Data to AI™ / Claude/Apple Search Ads
Let app marketers ask Claude where installs come cheapest, which campaigns stopped serving and which keyword bids look out of line, without exporting reports from the Apple Search Ads dashboard.
Better answers, fewer tokens.
Claude
Which countries brought installs at the lowest cost last month?
Canada had the lowest cost per install at $1.84, on a third of the UK's spend. The United States took the most budget at $2.61 per install.
| Country Or Region | Local Spend Amount | Total Installs | Total Avg CPI Amount |
|---|---|---|---|
| CA Report Campaigns | $3,120 | 1,696 | $1.84 |
| GB Report Campaigns | $9,480 | 4,190 | $2.26 |
| US Report Campaigns | $21,700 | 8,314 | $2.61 |
Figures come from Report Campaigns with Group By set to Country Or Region and a monthly breakdown, requested from Apple in UTC.
Apple Search Ads promotes iOS apps in App Store search results. Dataddo extracts three datasets: Campaigns with budgets and serving status, Report Campaigns with spend, taps, installs and cost per install by day, country or device, and Targeting Keywords Find with the bids and match types of one campaign.
From Report Campaigns grouped by Country Or Region, rank the last 90 days by Total Avg CPI Amount and flag countries where Total Install Rate fell month over month.
Before the quarterly planning session, the user acquisition lead asks Claude which markets earned more budget and which ones got pricier, and brings a short ranked list instead of a dozen dashboard screenshots.
In Targeting Keywords Find, which keyword Text values have a Bid Amount more than twice the campaign median, and which Match Type are they on?
A search marketer reviews outlier bids in one pass. Claude groups the expensive keywords by Ad Group Id, so the cleanup happens ad group by ad group rather than keyword by keyword.
Compare Local Spend Amount this month in Report Campaigns with Budget Amount and Daily Budget Amount in Campaigns. Which campaigns will run out early?
Claude drafts the pacing note for the marketing director, naming the campaigns to slow down and the ones with room to grow, with the figures it used listed underneath.
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 | |
|---|---|---|---|
| Country comparison | Someone downloads a report per country view from the Apple Search Ads dashboard and pastes the rows into one sheet before anyone can compare cost per install. | Report Campaigns can be grouped by Country Or Region at source creation, so Claude reads one table with spend, installs and CPI per market. | One table per question |
| Keyword review | Bids are checked by opening each ad group in the dashboard, and outliers in rarely visited ad groups stay unnoticed. | Targeting Keywords Find lists every keyword of a campaign with Text, Match Type and Bid Amount, and Claude sorts them in seconds. | Full bid list in one view |
| Campaigns that stop serving | A campaign stops serving after a budget or setting issue, and the team notices only when installs dip in the next weekly report. | Campaigns carries Serving Status and Serving State Reasons, so a quick question to Claude shows which enabled campaigns are not running and why. | Serving issues spotted early |
| Analysis in Claude Code | An analyst copies exported numbers into a notebook and redoes the column cleanup every time the export layout changes. | Claude Code queries the same datasets through the dataddo server while you write scripts, so field names such as Local Spend Amount come straight from the semantic layer. | Reusable install analysis |
| Token consumption | Claude spends tokens exploring raw Apple Search 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 Apple Search Ads datasets marketing teams use most, with their real field names. Claude queries them by name through the Dataddo semantic layer.
| Dataset | Metrics and attributes |
Ask Claude
|
|---|---|---|
| Report Campaigns |
|
Which Country Or Region had the lowest Total Avg CPI Amount last month, and how much Local Spend Amount went there? |
| Campaigns |
|
Which campaigns are enabled but have a Serving Status that is not running, and what Serving State Reasons does Apple give? |
| Targeting Keywords Find |
|
Which keywords by Text have the highest Bid Amount in this campaign, and are they on exact or broad Match Type? |
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 an Apple Search Ads source in Dataddo with Report Campaigns and Campaigns, then attach it to a Claude destination or AI Model. In Claude, open Customize > Connectors, find Dataddo Data Access in the connectors directory, click Connect and sign in to Dataddo. Claude Code signed in with your Claude account gets it too.
No. Claude reads Apple Search Ads data that Dataddo has already extracted, through a read-only semantic layer. Dataddo never accepts SQL from the model and nothing is written back to Apple, so bids, budgets and campaigns stay as they are.
Not from this connector. Report Campaigns holds performance at campaign level, split by date and optionally by Country Or Region or Device Class. Targeting Keywords Find lists keywords with bids, match types and status, but no taps or installs.
As far as you extract. Report Campaigns reads a relative date window that moves forward each run and replaces that window's data. To give Claude older months, run a full data re-sync with a wider date range.
No. Report Campaigns covers spend, impressions, taps, installs, new downloads, redownloads and cost per install. It has no revenue field, so return on ad spend needs revenue from another source in Dataddo.
On Pro and Max you add Dataddo Data Access from the connectors directory yourself. On Team and Enterprise an Owner adds it once for the organization, and each member signs in with their own Dataddo account, so they only reach data that account can see.
Yes, by not extracting them. When you create the Campaigns source, untick Client Name, Buyer Name, Buyer Email and Buyer Contact Email in the attribute list, and those columns never reach Dataddo or Claude.
Yes, if the data is extracted for all of them. Dataddo supports multi-account extraction for Apple Search Ads, enabled on request by the Solutions team, so Claude can compare cost per install across accounts in one question.
