Guides

Andromeda checklist for D2C ad accounts

Andromeda is the name of Meta's ad retrieval system. Before any auction runs, it cuts a very large pool of eligible ads down to a short list for each person. That means your account structure and the variety in your creative now count for more than clever audience splits.

This guide is a checklist to help you get that right. Each point shows you how to check it in your own Ads Manager export, so you are working from your own numbers and not guessing.

What Andromeda means for how you build campaigns

Your creative now does a lot of the targeting. The visual, the opening line and the offer tell the system who an ad is for. The more different ideas you give it, the more kinds of people it can match.

So the old habit of building many narrow ad sets with one or two ads each now works against you. Each small ad set gets too little data, and the system has too few different ads to show to different people. The checklist below fixes that.

• Use fewer, broader ad sets, and make sure each one gets a real share of the campaign's budget.
• Put several genuinely different creatives in every ad set.
• Do not split a campaign into many tiny ad sets.
• Do not run copies of the same creative in many places.

Pull the export you need

One file is enough for every check. Export at ad level so you can see each creative on its own.

A good first row looks like this: Campaign name, Ad set name, Ad name, Day, Amount spent (INR), Impressions, Reach, Link clicks, Results, Result type.

1. Open Ads Manager and pick the Ads tab. Ad level gives you the most to check.
2. Set the date range to the last 30 days.
3. Choose Breakdown, By time, Day.
4. Open the Columns menu. Make sure you have Delivery status and Cost per result. If you run video, add 3-second video plays and ThruPlays.
5. Click Export table data and save the file as .csv or .xlsx.
6. Open the file in Google Sheets or Excel.

Check 1: fewer, broader ad sets, not many tiny ones

Start with how each campaign's spend is split across its ad sets. Our audit's rule of thumb flags a campaign split into more than 5 ad sets that each take under 10% of its spend. This is our own house rule of thumb, not an industry standard, but it catches the pattern well.

For example, picture a made-up campaign with 9 ad sets. Two of them take 40% and 30% of its spend. The other seven share the last 30%, so each one takes under 10%. That campaign fails the check.

What to do: merge the small ad sets that reach similar people into one or two broader ad sets. Move in the ads that have results. Let the merged ad set build its data in one place. Make the change in one go, not through many small edits, because big edits can send an ad set back into learning.

1. Build a pivot table with Campaign name and then Ad set name as rows, and the sum of Amount spent as values.
2. Divide each ad set's spend by its campaign total to get its share.
3. Count the ad sets in each campaign that take under 10% of its spend.
4. If more than 5 do, that campaign is too fragmented.

Check 2: at least three genuinely different creatives per ad set

Our audit's rule of thumb flags any ad set with fewer than 3 live creatives. Live means the ad is actually delivering. So count using Delivery status, not the number of ads you once created.

Genuinely different means a different idea, not a new colour on the same image. Think in angles: a problem-led video, a review-style video, a product demo, an offer-led static. Three versions of one video with new captions are three ads, but treat them as one idea.

For example, an ad set has 4 active ads, and three of them are cuts of the same unboxing video. It really has only two ideas. Add one or two new concepts before you add more variations.

1. Filter Delivery status to active ads.
2. Make a pivot table with Ad set name as rows and a distinct count of Ad name as values. Each ad appears once per day in the file, so count distinct names, not rows.
3. Mark any ad set with fewer than 3.
4. For the rest, read the ad names and check that they are different ideas.

• When you add new video creatives, check their hook rate (3-second video plays divided by Impressions) and hold rate (ThruPlays divided by 3-second video plays).
• Our audit's rules of thumb flag a hook rate under 25% and a hold rate under 15%. They only judge video ads with at least 1,000 impressions.

Check 3: one creative, one home

When copies of the same creative run in several ad sets or campaigns, your own ads compete with each other. The data that one creative should be gathering also gets split. Our audit flags the same creative (copies) running in more than one place.

For example, a made-up video called Launch Video A runs in four ad sets. One copy has taken most of the spend and has the lowest Cost per result. The other three have spent little and brought few results. Keep the strongest copy, pause the rest, and let that one ad set carry the video.

1. Make a pivot table with Ad name as rows and a distinct count of Ad set name as values.
2. Any ad name that appears in more than one ad set is likely a copy.
3. Look for names ending in the word Copy, which Ads Manager adds when you duplicate an ad.
4. Compare Amount spent and Cost per result for each copy before you decide which one to keep.

• Renamed copies will not show up by name. If two ads look alike, open their previews and compare them.

Put the fixes in order

Fix structure first, then creative, then copies. Merging ad sets changes where your creatives live, so doing it first saves you work later. Change one campaign at a time and give it a full week of data before you judge it.

If you would rather not build pivot tables, the free Ads Audit at the free Ads Audit reads the same export. These structure checks are part of its eight checks. In about a minute you get a score out of 100 and your three biggest issues, and it never needs a login to your ad account. Your file is stored encrypted, never used to train AI models and deleted 30 days after upload.

• No campaign has more than 5 ad sets that each take under 10% of its spend.
• Every live ad set has 3 or more active creatives that are different ideas.
• Each creative runs in one place only.
• Video ads have been checked for hook rate and hold rate.

Questions

Does Andromeda mean I should run only one ad set?

No. The aim is fewer, broader ad sets, not just one. Separate ad sets still make sense when the budget, optimisation event or country is different. What matters is that each ad set gets a meaningful share of spend.

How many creatives should each ad set have?

Our audit's rule of thumb flags any ad set with fewer than 3 live creatives. After that, add new ideas when your current ones start to tire. Do not just pile on small variations.

Will merging ad sets reset learning?

Big edits can send an ad set back into learning. Do the merge in one go, then judge it after a full week of data, not day by day.

Can I run these checks without the video columns?

Yes. The structure checks only need Campaign name, Ad set name, Ad name, Delivery status and Amount spent. You only need the video columns for hook rate and hold rate.

Written by Monastic Media, Surat. The thresholds above are our audit's rules of thumb, not industry standards.