How We'll Find Your Buyers On Meta

The single most important shift in how Meta works since 2024 is that the words inside your ads are how the platform decides who sees them.

When you launch an ad, Meta reads the actual copy: the headline, the body text, the visual cues. It cross-references that against the patterns of who has clicked on similar copy before, then serves the ad to the slice of the population most likely to respond to that specific message, regardless of what you put in your demographic settings.

You cover the whole audience by varying what the audience hears.

This means a single generic ad written for "everyone in your market" only reaches a small slice of the people who could realistically buy from you. The algorithm serves it to whichever sliver of your audience the wording happens to resonate with, and stops there. The other slivers, who would also buy if they saw a different angle, never see the ad at all.

The way to widen the audience is to widen the messaging. Multiple ad angles, each one written to resonate with a different type of buyer, get served by the algorithm to different pockets of the market.

Before we write any ad copy, we sit down with your closed-won and closed-lost call recordings. The goal is to identify the ten or fifteen distinct versions of the person who buys from you, beyond the single "ideal customer" most ICPs collapse down to.

Most businesses think they have one ICP. When we go through the actual sales conversations, we consistently find between eight and fifteen distinct buyer profiles inside what looked like a single market. A customer who buys because their team is burned out from outbound is a different profile from a customer who buys because their last agency missed a launch deadline. Both end up in your business converting at high rates, but they show up looking for different solutions and respond to different language.

We consistently find between eight and fifteen distinct buyer profiles inside what looked like a single market.

Build a profile inventory. Each profile gets four fields: the trigger event (what made them start looking), the primary frustration (what they wish was different), the objection pattern (what almost stopped them buying), and the language they used in your real sales calls to describe all of it.

That inventory becomes the input to every piece of creative we write. Each profile gets its own set of ad angles, each one written in language we already know lands.

Each profile becomes the input to a creative cluster. Multiple ad angles, each one written in the voice and language of that specific profile, each one telling a different story.

One profile responds to a hook about wasted time, another to a hook about wasted budget, and a third only moves when a competitor moving faster is mentioned. The product is the same. What we say about it shifts to match the profile.

Build four to eight angles per profile, with both video and static formats. The full creative library across all profiles ends up between 60 and 120 distinct ads at any given time. That is the size of library Meta needs to serve your full reachable audience.

Most B2B accounts plateau on creative variety long before they plateau on budget.

B2B accounts plateau at 30 to 40 percent of their addressable market because of creative variety, well before budget becomes the constraint. The accounts that scale past that ceiling do it by widening the library.

On the campaign side, we run a layered audience architecture rather than a single targeting strategy. Three layers, each catching a different segment of your reachable market.

The cold layer covers people who have never heard of you. Inside cold, run three sub-audiences. Lookalikes (built from a closed-won customer list), interest stacks (Meta interest categories layered together for breadth), and broad (Meta's untargeted pool, where Andromeda decides). Each sub-audience reaches a different slice. Test all three until one or two emerge as the dominant performers, then concentrate spend.

The warm layer catches people who have already engaged with you. Page engagers from the last 365 days, video viewers from any of your existing content, and recent website visitors. We exclude past customers so you are not paying to advertise to people who already bought.

Three different starting points let the algorithm find the highest-converting buyers from each, with no second-guessing required.

The broad layer is the long-tail. Pure untargeted, with the algorithm doing all the audience selection from the ad copy alone. This layer is where Meta finds buyers your interest stacks and lookalikes miss. On accounts where the creative library is mature and the messaging is sharp, the broad layer becomes the top performer.

The architecture lets the algorithm do what it does best in each context, with three starting points covering different segments rather than a single targeting strategy doing all the work.

Lookalike audiences are Meta's way of finding new people who behave like your existing customers. The lookalike tier you choose controls a tradeoff between similarity and scale.

A 1 percent lookalike is the top one percent of the country who most closely match your customer list. Tightest targeting, highest similarity, smallest audience. CPMs are higher because everyone wants to reach this audience, but lead quality is the strongest. Every account should start on 1 percent lookalikes.

A 2 to 5 percent lookalike widens the net. The audience is several times larger, lead quality is still strong, and CPMs come down. Once a 1 percent audience starts saturating (every additional dollar of spend produces a higher cost per result), this is the next tier.

A 6 to 10 percent lookalike is for scale. Audience is huge, CPMs are cheapest, lead quality drops a tier. Use this when the other lookalike tiers have capped and the goal is reach. The trick is to pair this layer with sharper creative, because the broader audience needs more obvious targeting in the copy itself.

Meta cannot find lookalikes of bad data. The source list determines the quality of the audience.

Build lookalikes off your highest-quality data: closed-won customers when you have at least 100 of them. Below that, submitted applications or qualified leads work as a substitute. The source list determines the quality of the audience.

Phase 1 of the system maps your cost-per-call target to the financial model we built before launch. Phase 3 is where we hold the live ads to that target.

Cost caps are Meta's mechanism for telling the algorithm what an acceptable cost-per-result looks like. We set cost caps at two to five times your target cost-per-call, depending on the campaign maturity and how much data the account has.

The cap is not a hard limit. The algorithm uses it as a signal to stay inside auction ranges where your target is achievable. If the campaign stops spending, we toggle the cap up by 10 to 30 dollars at a time until spend resumes. If it spends but stays under the cap, we leave it alone.

The numbers we agreed on in week one are the numbers the algorithm operates within in week ten.

The reason this matters is unit economics. Without cost caps, a winning campaign drifts into auction ranges where cost-per-call doubles overnight. With cost caps tied to your financial model, the campaign stays inside the unit economics you signed off on. You are not waking up to surprise spend.

This is the layer that connects the ad system back to the financial model, holding live spend inside the targets we set before launch.

The biggest mistake a B2B account can make after launch is to set the creative once and walk away. Meta rewards new content with better distribution. The platform actively degrades the performance of ads that have been running too long.

We produce fresh creative every two weeks. New angles, new visuals, new hooks, sometimes new profile clusters as we surface fresh patterns from your sales calls. The new creative goes into the existing campaign structure, where the algorithm decides which old ads get retired and which new ones graduate to scaling.

Without refresh, the same library running for six months sees performance erode by 30 to 50 percent.

This is what makes the system compound rather than stall. Biweekly refresh keeps performance flat or improving as the algorithm accumulates more conversion data on what works, instead of letting the library decay through repeated exposure.

The refresh cadence is not optional. It is the price of admission for accounts that scale past six figures of monthly spend.

With the system running for ninety days, three things change about your Meta account.

Cost-per-call comes down. As the algorithm accumulates conversion data on the people who book and close on your funnel, it gets better at finding similar people. Cost-per-call drops 20 to 40 percent on most accounts between month one and month three.

Your reachable audience widens. The library has expanded from the initial 30 to 40 ads at launch to 80 to 120 ads by month three, covering more profiles and more angles. More of your addressable market sees your ads.

Scaling becomes a budget question. Once the unit economics are dialled inside cost caps and the creative library is mature, scaling means raising the budget. The structure does not need to change.

The system is built to convert better at 90 days than at 30, and better at 180 than at 90. Launch performance is the starting line.