Back to Sam Tomlinson
🎯
Sam Tomlinson

PPC & Google Ads

Issue #137 | Creative in the Age of Andromeda

Sam Tomlinson <sam@samtomlinson.me>
October 12, 2025

(

link

)

***********************

Happy Sunday, Everyone!

***********************

I trust you’re all enjoying Fall’s prime - the changing leaves,

the brisk mornings, the football-filled weekends and the

never-ending onslaught of conferences…all taking place in the

shadow of Q4 goals + BFCM. Somehow, Thanksgiving is already on

the extended forecast (the really, really extended one, but

still), with the winter holidays and 2026 not far behind.

During the first run of Fall conferences, I had the opportunity

to meet with several folks from Meta, who were kind enough to

share some insights on both the future of the Meta ads platform,

as well as specific insights on what they’re seeing perform since

the full-scale roll-out of GEM, Lattice & Andromeda over the past

~6 months.

If you’re not familiar (or frantically Googling, “What is Meta

Lattice”), let me save you the trouble:

* GEM = This is Meta Ads’ “brain” – it ingests trillions of data

points (everything from ad creatives, messages and offers to

offline events, behavioral sequences, personal characteristics

and more), maps the relationships between each one, then

leverages the model created by this process to predict which ads

are most likely to produce a desired action from each person, all

in near-real-time.

* Lattice = The easiest metaphor for Lattice is the Library of

Congress. Where most towns operate a library (or several), each

with some combination of the classics, some regional favorites,

and some picks from the librarian, the Library of Congress

contains copies of every single printed publication in the

history of the US. Before Lattice, Meta’s ad library operated

very similarly to those regional libraries - each one held ads

based on geo, objective, surface (IG, FB), industry. Lattice has

combined them all into one, super-massive library that can be

readily searched and analyzed by Meta’s ad systems. The end

result? Meta has a richer understanding of how different ads work

across objectives, and can more accurately predict which

creative(s) are more likely to convert.

* Andromeda = Andromeda is (as Meta describes it) a “personal ads

concierge” that leverages staggeringly sophisticated machine

learning to refine the pool of ads available to be shown to you

to only the ones most likely to resonate with you, based on your

tastes, preferences, behavioral history, connections,

relationships, etc. Think of it like your personal,

fully-customized ad filter: Andromeda uses its accumulated

knowledge about you to filter out billions of ads (literally),

leaving only a few thousand that are most likely to be relevant

to you. Only the ads that make it through Andromeda’s filter are

eligible to be served to you by Meta’s ad systems.

Each of these three systems is interconnected, working together

(along with sequence learning, which is a system that allows

every Meta ads model to consider the order in which events happen

before and after an ad exposure in order to enhance delivery +

relevance) to improve Meta’s ad delivery systems. At first

glance, each of these three systems probably seems quite good for

advertisers – after all, getting more relevant ads delivered

sounds like a huge win. Meta having a “better understanding” of

what ads are most likely to lead to certain outcomes probably

seems like exactly what we want – but the devil is in the

details.

And that’s what brings me back to the conversations with several

senior members of Meta’s Ads team. The big takeaway is that,

through these three systems, Meta is fundamentally rethinking

what it considers a “unique” or “distinct” ad. The days of

driving creative diversity via uploading step-change, iterative

variants of a winning ad are over. That’s a big deal (aside -

it’s also something I warned was coming way back in 2023 (

link

))

But the bigger deal? Meta is finally beginning to leverage the ad

set audience as the primary lever for creative diversity. The

impact of this change can not be overstated. It is a tectonic

shift in the Meta ads landscape.

And it’s happening right now. If you’ve been in Ads Manager over

the last few days/weeks (rollouts take time), you’ve likely seen

several changes to the “recommendations” section: (1) Creative

Similarity (the exact text: “Using creative that doesn’t appear

visually similar in your ads can help those ads resonate better

with audiences…”) and (2) Ad Fatigue (“creative fatigue occurs

when an audience has seen the same creative too many times.

People may be less likely to engage with your ad, which can

increase your cost per result…”). These are direct, downstream

effects of the above big changes to Meta’s ad systems….and we’re

still early in the rollout + fallout.

For today’s issue, I want to share what I’ve heard directly from

several Meta Product Leads, alongside data from accounts that

corroborates what I’ve been told and how we’re thinking about

modifying our creative strategies moving forward in response.

Let’s dig in.

The immediate implications of this are direct from Meta’s team:

Lattice + Andromeda functionally compress the number of ads

available to be served (Lattice accesses billions of ads in a

single library - meaning Meta can now “see” every single ad in

one view; Andromeda “dedupes” all of the similar ads then

filters/curates based on your tastes/preferences/history). The

primary reason for this is that it saves compute while

maintaining a viable set of available ads. The logic seems to be

that if Meta combines all the “similar” ads first, then filters

them, it (essentially) guarantees that it will have a

sufficiently large set of potential ads to enter into the

auction.

That means that Meta categorizes two ads as “identical” if they

share the same visual aesthetic (statics/carousels) OR a similar

first 3 seconds (video ads). Both Lattice + Andromeda are

visual-first systems; ads with similar visuals are considered

identical even if the copy or offer is wildly different.

Yes, you read that correctly.

Visual similarity now dominates narrative variation in Meta’s

assessment of distinctness across all ad formats (statics,

carousels, videos).

The implications for that change are staggering:

* Different copy/CTA on same visual = same ad

* The same creator (UGC, founder, VSL, whatever) in the same

scene (i.e. office

background, in front of a mirror, etc.) = same ad

* The same hook followed by two different stories = similar or

same ad (depending)

It’s abundantly clear that these changes are a direct response to

how advertisers tried to “game” Meta’s old ad system: find a

winning ad, create 30 variants of it, load ‘em all into ads

manager, profit. Well, thanks to Andromeda, those 30 variants are

now treated as 1 ad. This puts the onus back on advertisers to

create truly diverse, differentiated creatives for the system to

consider.

-------------------------------

Creative Relevance Is the Lever

-------------------------------

For what feels like ever, we’ve heard that creative is the

targeting lever for Meta. The implication of that was: don’t

worry about the audience, don’t worry about interest/lookalike

targeting, just go broad with creative and Meta will figure it

out. Obviously that was not (and is not) entirely true – but (as

with most things), it was half true. What we’ve found is that

creative relevance - the connection between creative and audience

- is the thing that matters. Align your creative - offer -

audience - experience (i.e. creative relevance) and you’ll see

improvements in engagement (CTR) and action (CVR) rates.

This isn’t new. If anything, this is Meta algorithmically

implementing a core tenet of behavioral economics: contextual

relevance drives engagement more than precision targeting. Users

respond when they see themselves - visually, emotionally and

motivationally - inside the ad. This is validated by three

behavioral economics principles: (1) the Self-Referencing Effect;

(2) Framing Theory and (3) the Mere Exposure Effect.

However, this does not imply that ad set-level targeting is

obsolete. In fact, persona-based ad sets, structured around a

behaviorally-defined audience, make more sense than ever when

paired with true creative diversity.

If each ad set targets a distinct persona and includes multiple,

visually divergent ads tailored to that persona’s fears, hopes,

dreams, challenges and/or motivations, the system has a clearer

signal to work with - Andromeda can then match the right angle to

each individual member of the targeted audience segment.

Ironically, the alignment of behavioral/audience-based targeting

and informed creative relevance improves ad relevance and drives

deeper personalization.

On the flip side, broad targeting remains useful for maximizing

reach with diverse creatives, but tight persona-level targeting

gains power when matched with informed, intentional creative

differentiation at the ad set level.

-------------------------------------------------------

When New Things Happen, Sometimes The Old Ways Are Best

-------------------------------------------------------

When big things change (and this certainly counts as a big thing

changing in a big way), my default strategy is to go back to what

never changes. It sounds crazy and counterintuitive, but often

there’s an answer for today’s crazy challenges in yesterday’s old

truths.

Case in point: as I was thinking about how to respond to these

changes, I wound up calling (yes, on the phone) a good friend,

Rabah, to catch up. Somehow, we ended up spending 30+ minutes

talking about the classics - Ogilvy, Bernbach and Burnett - the

old-school, mad men era guys. And ironically, I found the lessons

from those long-gone advertisers to be more applicable now than

they were back when they were first shared. In fact, how I’m

approaching the Andromeda era is actually shaped by some of their

lessons/principles:

“Creative is not creative unless it sells.”

One of my favorites. Ogilvy insisted that advertising must

ultimately drive results. Endless experimentation that doesn’t

scale or endure is not real creativity. In the current landscape,

your creative needs to do two things: immediately attract

attention and sustain relevance to the target audience over time.

That does not mean the ad needs to appeal to everyone - in fact,

the opposite is true: your creative should NOT appeal to people

outside your target audience. It does mean that your ad must

stand the test of time to earn sustained delivery – the quick,

flash-in-the-pan meme ads aren’t likely to sustain your account.

Focus on creating ads that are likely to work for a while, not a

minute.

Write Ads For Your Audience, Not For Everyone

Ogilvy’s most enduring lesson is brutal in its simplicity: pick a

position, pick a person and speak in their language. The most

effective advertising reads like a one-to-one letter, not a

stadium announcement. It uses the target audience’s words, frames

and fears to make a single promise unmistakably relevant to them.

When you create ads for “everyone,” you sand off the edges that

signal meaning to someone; relevance collapses into generality,

and generality doesn’t sell. Choose the audience, honor their

vocabulary and context and make the ad feel made-for-me to them -

because if it isn’t for a specific person, it isn’t for anyone.

If you can’t tell immediately who the ad was created for or who

it is speaking to, it is too generic.

The customer is not a moron.

Ogilvy famously cautioned against advertising that underestimates

or condescends to the audience. With audience-driven ads, respect

cognitive complexity. Don’t default to shouting features or

parading out awards/recognition. Instead, focus your ads on

meeting the audience segment’s emotional and identity-level

needs; assume the audience is intelligent, somewhat informed and

capable of understanding how your product/service can improve

their lives.

Focus on the Fundamentals of the Human Experience

One of Bernbach’s most enduring observations was that while

everything else may change, people don’t – we’re still motivated

by the same handful of things that we always have been. The

durable advantage is creating ads based on a single, timeless

human drive, then expressing it in the audience’s language,

visuals and context.

Survival: turn fear into agency

Identify a real, near-term risk, then show the path to safety

(via your product/service).

* Hook: the avoidable mistake or looming loss.

* Visuals: problem-first, high-contrast; fast cut to resolution.

* Proof: demo the safeguard; reassure with clarity (guarantee,

fail-safe).

Admiration: signal status

Show (don’t tell) how the product/service confers belonging to

an admired group.

* Hook: “for the ones who…” (set the bar, lead the field).

Visuals: elevated design, distinct palette, aspirational

archetypes.

* Proof: tasteful third-party signals; social proof as texture,

not trophies.

Success: show transformation

Walk the viewer from “now” to “next,” with measurable progress.

* Hook: time, money, or outcome delta (“Go from 3 hours to 15

minutes,” “+27% ROI”).

* Visuals: timelines, side-by-sides, dashboards, time-lapse.

* Proof: specific numbers and receipts; the user is the hero,

product/service the enabler.

Love: create connection and belonging

Make the audience feel seen, supported and not alone. This

doesn’t have to be purely romantic love - but rather any form of

connection and belonging.

* Hook: empathetic mirror (“If this sounds like you…”).

* Visuals: lived-in settings, real people, warm tone;

micro-moments > polish.

* Proof: authentic testimonials, DMs, UGC clips; speak plainly.

Protection: defend the tribe and its values

Draw a line around standards, origins and what you won’t

compromise.

* Hook: conviction (“Made for [community], not for everyone”).

* Visuals: craft, provenance, badges; behind-the-scenes rigor.

* Proof: guarantees, certifications, founder voice; expect

(healthy) polarization.

Test For Insights & Learning, Not P-Values

Ogilvy’s enduring lesson on testing remains urgent in the

Andromeda era: experimentation must be systematic, intentional,

and designed to uncover insight, not to chase statistical

significance. The goal is not to prove that one variant “wins” at

the 95% confidence level, but to understand why an ad resonates,

with whom and under what contextual or emotional conditions.

Modern creative testing should function as behavioral research:

mapping motivators, decoding visual triggers and refining

hypotheses about audience psychology. In other words, data should

fuel creative intelligence, not just validate it.

From those principles, we’ve evolved our approach to creative

strategy:

Creative Must Be Audience-Centric

---------------------------------

The primary reason why I’m obsessed with audience understanding

isn’t because I love SparkToro (though I do) or doing research

(which I don’t) – it’s because understanding the audience at a

deep, human level massively increases the odds of creating hit

ads. It’s really that simple.

From a practical standpoint, that means that we spend an

inordinate amount of time on audience understanding + insight,

then design ads around psychographics, motivations + audience

preferences (what sources do they trust? Where will they get this

information? Do they care about third party credibility or

prioritize personal relationships?) – not visuals or messages.

That’s a radical departure from how most agencies manage creative

(which is disproportionately focused on output + iteration) - but

one that is now being rewarded by both audiences AND platforms.

If you don’t truly, madly, deeply, viscerally understand your

audience - to the point where you can empathize with them,

predict their responses, speak directly to their pains,

challenges, hopes and dreams - you’re not going to be able to

generate the caliber of diversity Meta is seeking.

If you’re curious about how we conduct audience research, I wrote

about that here (

link

).

Immediate Visual Diversification Of “Hit” Ads

---------------------------------------------

When a “hit” ad emerges, the natural tendency for most

advertisers is to immediately begin making small, incremental

changes to it – a slight adjustment to the hook, a different CTA,

a new value prop, alternative proof points, whatever. That’s

never been something I’ve gravitated toward, because

(overwhelmingly) my thesis is that climbing a little higher on

that particular mountain isn’t likely to be a positive expected

value bet – I’m more likely to spend a bunch of time/money/energy

to get a bunch of derivatives that perform pretty much the same

as the original.

In fact, a few issues ago, I shared how we often will take the

script from a “hit” ad, upload it (alongside our target audience

details) to an AI system (Gemini or GPT), and have it create 5-10

diverse, new scripts based on the script + insights it can mine

from the audience profile we provided. We’ll then give those

scripts to different creators.

* Swap locations (indoor to outdoor, studio to field)

* Change spatial composition (wide, tight, split screen, motion)

* Replace set, camera style, framing

* Swap creators + formats

* Modify the drive (same audience, different approach)

In each case, we’re treating the original “hit” ad as the

creative center of gravity, then branching out to different

worlds via diversification. That’s a radically different approach

vs. simple iteration on a winner (i.e., treating the winner as a

template to be copied).

Spend 80% Of Your Effort on the Hook

------------------------------------

One of the primary takeaways from the Meta team on Andromeda is

that the system disproportionately determines “sameness” based on

the hook - the first ~3 seconds of the ad. This creates a massive

opportunity for advertisers to unlock new segments of your

audience simply by adjusting the hook. The rest of the ad can

change slightly, but (as I wrote here (

link

)), spend 80% of your time focused on the hook.

Proactively attempt different hook styles - immediate action,

result-and-rewind, secrets, stories, non-traditional narrative

arcs, hopeless-until-solution, before-and-after, bold statements,

whatever - to the fullest extent possible.

As a bonus, the beautiful thing about a hook-first approach is

that the hook will dictate much of what follows, restricting your

set of possibilities in the middle/end of the ad from a

narrative, logical and tonal perspective.

The analogy I’ve used before (and which still holds) is that of a

chess game: the hook is your opening. It shapes what happens in

the mid game, which (in turn) dictates the end game. While a

fantastic opening does not guarantee victory in chess (and,

conversely, a brilliant hook does not guarantee the ad will be a

hit), it certainly does improve the odds.

Smart Iteration + Repackaging

-----------------------------

About 50 issues ago, I shared our creative matrix – how we think

about Structural Iteration in ads (you can view the full article

here (

link

)):

And this framework is more valuable today than it was when I

shared it years ago, simply because it leans in to the three

things that will thrive in the age of Andromeda: (1) audience

centricity (the inputs + idea generation); (2) the concept of

tokenizing ads into component parts (i.e. the red part of the

diagram above) and (3) recombining those components to create

maximum diversity while minimizing creative costs.

If I was going to add anything to this, it would be to add format

as an axis for diversity (something I wrote about here (

link

)) – more brands should be trying more different types/styles of

ads. Too many brands are too stuck in their ways, with each new

creative following a relatively established, semi-predictable

script. The brands with true diversity scrap that convention +

create ads that are wildly different, to the point where the

audience is surprised they’re all from the same brand.

Create Worlds Where Your Ads Can Take Place

-------------------------------------------

Another big shift we’ve begun to adopt is the “worlds” framework.

This is born out of the observation that differentiated creative

must simultaneously create distinctiveness and separability for

Meta while maintaining salience for the audience (basically,

people like familiarity - just not too much).

For small teams, achieving that balance can be somewhat easy –

the same creatives tend to work on the same projects, so they can

reference/create that feeling of consistency. As you scale, doing

that becomes much more difficult. The answer to that is worlds. A

world is a repeatable, rules-based environment that governs how a

creative concept looks, sounds, feels and operates. It’s the

combination of setting, lighting, palette, cast (including

creators/talent), voice, camera, symbols/artifacts and proof

style. Each world has its own set of rules that dictate where and

when it lives (setting, time of day), how it feels to the

audience (palette, lighting, cut cadence), who speaks and how

they speak (cast, dialect, cadence, language), what carries

meaning (props, iconography) and how truth is shown (demo,

receipt, stopwatch, testimonial).

Examples of this might be: “The Kitchen Table At Night” or

“Morning Mirror” or “Workshop Desk” or “The Corner Office in the

Afternoon” – each one immediately creates a specific feeling,

while giving you endless opportunities to express the same core

concept in materially different ways.

The inspiration for this was sitcoms – we’ve all watched Big Bang

Theory and immediately knew what kinds of things might happen in

Leonard & Sheldon’s apartment, or the Cheesecake Factory or the

Comic Book Store (in fact, if you watched BBT, you can probably

picture each of those scenes in your mind right now). The reality

is that what happened in each of those places in each episode was

quite different - but the “world” where the action took place

created a sense of both narrative and visual coherence and

familiarity.

That’s exactly what we’re doing here - but instead of a show,

it’s your advertising. From Meta’s perspective, two ads that

share the same copy/messaging but live in different worlds are

distinct – Andromeda + Lattice won’t collapse them into the same

concept. That provides you with the ability to scale while

maintaining audience familiarity.

Adopt A Perpetual Creative Graph

--------------------------------

Linear campaign planning assumes finite stories, fixed audiences

and a platform that rewards cadence (i.e. shipping more ads over

a certain period of time). But, Meta now rewards something very

different: differentiation, specificity and diversity.

Fundamentally, that means the operating model we use to plan +

execute must evolve.

Instead of the old, outdated “campaign planning” approach, think

of your creative system as a perpetual graph - a living network

of ideas and evidence - rather than a sequence of launches or

iterations.

The building blocks become nodes: the audience, the concept (big

idea), the human drive it taps, the hook, the world, the proof

type and your distinctive brand cues. The edges, then, are the

relationships between those nodes. “Hook A × Survival Drive ×

World 2 × UGC × Bundle Offer 6 produced a “hit” ad for the

Stressed/Overworked Toddler Mom Audience in Q4” isn’t some random

thing; it’s actionable audience insight. Over time, the edges

become IP: a proprietary map of what works for whom, when and

why.

The result of this is that you stop “launching campaigns” and

start “querying the graph” – given a specific audience and

objective (sell product X or generate leads for service Y), you

ask: What is the next best expression Meta will immediately

recognize as new, relevant and distinct? The answer returns a

small set of combinations - hook × drive × world × format ×

Product/Service/Offer - that are both available (i.e. we have or

can get the assets) and untested (we have not created an ad like

that yet). Those become the next creatives for the ad account.

This also fundamentally changes how we evaluate what to create

and launch in the ad account – the tests now are simple:

* What is the probability that Meta will recognize this as

meaningfully different from the existing creatives in this ad

set?

* Does this combination align with our understanding of this

audience? If not, is there something we could have missed in our

research that suggests it is worth a try?

* Even if the performance of this concept is merely average, does

testing the combination generate learnings we can leverage later?

This completely flips how you approach creative production –

because instead of focusing on raw output or creative iteration,

we’re now solving for two things: (1) expected value and (2)

learning. The end result is a compounding advantage. As you map

the edges in the graph, the areas of opportunity become more

obvious. True creative differentiation becomes easier. Results

become more predictable.

The Bottom Line:

----------------

The entire goal of Andromeda (and Lattice, and GEM) is to force

advertisers to get back to being advertisers – to create truly

unique, differentiated ads that actually resonate with different

sub-segments of their target audience. Fundamentally, that means

the days of small, incremental tests are over. The accounts that

will thrive are the ones that can blend bold, big-picture

variation with structured experimentation.

If and when you see some of your ads stop performing (or your

iterations never serving), know that it is not because Meta is

broken. That is the system working exactly as intended. It is

more discerning + more selective.

In the new landscape, creativity is not more important than

before - it’s importance is just more obvious. The key to winning

is not to do some new-fangled thing or try to “hack” the system;

it’s to go back to the old-school, tried-and-true principles.

Meta just wants advertisers to go back to being advertisers – and

they’re aligning incentives + systems accordingly.

Sometimes, the path of least resistance AND the right thing to do

are the same thing. This is one of those times.

Until next week,

Sam

This week’s issue is sponsored by Optmyzr.

------------------------------------------

Most teams still think of Optmyzr simply as “the search tool.” In

reality, their team has quietly built a phenomenal Amazon Ads

management suite, too. I genuinely wish more teams would give it

a try, because it’s a superb product.

The quick highlights: Optmyzr for Amazon gives you the levers you

wish Seller Central had: manage Sponsored Products, Brands and

Display from one place, centralize what matters and act fast when

things get a little crazy (in either a good or bad way).

Optmyzr’s Amazon suite still has the wonderful Rule Engine for

automating the rote work like pausing expensive ASINs or

keywords, adjusting bids when ACoS goes outside of a prescribed

threshold, adding more budget to scale winning segments, all

without handing the keys to your account over to a black-box

automation software. You create and control the rules (it’s a

simple, drag-and-drop editor), you get notified when they run,

and you remain firmly in control of the big picture strategy.

Optmyzr just ensures everything runs smoothly.

Beyond the basics, you can layer in inventory alerts so you stop

paying for clicks on items that are running low, along with geo

modifiers, external data (including weather) and just about any

other guardrail you can imagine. If you run Amazon ads, it’s one

of those things that will save you hours of time every week (and

prevent more than a few grey hairs).

If you’re curious and want to see what Optmyzr can do for your

Amazon campaigns, give it a try for 2 weeks free here:

-->Try Optmyzr For 14 Days Free (

link

)

Try Optmyzr For 14 Days Free (

link )

Loving The Digital Download?

Share this Newsletter with a friend by visiting my public feed.

---------------------------------------------------------------

-->View the Newsletter Feed (

link

)

View the Newsletter Feed (

link )

Follow Me on my Socials

(

link

) (

link

) (

link

)

1700 South Road, Baltimore, MD 21209 | 410-367-2700

Unsubscribe (

link

) | Manage Preferences (

link

)

Want to read more from Sam Tomlinson?

Join Ads to AI to get full access to all 67 articles plus 500+ more from top AI and marketing thought leaders.

Join Ads to AI →