Data to AI™ / Gemini/Google Merchant Center
Give the engineers who build your product feed the current item-level issues, prices and identifiers in the terminal or VS Code, next to the code that produces the feed.
Better answers, fewer tokens.
Gemini CLI
Which feed issues keep the most of our products out of Shopping right now?
Invalid GTINs affect 412 of 5,020 products (8.2%), more than price mismatches and small images combined (283).
| Issue | Item Level Issue Attribute Name | Products | Share of catalog |
|---|---|---|---|
| Invalid GTIN Product statuses | gtin | 412 | 8.2% |
| Price mismatch Product statuses | price | 186 | 3.7% |
| Image too small Product statuses | image_link | 97 | 1.9% |
Counted from the latest Product statuses run, which pulls every current product; the dataset has no date range.
Google Merchant Center holds the product listings that appear on Google Shopping. Dataddo extracts two datasets: Products, with feed attributes such as price and availability, and Product statuses, with approval status and item-level issues.
Before I merge this feed generator change, count products per Item Level Issue Code and save the result to checks/gmc-issues-before.csv.
An ecommerce engineer records the baseline in Gemini CLI, ships the change, and after the next scheduled extraction writes a second file. Diffing the two shows whether the release cleared issues or created new ones.
Compare Price Currency per Target Country in Products with the currencies in config/markets.json and list every Offer ID that does not match.
A developer working on market settings asks Gemini Code Assist in agent mode. It reads the config file from the open workspace and the offers from Dataddo, then points at the markets whose prices would show in the wrong currency.
Take 20 Item Group ID values with their Offer ID, Title, Condition and Availability and turn them into a JSON fixture for the feed builder's unit tests.
Instead of invented sample rows, the feed builder's tests run on variant groups shaped like the real catalog, so edge cases such as mixed Availability inside one group are covered.
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 | |
|---|---|---|---|
| Feed health in code review | Engineers open Merchant Center diagnostics in the browser and paste screenshots into the pull request, which reviewers cannot search or compare. | Gemini CLI writes issue counts per Item Level Issue Code into a CSV on the branch, so the numbers appear in the diff next to the code change. | Feed impact visible in the PR |
| Knowing when the catalog was read | A downloaded export gives no clue whether it reflects the feed before or after the last upload, and the next person guesses. | Neither dataset uses a date range; each scheduled run reads every current product. Turn on the optional Dataddo Extraction Timestamp column and Gemini can write it into each file. | Snapshot time on every row |
| Control over tool calls | A home-made script with broad access runs unattended, and nobody sees which product data it pulled or when. | Gemini CLI asks you to confirm each Dataddo tool call before it runs, and you choose Always allow only for the tools you trust. | Every lookup approved |
| Token consumption | Gemini spends tokens exploring raw Google Merchant Center 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 Merchant Center 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
|
|---|---|---|
| Product statuses |
|
Which Product ID values switched to a disapproved Status since the snapshot CSV we committed last week? |
| Product statuses: item-level issues |
|
Which Item Level Issue Description texts point at attributes our feed export script in this repo generates? |
| Products: catalog |
|
Which Item Group ID values contain variants with a different Brand, which our feed builder should never produce? |
| Products: price and availability |
|
Does every Price Currency in the feed match the currency our storefront config sets for that market? |
| Products: identifiers and targeting |
|
Which Target Country and Content Language pairs in the feed are missing from the locale list in our storefront repo? |
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 →Create a Google Merchant Center source in Dataddo with the Products and Product statuses datasets and attach it to a Google Gemini destination or AI Model. In ~/.gemini/settings.json, add a dataddo entry under mcpServers with httpUrl https://headless.dataddo.com/mcp-data, restart Gemini CLI and run /mcp auth dataddo to sign in.
Not in this setup. Dataddo extracts Products and Product statuses on your schedule and serves them through the Data Access MCP server. Gemini CLI signs in there with your Dataddo account, not with your Merchant Center login.
After the next scheduled run of the Merchant Center source in Dataddo. Each run pulls all products and statuses currently in Merchant Center, so a daily schedule means the fix shows up in the following day's data.
Yes. Product statuses carries Item Level Issue Code, Item Level Issue Description, Item Level Issue Attribute Name and Item Level Issue Servability, so Gemini can say which attribute is wrong and whether it blocks the listing.
Yes. Put the dataddo entry in .gemini/settings.json inside that project instead of ~/.gemini/settings.json. Gemini Code Assist in VS Code reads the same file; in IntelliJ it uses its own mcp.json.
Sign-in opens the Dataddo page in a web browser, so Gemini CLI needs a browser on the same machine. You sign in with your Dataddo email and password, with MFA if prompted; Google and Microsoft SSO are not offered here.
Every flow or Context attached to an AI destination or AI Model in the Dataddo account you signed in with, not only those attached to Google Gemini.
No. Access is read-only through the semantic layer. Gemini requests entities and fields by name, Dataddo never accepts SQL from the model, and nothing is written back to your Merchant Center account.
