Meta Ads creative after Andromeda

Start with what Meta actually says

Meta described Andromeda in December 2024 as a personalized ads retrieval system. Retrieval is an early stage that selects a smaller set of relevant ads for later stages of recommendation. The engineering account describes improved personalization and capacity to handle a larger volume of creative.^[1]

That description supports taking the range and relevance of your creative seriously. It does not establish a universal number of ads to produce, a fixed replacement schedule or a rule that a small visual edit will be rejected. Those claims need their own evidence.

For a marketer, the useful question is practical. What different reasons could the right customer have to choose this product and have we made those reasons clear enough to test? The suggestions below are a working approach that needs the context of your own account.

Find the question behind the creative

Start with the things a buyer has to understand or believe before taking the next step. A useful brief might focus on a frustration, a specific use case, an objection or a piece of evidence that makes the offer easier to assess.

Consider a reusable bottle. One customer may want to stop buying water on the commute. Another may care about carrying it in a laptop bag. Someone else may need to see how easy it is to clean. The product is the same, but the reasons to pay attention are different.

These are illustrative ideas. Before turning one into an ad, check that the product can support the claim. A leak demonstration needs truthful conditions. A cost comparison needs defensible assumptions. A customer quote needs permission and an accurate context.

Sales questions, support conversations, product reviews and the founder’s experience can help identify useful angles. Keep the distinction between a customer’s actual words and your interpretation. An invented quote should never be presented as customer evidence.

Give each concept a clear job

Write down the customer motivation before deciding on the format. Then choose the opening, visual treatment and explanation that make the idea clear. A direct demonstration, a founder explanation and a simple still image can be sensible treatments for different briefs.

Keep track of the idea separately from its production versions. A crop for another placement, a shorter edit or a different caption can serve a useful delivery purpose. Recording them under the same concept makes the creative inventory easier to understand.

For example, an account may contain twelve exported assets but only three customer motivations. That is a useful distinction when deciding what to make next. The point is to understand the range of ideas you have actually covered.

Give the production team enough information to make good decisions. Include the audience, the reason to care, the claim you can support, the opening moment, the intended next step and the required versions. Keep the brief short enough that someone will use it.

Set up a test you can interpret

Decide what kind of answer you need before launch. You may be looking for ideas that deserve more development, or you may need a stronger comparison between two defined alternatives. Those are different decisions and can call for different test designs.

When you need a causal comparison, use a properly controlled experiment with random assignment and a defined outcome. Randomization is a basic principle of experimental design. It helps reduce the influence of other factors on the comparison.^[2]

For ordinary campaign results, be careful about unequal exposure. Two ads that received different spend, impressions or audience mixes have not necessarily had a comparable opportunity. Treat a low-spend ad as a weakly observed idea until you have enough context to say more.

Agree the primary outcome, budget guardrails and review plan in advance. Choose a test you can afford to observe. A large list of concepts may divide a modest budget so thinly that none provides a useful answer.

The right observation period depends on the amount of data, the size of the difference you need to detect and the customer’s time to act. A fixed number of days is a planning convenience, not proof that a result is settled. Record uncertainty rather than dressing a small difference up as a firm finding.

Read attention and acquisition together

Attention measures help you ask better questions about the creative. A weak opening may suggest the scene is hard to recognize or the message takes too long to arrive. Strong viewing followed by weak response may point you towards the offer, the next step or the destination page.

Treat these as hypotheses to investigate. A viewing metric alone cannot explain the full path to a purchase. Click behavior and conversion behavior also need the context of the placements, objective, audience and offer.

For an e-commerce business, compare acquisition cost with the economics of the order. Account for the business data you actually have, including product margin, fulfilment costs and returns where relevant. Keep new-customer acquisition distinct from repeat purchases when you can measure the difference reliably.

For a service or SaaS business, involve the people who handle the enquiries. Ask how many fit the offer, progressed to a useful conversation and became customers. A lower cost per form submission can still leave the sales team with less useful work.

Keep the attribution source visible. A platform report, your order system and your sales records may answer different questions. State which view you are using and avoid presenting a platform-attributed result as proof of incremental sales.

Investigate a decline before naming it

A previously useful ad can stop meeting the goal for several reasons. The creative may have become too familiar to the people seeing it. The offer, stock, landing page, competition or measurement may also have changed.

Look at the timing. What happened to delivery, audience exposure and the wider business around the decline? Are several concepts changing together, or is one treatment behaving differently? Those questions help narrow the next investigation.

A new creative brief should follow from what you learned. You might need a fresh customer motivation, clearer evidence, a different format or a more relevant opening. Replacing material on a fixed calendar can be a poor substitute for understanding the problem.

Keep useful ideas in the record even when they stop receiving spend. Note what the creative tried to communicate, where it ran and what you could reasonably infer. A later team member should be able to understand the decision without reconstructing the account.

Leave a record the next brief can use

A simple creative log can make the work more useful over time. It gives the media buyer and the production team a shared account of the idea, the evidence and the decision.

For each concept, keep the customer motivation, the question being tested, the supporting assets and the outcome you watched. Add the conditions that mattered, such as the offer, audience context and limits of the data. Finish with the next decision and its reason.

That may lead to another version, a fresh angle or a decision to investigate something outside the creative. The value is in making the next step easier to explain.

The Meta Ads service at edgeshift covers creative decisions alongside campaign management. A separate production brief can cover the concepts, filming and edits when content is the work you need.