How the AI Assistant treats your data

Every answer from the AI Assistant is built from the account data of the campaign group you select.

This article explains the rules behind the numbers: which reporting time and attribution settings apply, at which level an analysis runs, why low-volume items are left out, how missing Impression Share values are handled, and how Overall Impression Share is calculated.

If you only need a short answer, the AI Assistant FAQ covers the most common questions. This article is for when you want the exact rule.

Reporting time and attribution

The assistant reports every metric by install time. Apple Ads metrics such as impressions, taps, installs, and spend are reported on install time, and for MMP-sourced metrics such as revenue, ROAS, and ROI, the assistant's default attribution settings also use install time as the reporting time. Whether your question involves Apple Ads data or post-install data from your MMP, reporting is always based on install time.

For MMP (Mobile Measurement Partner) metrics, the following default settings apply. The assistant states them in every relevant answer.

Analysis levels

Every conversation is tied to the one campaign group you select at the top right. Within that group, analyses run at different levels depending on the question:

• If you name a specific campaign or keyword in your question, the assistant analyzes at that level instead of the campaign group.
• Campaign group level totals are not filtered by the minimum data thresholds described below.
• Campaign level analyses include only campaigns above the campaign spend threshold (more than $100 per 7 days).

The assistant cannot combine several campaign groups in a single chat. To compare groups, open a separate chat for each.

Minimum data thresholds

Ratio metrics such as CPA, CPT, TTR, CPM, ROAS, and ROI are unreliable at low volume. A keyword with 3 installs can show a $2 CPA by chance, and $50 revenue on $10 spend can be a single purchase. To avoid misleading conclusions, the assistant applies a minimum data threshold and leaves out anything below it, so the results you see are statistically meaningful. Excluded items are not shown in the results.

At keyword level, the thresholds are defined for a 7-day window and scale with the length of your time window:

For a custom window, each threshold scales by the number of days. For example, a 45-day window requires about 129 installs, 321 taps, or $321 spend, depending on the metric.

Campaigns follow a similar rule. Day parting and budget analyses include only campaigns with more than $100 spend per 7 days (for example, more than $400 for 28 days), because smaller campaigns cannot produce a reliable hourly pattern or budget signal.

How to get more results:

• Choose a longer time window so more items pass the threshold.
• Check the answer, which states how many items passed the threshold. Excluded items are not listed one by one.

The threshold does not apply everywhere. Campaign group level totals, campaign level spend change detection, keyword suggestions, and Impression Share drop questions are not filtered this way.

Missing or N/A Impression Share

N/A means Apple Ads returned no Impression Share data for that keyword on that day. It does not mean 0%. A keyword showing N/A is not necessarily invisible; there is simply no value to report.

How the assistant handles it:

• N/A days are left out of calculations instead of being counted as zero, so they do not pull averages down.
• If you ask about a keyword that has no Impression Share data at all, the assistant tells you so instead of guessing.
• Impression Share is reported per keyword and per storefront, so the same keyword can have data in one storefront and N/A in another.

What you can do:

• Do not read N/A as zero visibility.
• Try a longer time window, since a keyword may have data on some days but not others.
• Ask about another keyword or storefront if you need a value to compare.

Overall Impression Share

Impression Share shows how much of the available impressions your ads received. Apple Ads reports it only for individual keywords in a storefront, so there is no single number that tells you how visible your campaign group is as a whole. Overall Impression Share fills that gap. It is one score for your campaign group, built from the Impression Share of your highest-spending keywords.

How it is calculated:

• The last 28 complete days are used as the baseline.
• All keyword and storefront pairs in the campaign group are ranked by spend.
• The 25 highest-spending pairs are selected. To be selected, a pair must have valid data on at least 20 of the 28 days; pairs that do not meet this requirement are skipped.
• The daily market size is calculated for each pair as Impressions Received divided by Impression Share.
• The total impressions and total market estimates of all pairs are summed and combined into a single ratio: Aggregated IS = (Total Impressions Received / Total Market Estimate) × 100.

The result is rounded to the nearest whole number. Keywords in larger markets carry more weight in the score, so it reflects the actual market position of your campaign group.

How to read it:

• It is a snapshot of the last 28 complete days, not a live value.
• It reflects your biggest spenders. Small keywords outside the top 25 keyword and storefront pairs do not move it.
• Keywords in larger markets carry more weight, so a high score means you hold a strong position where most of the demand is.
• Like keyword-level Impression Share, it is based on the minimum value Apple Ads reports, so treat it as a conservative estimate.
• A change is expressed in percentage points. Going from 20% to 25% is +5 points, not +25%.

When to use it:

• Use Overall Impression Share to track your general market presence over time.
• Use keyword-level Impression Share to find where you are losing visibility.

Related articles

AI Assistant FAQ

What the AI Assistant can and can't answer

Need more help?

If you have any questions, feel free to reach out to your dedicated Customer Success Manager or contact us through live chat.