line items |
|
Output every Order Name with its Line Item Name entries as JSON so a script can check our naming convention. |
| Inventory dimensions |
|
Which Ad Unit Name values in the report are missing from the ad slot list in our site code? |
| Line item attributes |
|
Which line items have a Line Item Start Date Time later than their Line Item End Date Time? |
We asked 43 pre-registered questions of three real datasets - Google Search Console, HubSpot deals and Google Ads - and delivered the same rows to the same model in four ways. Only the data layer changed.
of questions answered correctly - the best of the four delivery methods tested
fewer confidently wrong answers - 14.0% of runs against 24.0% and 26.3%
more correct answers on messy CRM data - better on 8 questions, worse on none
Same model, prompts and rows in every condition; 759 graded runs, ground truth frozen before any run. On Google Search Console alone, documented CSV files did as well as Data to AI™ - the advantage comes from messier data such as HubSpot deals.
Read the benchmark →Set up a Google Ad Manager source in Dataddo with Line Item Name, Line Item Start Date Time and Total Impressions, and attach it to a Google Gemini destination or AI Model. Put an httpUrl entry for https://headless.dataddo.com/mcp-data under mcpServers in ~/.gemini/settings.json, restart Gemini CLI and run /mcp auth dataddo.
Not one this setup uses. Dataddo extracts your Ad Manager report into a flow on a schedule, and Gemini CLI reaches that flow over the Dataddo Data Access MCP server.
Add a dataddo server under mcpServers with only an httpUrl key set to https://headless.dataddo.com/mcp-data, with no headers or oauth block. The url key falls back to SSE, which Dataddo does not support.
No. The sign-in uses your Dataddo email and password, with MFA if prompted. Gemini CLI opens the page in a browser, so the machine running it needs one.
Yes, once you add the Dataddo server to the mcp.json of Gemini Code Assist in IntelliJ. In VS Code, agent mode already reads the same settings.json as Gemini CLI.
No. Gemini only reads extracted Ad Manager report data through the semantic layer by entity and field name. Dataddo never takes SQL from the model, and no change reaches your network.
The session probably expired or was revoked. Tokens live in ~/.gemini/mcp-oauth-tokens.json and are normally renewed on their own; run /mcp auth dataddo again to sign back in.
