meta andromeda creative diversification: how many creatives you actually need, by budget

most accounts do not need more ads. they need more concepts. here is how to count a concept, build a matrix that produces them, test without resetting learning, and the weekly production plan we would run at ₹50,000, ₹1.5 lakh and ₹5 lakh a month.

our meta andromeda update guide sets out the account structure and a rule-of-thumb table of concepts by budget, within our meta ads strategy for 2026. this post goes one level down on meta andromeda creative diversification: how to count a concept, produce enough of them, test them, and read the fatigue and similarity signals meta now shows.

• meta caps an ad set at 50 ads, per its marketing api ad reference.
• meta publishes no concept count for lead campaigns. ours is a working rule, matching the parent guide.
• thin ad sets cost more, meta now says. as reported by ppc land, meta said ad sets under about 50 conversions a week ran about 28% dearer per purchase. it gave no method.

meta andromeda creative diversification: what counts as a concept, and what is only a variation?

a concept is an ad that would appeal to a different person for a different reason. a variation is the same idea with a new hook line, headline, crop, length or colour. under meta andromeda creative diversification only concepts widen reach, so plan in concepts and use variations to keep a working concept fresh.

meta's performance marketing page lists creative diversification as a best practice. its guidance, as quoted by mockingbird, warns that ads which are too similar may not be recognised as different and may reach the same people. andromeda reads ad content at scale (meta engineering), so near-copies add cost, not reach.

our test: would it pull a different person, or the same person for a different reason, and would its first frame look different in a thumbnail grid? if not, it is a variation.

marut drones shows the concept side. creative and forms were state-specific across 13 states, so language and local context changed, not colour. ₹25 lakh of spend produced 23,420 leads and ₹8.7 crore in revenue, peaking at 36x roas (case study).

how do you build a concept matrix of persona, problem, format and hook?

a concept matrix lists your personas, their problems, the formats you can make and the hooks or languages you can use, then picks combinations far apart. three personas, three problems, three formats and two languages give 54 cells. you only need six to ten live, so the matrix is a menu, not a quota.

rows are persona and problem pairs, columns are formats, and each cell holds a hook. no two live concepts should share both persona and problem.

an illustrative example for a rooftop solar installer in maharashtra and gujarat, not client data:

language is the fourth axis: a marathi and a gujarati talking head with local faces are two concepts. take personas from sales calls; our meta ads by industry guide has starting personas per sector.

how many creatives per ad set does meta allow, and how many should you run?

meta allows up to 50 ads in one ad set, but budget sets the useful number. run four to ten concepts per ad set depending on spend, with one or two variations each, so a typical ad set holds 6 to 20 ads. more ads than your budget can fund means most never get a fair look.

meta's marketing api ad reference lists 50 non-deleted ads per ad set; we saw it through search summaries. a figure of at least 20 diversified ads applied to advantage+ shopping campaigns in november 2024, as reported by mockingbird; we could not open meta's original.

meta's october 2026 guidance, as reported by ppc land, gives a formula: weekly budget ÷ (cost per result × 50) is the most ad sets your budget supports. our working rule for concepts is three to five times your target cost per result for a new concept's first week before judging it.

if an ad set nears 30 ads, you are storing old ads, not running concepts.

meta creative testing 2026: one ad set or a separate testing campaign?

below ₹1.5 lakh a month, add new concepts straight into the live ad set in one weekly batch. above that, a small testing campaign or meta's built-in creative test earns its place. use a full a/b test only when you need a clean answer to one question, because it splits audience and budget.

meta's a/b testing help says to test one variable, keep budgets equal and not switch ad sets on and off by hand. its newer creative testing feature runs at the ad level inside an existing campaign; jon loomer reports two to five ads, up to ten in some accounts, using part of the budget (jon loomer).

meta's significant edit rules count adding creative, so we batch: one weekly reset at worst. move winners as the same post, so comments travel with them.

what production and refresh cadence do you need at each budget?

budget sets the cadence: about one new concept a week at ₹50,000 a month, around three a fortnight at ₹1.5 lakh, and five or so a fortnight at ₹5 lakh. give each new concept one or two variations, launch them in a weekly batch, and refresh a winner's hook every two to three weeks.

the parent guide's new concepts per fortnight are the base. we run them as a weekly loop: read on monday, brief tuesday, produce midweek, check compliance thursday, launch friday. our performance creative guide covers the loop and the hook bank behind it.

retire by concept, not by ad: if two fresh variations fade within a week, the idea is tired. keep two briefed concepts on the bench.

how do you read creative fatigue in meta ads before it costs you?

creative fatigue in meta ads shows as frequency rising, click-through rate falling and cost per result climbing for the same ad over several days. meta's delivery column may flag "creative limited" or "creative fatigue", but the label arrives late. watch the trend weekly and replace by concept, not colour.

meta's creative fatigue help defines fatigue as people seeing the same creative too many times. as reported by ppc land, "creative limited" shows when cost per result is above your past ads but under double, and "creative fatigue" at double or more.

october to december auctions in india are crowded; our facebook ads cost guide explains why. check lead quality before killing a concept: dearer leads that qualify better make a winner (sending quality stages back).

creative similarity in meta: what the "creative diversity" column tells you

creative similarity in meta is now visible through a "creative diversity" column that rates each ad set high, medium or low. as reported, it compares the images and video thumbnails in the ad set. a low rating means your ads look alike at first glance. treat it as a prompt to check, since meta calls it estimated.

as reported by common thread collective and mockingbird, it arrived around late august 2026, sits under columns then customize columns, and is labelled estimated and in development. meta has not published how it scores, and we found no meta source for a separate "too similar" warning, a public threshold, or the "entity id" clustering some practitioners describe.

twenty ads with the same product shot, opening, spokesperson and message offer little variety in meta's eyes (search engine land). a low rating usually traces to:

• same first frame. different scripts look alike if they open on the same shot.
• same person everywhere. add a customer, an engineer or a dealer.
• text-only edits. a new headline on the same visual groups with the original, agencies report.

a high rating does not prove your messages differ.

does advantage+ creative count towards your concept total?

no. advantage+ creative enhancements make versions of one ad: text variations, visual touch-ups, music, animation and static-to-video. they help a concept fit more placements and people, but they do not change the persona or problem. count them as variations, check each toggle, and keep your own concept pipeline running.

meta's advantage+ creative page and creative enhancements help describe ai that adjusts versions of your images, videos and text. agencies disagree on which are on by default, so open each toggle before launch.

we keep touch-ups that leave the message intact, test generated text and music first, and switch text changes off where a word alters legal meaning, such as rera lines. generated images and voices still fall under asci rules; see our ai generated ads guide. this is general information, not legal advice.

illustrative example: production plans for ₹50,000, ₹1.5 lakh and ₹5 lakh a month

at ₹50,000 a month, run one ad set with four concepts and add one a week. at ₹1.5 lakh, run one ad set with six concepts, three new a fortnight and a small test line. at ₹5 lakh, split only on a hard limit, such as language, and run six to eight concepts in each ad set.

these are illustrative examples with made-up costs per lead, not client data. weekly budget is monthly spend × 12 ÷ 52.

₹50,000 a month: one ad set, four concepts

about ₹11,500 a week. at ₹250 a lead that is about 46 leads. meta's formula gives 0.92, so one ad set and no more. optimise for leads, not a deeper stage.

₹1.5 lakh a month: one ad set, six concepts, a small test line

about ₹34,600 a week. at ₹400 a lead that is about 86 leads, enough for one strong ad set. two prospecting ad sets would leave each near 43, under 50.

₹5 lakh a month: two language ad sets

about ₹1,15,000 a week, about ₹95,000 of it on prospecting. at ₹500 a lead, 95,000 ÷ (500 × 50) is about 3.8, so up to three ad sets fit. language is the hard limit: hindi and marathi ad sets on a 60:40 split get about 114 and 76 leads a week, both past 50.

five new concepts a fortnight sits inside the parent guide's four to six, made with local faces, not subtitles. retargeting needs its own objection-led concepts, and it can carry real weight: for one d2c client, ai retargeting delivered 4.5x roas. scale also does not require many campaigns: for a luxury retail client, one meta campaign drove ₹2.1 crore in sales.

who wrote this, how and why

who: puneetraj chavan, founder of rapchik media, with 8+ years running performance marketing and automation for 30+ brands across india and the us.

how: sources were checked on 10 november 2026. figures marked "as reported" come from search summaries or named trade press, as several meta pages could not be opened. platform-reported figures were not independently audited; budget plans are illustrative.

why: most andromeda advice says "add more creative" without saying what counts as one, and that gap is where budgets go to near-copies.

• meta: engineering: andromeda · performance marketing · advantage+ creative · marketing api ad reference
• meta business help: creative fatigue · a/b testing · last significant edit · creative enhancements
• trade press: ppc land on 50 conversions · ppc land on fatigue · search engine land · jon loomer · common thread collective · mockingbird

want a concept matrix and weekly production plan built for your account? see our services or talk to us.

found this useful? share it

written by puneetraj chavan

founder of rapchik media. 8+ years scaling 30+ brands across india and the us, from agri-drones to a us job portal. strategy and execution, same brain.

more about me →

want this on your brand?talk to puneetraj on whatsapp reply within 24h →