Data to AI™ / Gemini/Amazon Ads
Join ad spend with product costs, keep attribution windows straight in code and check report freshness before a job runs, using Amazon Ads data from Dataddo in the terminal or VS Code.
Better answers, fewer tokens.
Gemini CLI
Which Sponsored Products campaigns had the highest ACOS on 14-day sales last month?
Generic - Kitchen ran the highest ACOS at 42.0%, more than four times Brand - Exact at 10.0%. Competitor ASINs came second at 33.0%.
| Campaign Name | Spend | Sales 14d | Acos Clicks 14d |
|---|---|---|---|
| Generic - Kitchen Sponsored Products: Campaign | 4,120 | 9,810 | 42.0% |
| Competitor ASINs Sponsored Products: Campaign | 2,460 | 7,450 | 33.0% |
| Brand - Exact Sponsored Products: Campaign | 1,350 | 13,500 | 10.0% |
Spend and Sales 14d use the same 14-day attribution window; Sales 7d is never mixed in.
Amazon Ads reports cover Sponsored Products, Sponsored Brands, Sponsored Display and Amazon DSP. Each source reads one report type for the profile you pick, daily or as a summary for the date range.
Pull Advertised Asin, Spend and Sales 14d from Sponsored Products: Advertised Product for last month, join them with data/cogs.csv and write profit after ad spend per ASIN to reports/asin-margin.csv.
A retail media analyst combines Amazon ad data with product costs that never leave the repository. Gemini CLI reads the ad side from Dataddo and writes the join, so the margin view is rebuilt in seconds each month.
src/amazon/acos.ts divides Spend by Sales 7d. Describe the Sponsored Products: Campaign entity and switch the formula to Sales 14d so it matches Acos Clicks 14d.
In Gemini Code Assist agent mode, a developer aligns the dashboard formula with the window Amazon reports. Mixed windows are caught in the pull request instead of in a quarterly review where ACOS looked better than it was.
Run data_status on the DSP: Campaign flow. If last week is complete, write Total Cost, Impressions and eCPM per Order Name to reports/dsp-weekly.md.
Amazon builds reports asynchronously, so the job first confirms the data is in. The programmatic team gets a complete file or a clear note on what is still missing, never half a week presented as a full one.
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 | |
|---|---|---|---|
| Report types in one project | Each Amazon report type means a separate download, and scripts juggle files from Sponsored Products, Sponsored Brands and DSP. | Every report type is its own Dataddo source, and a single dataddo entry in settings.json lets Gemini reach all of them that are attached. | One server entry, all reports |
| Attribution windows | A notebook divides Spend by whichever sales column comes first, and 7 day and 14 day figures end up in one ratio. | Gemini reads field descriptions from the semantic layer, so code written with it keeps Sales 14d with Acos Clicks 14d and Sales 7d with Purchases 7d. | Consistent ACOS and ROAS |
| Late report data | A scheduled script runs on schedule even when Amazon has not finished yesterday's report, and nobody notices the gap. | Gemini can call data_status to check when each Amazon Ads flow last received data before writing any file. | No silent partial weeks |
| Token consumption | Gemini spends tokens exploring raw Amazon Ads 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 Amazon Ads 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
|
|---|---|---|
| Sponsored Products: Campaign |
|
Export Campaign Name, Spend, Sales 14d and Roas Clicks 14d for last month to a CSV for our bid notebook. |
| Sponsored Products: Search Term |
|
Write every Search Term with Cost above 20 and no Purchases 7d to a negatives CSV, with its Match Type. |
| Sponsored Products: Advertised Product |
|
Return Spend and Sales 14d per Advertised Asin as JSON for our product margin script. |
| Sponsored Products: Campaign Placement |
|
Export Top Of Search Impression Share per Campaign Name and Placement Classification for the last 8 weeks to a CSV. |
| Sponsored Brands: Campaign |
|
Write weekly Cost, Sales and New To Brand Sales per Campaign Name to a CSV for the brand notebook. |
| Sponsored Display: Campaign |
|
Which Sponsored Display campaigns spent Cost last week with zero Purchases? |
| DSP: Campaign |
|
Export Order Name, Impressions and eCPM per day to a CSV so a script can flag eCPM spikes. |
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 one Amazon Ads source per report type in Dataddo, such as Sponsored Products: Campaign and DSP: Campaign, and attach them to a Google Gemini destination or AI Model. Add dataddo with httpUrl https://headless.dataddo.com/mcp-data under mcpServers in ~/.gemini/settings.json, restart Gemini CLI and run /mcp auth dataddo.
Yes. Amazon serves each report type from its own API endpoint, so Dataddo extracts each into a separate source. The single dataddo server in settings.json reaches every one of them attached to an AI destination or AI Model.
Amazon generates reports asynchronously and busy API traffic can delay them, so new rows sometimes arrive after the scheduled run. Ask Gemini to run data_status first and it reports the freshness of each flow.
The one you name in the prompt. Fields carry their window, such as Sales 7d or Acos Clicks 14d, so state it and keep cost, sales and ACOS on the same window throughout the script.
No. Gemini only reads report data Dataddo has extracted, through the semantic layer, and nothing is sent back to Amazon. Bids, budgets and negatives are still changed in the Amazon Ads console.
Gemini CLI opens the Dataddo sign-in page in a browser on the same machine, so it needs one there. You sign in with your Dataddo email and password, with MFA if prompted; Google and Microsoft SSO are not offered.
Yes. Each Dataddo source reads one profile, which is one account in one marketplace. Create a source per profile and report type, and load older periods once with a full data re-sync.
