Meta Andromeda.Understand the shift.
Build better campaigns.
A practical guide to what changed—and how to run ads for online stores, service businesses, and WhatsApp.
What is Meta Andromeda?
Meta confirmedAndromeda improves the first stage of Meta’s ad recommendation system: finding potentially relevant ads for each person. Meta calls this retrieval. It publicly described the system in December 2024. [1]
Example: A shop assistant selects suitable shoes from a warehouse before the final choice. Andromeda helps with that first selection.
Meta introduced it to improve personalization and handle more ads and automated variations. You do not switch it on in Ads Manager. [1]
Automation existed before Andromeda; other delivery systems also keep evolving. [2] [3]
Best Practice: Choose a clear campaign goal and create ads that explain why someone should buy or enquire. Do not rebuild a profitable account just because an update has a name.
Old approach vs recommended approach for 2026
These are changes in working habits, not a list of techniques that stopped working. Keep an older approach if your results still support it.
| Old Approach | Recommended Approach for 2026 |
|---|---|
| Find the perfect interest combination | Look beyond interests: improve your offer and ads, and check customer quality. Test one clear question at a time. |
| Rely on one winning ad | Develop several reasons to buy |
| Create many similar ad sets | Create a separate ad set only when it has a clear purpose |
| Celebrate the cheapest lead | Measure qualified leads and paying customers |
| Change settings every day | Allow time for results before changing settings |
Meta confirmed Meta’s Performance 5 guidance includes simplification, automation, creative diversification, data quality, and results validation. These are broader principles—not all new Andromeda rules. [4]
Creative diversity: different ideas, not just different designs
Creative means an ad’s image or video, words, and message. A concept is the main idea behind it—for example, saving time or proving durability.
Industry interpretation Different concepts give delivery systems more useful options for different people. Changing only the colour or headline adds less variety than showing a different benefit or customer problem. [5]
Creative diversity example: a social media marketing course
One course, five customer needs. These groups overlap. They are creative angles, not five required ad sets.
| Customer need | Example opening | Proof to show |
|---|---|---|
| Freelancers Promote their services | “Learn to build ads for your freelance offer.” | Lesson demo: offer → ad → enquiry. |
| Students Learn a practical skill | “Build a sample campaign as you learn.” | A real practice project and beginner lesson. |
| Parents Evaluate training for an adult child | “See the curriculum, projects, and learning support.” | Instructor walkthrough, feedback process, and fees. |
| Business owners Understand their own advertising | “Learn to plan your ads and read the results.” | A business campaign and results review. |
| Aspiring owners / job holders Prepare to promote an online venture | “Learn how to reach customers before you launch.” | A realistic learning plan and promotion exercise. |
Freelancers
Promote their services
- Example opening
- “Learn to build ads for your freelance offer.”
- Proof to show
- Lesson demo: offer → ad → enquiry.
Students
Learn a practical skill
- Example opening
- “Build a sample campaign as you learn.”
- Proof to show
- A real practice project and beginner lesson.
Parents
Evaluate training for an adult child
- Example opening
- “See the curriculum, projects, and learning support.”
- Proof to show
- Instructor walkthrough, feedback process, and fees.
Business owners
Understand their own advertising
- Example opening
- “Learn to plan your ads and read the results.”
- Proof to show
- A business campaign and results review.
Aspiring owners / job holders
Prepare to promote an online venture
- Example opening
- “Learn how to reach customers before you launch.”
- Proof to show
- A realistic learning plan and promotion exercise.
Remember This: Change the customer problem, message, and proof—not just the poster colour.
Only advertise what your course delivers. Do not promise clients, jobs, or income.
How these ideas fit into a campaign
Best Practice: Start with 2–3 relevant concepts if budget is limited. Test the others later. Meta may show any concept to people beyond the group you imagined.
Choose your audience and campaign structure
Best Practice: Start with a broad audience: keep necessary limits such as location, without adding many interests. A Lahore service provider still needs the correct service area; an online store still needs delivery restrictions.
- Interests and lookalikes: Interests describe topics people may care about; lookalikes find people similar to a source audience, such as customers. Test them when you have enough budget to compare lead or customer quality.
- Suggestions vs controls: A suggestion guides Meta; a control limits who can see your ads. Some Advantage+ audience inputs are suggestions, so check which settings are firm limits in your setup. [6]
- Advantage+: A family of automation tools—not another name for Andromeda. Review automated creative changes and use suitable placements. [7]
Campaign → Ad Set → Ads
Campaign budgets let Meta allocate money across ad sets. Use ad-set budgets when a test or business requirement needs controlled spending.
Avoid splitting your audience and budget too much
Fragmentation means too many small audience or budget divisions. PKR 3,000/day across six ad sets leaves PKR 500 each. At PKR 1,500 per purchase, that is roughly one purchase every three days per ad set—slow evidence for decisions.
| Keep together when… | Separate when… |
|---|---|
| The offer, market, goal, and sales process are the same | Goals differ, or a business line needs its own budget |
| Several nearly identical interest groups | Profit margins, locations, or sales processes differ |
| Ad sets are copies made only to restart delivery | A fair test needs a separate setup |
Remember This: One campaign is a useful starting point—not a universal rule.
Campaign setups you can start with
These structures and PKR budgets are examples, not Meta requirements. Budget for your expected cost per purchase or qualified lead.
A. Ecommerce: website purchases
- Starting ads and budget: 3–5 concepts; PKR 5,000/day is an illustrative budget. Consider catalog ads with an accurate product feed.
- Use separate campaigns or ad sets when: Markets, margins, or protected product budgets differ.
B. Lead generation: forms or website
A qualified lead is an enquiry from someone who fits your service and has a realistic need, budget, and timeline. An Instant Form opens inside Facebook or Instagram.
- Starting ads and budget: 3–5 concepts; for example PKR 3,000/day. Start broad within your service area.
- Use separate campaigns or ad sets when: Different services, branches, destinations, or sales teams require it.
Meta confirmed Supported CRM systems can report lead outcomes through Conversions API, Meta’s connection for business results. This can help supported campaigns optimize for lead quality. [8]
C. WhatsApp: turn chats into customers
Meta confirmed Eligible accounts can optimize for WhatsApp leads or purchases. Meta documents Leads for lead optimization and Sales or Engagement for purchase optimization. Follow the supported setup; a conversation goal is a starting option when a suitable deeper goal is unavailable. [9]
- Starting ads and budget: 3–4 concepts; for example PKR 2,000/day.
- Use separate campaigns or ad sets when: Numbers, branches, languages, or response teams need separate routing.
Common Mistake: Celebrating cheap “Hi” messages while nobody buys. Prepare qualification questions and use supported outcome reporting.
Adjust the structure to your budget
| Budget example | Campaign → Ad set → Ads | What to do |
|---|---|---|
| Small PKR 1,000/day | One priority campaign → usually one ad set → 2–3 different ad concepts | Broad within limits; test fewer ideas, usually skip separate retargeting, scale after real results. Split only for essential requirements. |
| Larger PKR 50,000/day | Main campaign + separate testing campaign + optional retargeting → ad sets with clear purposes → proven ads plus small batches of new ideas | Example: 37,500 main / 10,000 testing / 2,500 optional retargeting. Start with 5–8 main concepts and 2–4 test concepts if affordable. Scale profitable markets and offers. |
Small
PKR 1,000/day
- Campaign → Ad set → Ads
- One priority campaign → usually one ad set → 2–3 different ad concepts
- What to do
- Broad within limits; test fewer ideas, usually skip separate retargeting, scale after real results. Split only for essential requirements.
Larger
PKR 50,000/day
- Campaign → Ad set → Ads
- Main campaign + separate testing campaign + optional retargeting → ad sets with clear purposes → proven ads plus small batches of new ideas
- What to do
- Example: 37,500 main / 10,000 testing / 2,500 optional retargeting. Start with 5–8 main concepts and 2–4 test concepts if affordable. Scale profitable markets and offers.
At PKR 2,000 per purchase, PKR 7,000/week supports roughly 3–4 purchases. More ad sets cannot fix that arithmetic. Reallocate unused retargeting budget to a better opportunity.
Graphic designers, developers, and consultants
You do not need an ecommerce store to advertise a service. If a customer needs a discussion or quote, start by generating enquiries, then follow up to turn suitable enquiries into clients. If they can purchase a defined service online and purchases are tracked, a Sales campaign may fit.
Best Practice: Advertise a specific package or service. “We do everything digital” gives buyers little reason to enquire.
| Service | Clear offer example | Three distinct ad concepts |
|---|---|---|
| Graphic design | Monthly social-media design package for restaurants | Portfolio walkthrough · how ordering works · package scope and price |
| Web development | A mobile-friendly website that helps local service businesses collect enquiries | Live site demo · project case study · process and timeline |
| Consultancy | Initial consultation about a defined business problem | Explain a common mistake · show how you assess the problem · answer a buying objection |
Graphic design
- Clear offer example
- Monthly social-media design package for restaurants
- Three distinct ad concepts
- Portfolio walkthrough · how ordering works · package scope and price
Web development
- Clear offer example
- A mobile-friendly website that helps local service businesses collect enquiries
- Three distinct ad concepts
- Live site demo · project case study · process and timeline
Consultancy
- Clear offer example
- Initial consultation about a defined business problem
- Three distinct ad concepts
- Explain a common mistake · show how you assess the problem · answer a buying objection
Use truthful examples and client work you have permission to share. Do not invent testimonials or imply guaranteed financial results.
Turn enquiries into suitable clients
- No website yet? Use an Instant Form or a supported WhatsApp setup, with relevant proof in your ad and profile.
- Qualification questions: What business do you run? What do you need? What is your budget range? When do you want to start?
20 enquiries → 5 qualified leads → 1 paying client
Cost per enquiry: PKR 700 · Qualified lead: PKR 2,800 · Client: PKR 14,000
Compare client acquisition cost with profit after delivery. A PKR 50,000 project costing PKR 30,000 to deliver leaves PKR 20,000 before ads and overheads. Allow weeks for longer sales cycles.
Best Practice: Track enquiry → qualified → proposal → paid in your CRM or a sheet. Report supported outcomes to Meta.
Creative testing, audience testing, and scaling
Best Practice: First find a convincing message. Then test audience choices when you have enough budget. Increase spending only when the business results justify it. These are working recommendations, not a special Andromeda formula.
Meta’s broader guidance supports simpler accounts, varied creative, useful data, and results validation. [4]
A. Creative testing: find the message
Compare distinct concepts while keeping the course, price, form, audience, and performance goal consistent.
- Start small: Test 2–3 concepts on a limited budget. Keep promising ads running.
- Measure: Cost per suitable enquiry and paid enrolment. Clicks alone do not prove success.
- Low spend means unproven: Meta does not give every ad an equal test.
- Compare fairly: Use Meta’s A/B tools for a controlled test. Refine headlines after finding a promising concept.
B. Audience testing: compare setups
Compare broad targeting with one alternative: a customer lookalike or relevant interest suggestions.
- Hold other settings steady: Same geography, ads, offer, goal, and timing. Set a spending limit.
- Check expansion: Suggestions can reach beyond the named audience. This compares setups, not necessarily distinct groups of people. [6]
- Control spending: Use the experiment’s budget controls; free campaign allocation does not guarantee a fair comparison.
- Skip underfunded tests: If each version cannot generate useful results, invest in better creative first.
C. Scaling: grow without losing the economics
Example: With profitable enrolments, try increasing PKR 3,000/day to PKR 3,300–3,600. Review over the sales cycle. This 10–20% step is a recommendation, not a Meta rule or protection against renewed learning.
- Keep what works: Increase the successful campaign or ad-set budget. Avoid duplicate ad sets as a default.
- Watch new results: A strong historical average can hide expensive additional customers.
- Expand thoughtfully: Test fresh concepts, markets, or offers when your capacity allows.
- Pull back when needed: Allow for sales delays, then reduce spending if profit or lead quality deteriorates.
Profit check: PKR 15,000 extra spend ÷ 3 extra enrolments = PKR 5,000 each. Compare that with profit after teaching, support, fees, and refunds. Attribution alone does not prove the ads caused every sale.
Budget, learning phase, and review checklist
The learning phase is when delivery explores how to get your chosen result. Significant edits can restart learning. Around 50 optimization events per ad set in seven days is a longstanding industry reference; account displays and tests may vary. It is not a new Andromeda requirement. [10]
An optimization event is the result Meta is trying to get, such as a purchase or lead.
PKR 14,000 ÷ PKR 1,000 per lead ≈ 14 leads
More ad sets do not create more budget. Combine similar setups where sensible; do not switch to cheap clicks just to leave learning.
| When | What to check |
|---|---|
| After launch | Ads, tracking, forms, website, and chats work. |
| Daily | Unusual spending, failures, enquiry handling, and customer feedback. |
| Performance review | Allow for buying and reporting delays. A 3–7 day window may suit an initial review of short buying cycles; services may need longer. |
Spending 2–3 times your affordable cost per result without a result is a reason to investigate, not an automatic pause rule. Set limits you can afford before testing.
When to use a separate retargeting campaign
Retargeting means advertising to people who have already interacted with your business. Broad campaigns may already reach these people.
- Separate it when you need a different message or controlled budget.
- Start with one useful audience rather than many tiny groups.
- Choose how far back to include visitors based on how long people take to buy. A quick product purchase and a consultancy decision may need different time windows.
- Check that Meta will stay within your chosen audience. If your Custom Audience is only a suggestion, ads may reach other people too.
- Exclude people who already bought or enquired when the same offer is no longer useful to them.
Example: A backpack visitor needs sizing reassurance. A website prospect needs portfolio proof and a clear project process.
Common Mistake: Assuming every reported retargeting sale is an extra sale. Some customers would buy anyway; assess total business results.
Andromeda myths vs reality
| Myth | Reality |
|---|---|
| Targeting is dead | Business limits and audience strategy still matter |
| Interests never work | Interests can still work. Test them and check whether Meta can reach beyond them. |
| Everything must be in one campaign | Business needs can justify separation |
| You need 50 ads | Meta’s cited Andromeda documentation does not require every advertiser to run a fixed number of ads. |
| Every decline is Andromeda | Investigate offer, competition, creative, tracking, website, and follow-up |
10 common mistakes—and what to do instead
| Don’t: common mistake | Do: better action |
|---|---|
| 1. Rebuild because of hype | Change based on evidence |
| 2. Duplicate identical ad sets | Create separate ad sets only for a clear reason |
| 3. Split tiny budgets into many tests | Prioritize one question |
| 4. Confuse cosmetic changes with diversity | Test different buying reasons |
| 5. Upload more ads than you can evaluate | Run only as many ads as your budget can meaningfully test |
| 6. Judge success by cheap clicks or chats | Measure customers and qualified leads |
| 7. Edit every day | Use planned reviews |
| 8. Declare winners after one sale | Look for repeated results across a suitable review period |
| 9. Assume automation fixes measurement | Verify reported outcomes |
| 10. Scale before operations are ready | Check fulfilment and follow-up |
Pre-launch checklist
“Andromeda improves how Meta shortlists ads. My job is to provide a clear goal, different reasons to buy, accurate tracking, and a sensible budget.”
Sources & evidence
Official publications support platform explanations. Practitioner sources provide interpretation and interface context. Structures, counts, budgets, and operating examples are recommendations, not Meta requirements. Features can vary by account and eligibility.
- Meta Engineering: Andromeda’s role in ad retrieval — December 2024.
- Meta: Advantage+ automation launch — August 2022.
- Meta Engineering: subsequent ads-ranking improvements — August 2026.
- Meta Blueprint: Performance 5.
- Jon Loomer: Why creative diversification matters — industry interpretation.
- Jon Loomer: Advantage+ campaign setup and audience controls — practitioner guide.
- Meta Blueprint: automation tools.
- Meta: lead-generation guide — CRM and Conversions API principles.
- WhatsApp Business: lead and purchase optimization.
- Jon Loomer: learning thresholds and account tests — industry reference, not a newly verified universal 2026 threshold.