Meta Ads AI: Ask Your Ad Account Questions

Meta Ads Manager answers questions one filter at a time. The faster path is to ask across campaigns, ad sets and ads, as long as you know how Meta counts a purchase.

By Ed · Published October 25, 2026 · 8 min read

Twenty-five years starting companies, consulting, owning a multi-million-dollar P&L and leading business intelligence for large enterprises.

You can ask a Meta ad account questions in plain English and get reliable answers as long as the AI reads your actual campaign, ad set and ad rows, and you know that purchases are counted under your account's attribution setting. The questions that pay off compare cost and result across many ads at once, which is the part Ads Manager makes slow.

For the manual version, see the Facebook ads report guide.

Three levels, three kinds of decision

Meta structures an account as campaigns, ad sets and ads. Each level supports a different decision, and asking at the wrong level gives a confusing answer.

Campaign: where does the budget go, and what does each objective return?
Ad set: which audiences and placements are efficient?
Ad: which individual ads spent a lot and earned little?

A common mistake is judging an ad set by one campaign-level number. If a campaign has a good blended cost per purchase because one ad set carries it, the average hides a weak ad set that is eating budget.

How Meta counts, and why it matters

Three counting details change what a question means.

Attribution setting. Purchases are credited to ads within the window your account is set to. Changing the setting changes the history of reported purchases, and that is before any real change in sales.
Clicks. Meta reports several kinds. Link clicks are visits to your destination. "All clicks" includes reactions, shares and profile clicks. Cost per click and click-through rate mean different things depending on which you use. Link clicks are the one that corresponds to traffic.
Overlapping action types. Meta lists purchases under several overlapping action types in one response. Adding them double counts. A careful reader takes one, not the sum.

Questions worth asking

How much did we spend each month this year, and on which campaigns?
Which campaigns had the best ROAS last month, with their spend?
Which ad sets had the lowest cost per purchase in the last 30 days, excluding anything under a minimum spend?
Which ads spent the most last month, and what did they return?
How many leads did we get per campaign, and at what cost per lead?
Did cost per purchase rise because link click cost rose or because fewer clicks turned into purchases?

The last question matters because the fix differs. Rising click costs point at audience saturation or competition. A falling purchase rate per click points at the landing page, the offer or tracking. See what CTR, CPC and CPM tell you for how to separate these.

How Quiriz reads a Meta account

Quiriz has a Meta Ads connector, in beta. You sign in with Facebook, choose one ad account, and Quiriz reads 13 months of insights using the attribution setting from Ads Manager. It refreshes daily, is read-only and is included in paid plans. Datasets:

Campaigns: status, objective, buying type and budgets.
Campaign performance (daily) and Ad set performance (daily): impressions, link clicks, spend, purchases, purchase value and leads.
Ad performance (monthly): the same measures per ad, for the top 2,000 ads each month by spend.

Metrics are defined once: Meta Ads spend, impressions, clicks (link clicks), conversions (purchases), conversion value, leads, ROAS, CPC, CTR and CPA, with ratios computed from totals. Purchases come from one Meta action type, not a sum of overlapping ones. You can ask:

"Which Meta campaigns had the best ROAS last month?"
"Which ad sets had the lowest cost per purchase in the last 30 days?"
"Which Meta ads spent the most last month, and what did they return?"

Answers include the table, so you can see the spend behind each ratio.

An illustrative example: a campaign that looks fine

A campaign shows a cost per purchase of 28 against a target of 35 (example numbers). Looks healthy. Ask the same question at ad set level with a spend floor and you might find three ad sets: one at 18 on a small budget, one at 24 on the biggest budget, and one at 61 that took a third of the spend. The campaign average hides the 61. The fix is not to switch off the campaign but to move budget away from the weak ad set and see whether the blended figure improves over the next two weeks. Asking the question at three levels, campaign then ad set then ad, is the single most useful habit in Meta reporting.

One more precaution: ask for the spend alongside every ranking. A list of ad sets sorted by cost per purchase is only trustworthy when you can see that each one had enough spend and purchases behind it. If the answer you get back omits those columns, ask again and request them.

Limits

Numbers, not creative. You get ad names and results. Images, videos and copy are not read, so a question about which visual style works needs your own labeling in the ad names.
No reach or frequency. They cannot be added across days, so they are left out.
No placement, age, gender or country breakdowns.
The ad dataset is capped at the top 2,000 ads per month by spend, so its total can be below the campaign total.
Purchases are as Meta counts them. They will not match your store. Here is why.
Reconnecting. Facebook sign-ins expire after roughly 60 days and cannot be renewed silently, so expect to sign in again now and then.

Next, see a Meta Ads dashboard or compare platforms in Facebook Ads vs Google Ads ROAS. Paid media teams can start at Quiriz for paid media.

Ask your Meta ad account a question

Connect an ad account and ask which ad sets had the lowest cost per purchase in the last 30 days.

Try Quiriz free →

Frequently asked questions

Can AI analyze my Facebook and Instagram ads?

Yes, when it has the campaign, ad set and ad rows from your account. It can rank by cost per purchase, compare months, and surface ads that spent a lot and returned little. It cannot judge the creative itself unless it can see the images, so treat it as an analyst of numbers rather than a creative reviewer.

How does Meta count purchases?

Meta reports purchases according to the ad account's attribution setting, which defines how long after a click or other interaction a purchase is credited to the ad. The same actual orders can produce different purchase counts under different settings. Always know which one your account uses before comparing periods.

Why is my Meta ROAS different from my store revenue?

Meta credits purchases it believes the ads influenced, within its attribution window, and a store counts every order regardless of source. Other channels also claim the same orders. The two figures answer different questions, and the gap is worth tracking.

What is a good minimum spend before judging a Meta ad?

There is no universal number. A practical approach is to require spend of at least two to three times your target cost per purchase before ranking an ad, since an ad with one purchase on a tiny budget will look better than it is.

Why do I need to reconnect Meta periodically?

Facebook sign-in tokens last for a limited time, roughly 60 days, and cannot be silently renewed. When one expires, a connected tool has to ask you to sign in again. It is a Meta platform behavior, not an error.