Playbooks / Creative Intelligence

Five sources. One sequence.

Creative coverage maps the territory an account needs to exist. This process determines what fills it: the angles, personas, and hook language that go into every brief before production starts. Run in order, on a weekly cadence, from external signal rather than internal opinion.

Part of The Creative System — Layer 00, the input layer. This is where angles, personas, and hook language come from before a brief is written. It feeds into Coverage, which maps how that material gets distributed. Three supporting playbooks go deeper on specific sources within this process: Comment Mining, AI Workflow, and Claude in Creative Ops.

Channel

Meta (Facebook & Instagram)

Applies to

Lead Gen · eCommerce

Focus

Pre-brief research · Signal sourcing

Research stack — source → output

Reputation audit

Awareness layer

Review mining

Hook language

Persona mapping

Buyer priority

Past performance

Angle leaderboard

Competitor audit

Gap map

The rule: If you cannot point to where an angle came from — a review, a comment, a performance data point — it is still a guess.

Signal sources

1×

Weekly cadence

Brainstormed angles

The stack

The order matters as much as the sources.

Running these five sources in isolation produces five disconnected lists. Running them in sequence produces a brief with a clear point of view. Each source builds on the one before it. By the time you reach competitors, you have enough of your own signal that you're checking for gaps, not looking for ideas. That distinction changes what you come away with.

Source + how to run it

What to extract

What it produces

First

Reputation audit

Google the brand name, the category, and the main problem the product solves. Read the first two pages of results as a skeptical buyer doing pre-purchase research. Check BBB, Trustpilot, Reddit threads, and any review aggregators in the category. Note every friction point, every trust signal, every question the results don't answer.

What to extract

What objections exist before the buyer reaches your brand. What third-party narratives dominate the category. Whether the category itself has a credibility problem that predates your client. In home services, this step often reveals that the buyer's first search surfaces contractor horror stories. If the ad doesn't acknowledge that skepticism, it is competing on the wrong level entirely.

What it produces

An awareness level assessment. What does this buyer already believe, and what are they afraid of, before they see the ad? This determines where the hook needs to enter the conversation. Not at the product, but at the buyer's existing mental state.

Second

Review mining

Export reviews from Google, Yelp, Amazon, app stores, and post-purchase surveys. The goal is verbatim customer language, not sentiment summaries. Look for the moment-of-decision language: what finally made them act. Look for failed-solution language: what they tried before this. Flag anything with unusual emotional specificity. That is the raw material for hooks.

How to mine ad comments specifically → Comment Mining playbook

What to extract

The specific phrases customers use to describe their problem before finding a solution. "My garage floor looked like a gas station" is a hook. "Suboptimal flooring aesthetics" is not. The rougher and more specific the language, the closer it is to what a real buyer will recognize as their own thought.

What it produces

A raw language bank. 20 to 40 angle candidates in the customer's own words. For processing volume at speed, AI earns its place here: it can cluster hundreds of reviews by pain point, trigger, and transformation language in minutes. The judgment about which phrases are worth pursuing stays yours.

How AI fits into this step → AI Workflow playbook

Third

Persona mapping

Build two distinct sets of personas. First: who the account appears to be targeting based on live creative and demographic data. Second: who is actually buying based on reviews, post-purchase surveys, and sales call notes. These are frequently different people. The gap between them is usually the single largest untapped creative opportunity in the account.

What to extract

Life stage, decision context, and the specific trigger that moves each persona from consideration to action. A 54-year-old homeowner replacing windows after a storm is not the same buyer as a 38-year-old doing a planned renovation. The same ad will not work for both. Most accounts run one ad for both and wonder why CPL looks inconsistent.

What it produces

A ranked buyer map: which persona represents the highest-value opportunity, which one is currently under-served by creative, and which angles are missing for each. In a new account with no prior data, this map becomes the creative launch sequence. It determines which persona gets brief priority before a dollar moves.

Fourth

Past performance data

Pull account data across three windows: last 30 days, last 6 months, all-time. Rank by spend for each window. For the top 3 and bottom 3 in each window, name the underlying angle. Not the hook line, not the format. The claim or belief the ad builds. Then look for proven angles that only ran once and were never iterated on. Those are the most underexploited resource in most accounts.

What to extract

Which angles have confirmed signal that hasn't been fully exploited. Whether performance clusters around a persona or awareness level the account is over-indexing on, or quietly abandoned. Whether what looks like a creative problem is actually a structural one: testing and scaling sharing the same campaign, starving new concepts of spend before they can show anything.

What it produces

The account's own angle leaderboard. First-party signal no competitor has access to. This step also functions as a structural diagnostic. Read the data here before briefing anything new. A brief written without this step may produce creative that duplicates an angle that already failed, or that tests something the account already knows works but never scaled.

Last

Competitor audit

Open Meta Ads Library. Filter for competitors running 90-plus days. Those are their confirmed winners. Note the lead angle, hook type, and format for each. Then write down what they are not saying. The gap between what every ad in the category says and what buyers actually feel is where differentiation lives. Run this step last: doing it first installs a copy-by-default instinct that overrides your own data.

What to extract

Category-wide patterns: what angle every competitor leads with (usually price or speed), what the creative format default is, and whether anyone is speaking to the skeptical or solution-aware buyer. Look for the angle the whole category avoids. That avoidance usually comes down to the brand having to acknowledge something uncomfortable about the category or the buyer's experience. That is exactly why it tends to be the most effective opening.

What it produces

A differentiation brief. By this stage you already have validated angles from your own data and customer language. Competitors are a gap-check, not an idea source. The question is not what they are doing. It is what they are not doing, and whether your own data points to a buyer sitting in that space.

By vertical

Same sources. Different extraction priority.

The five sources don't change across verticals. What changes is where the highest-signal material lives, how long the buyer consideration cycle runs, and which source to weight most heavily when building the first brief. A $30 DTC product and a $15,000 bath remodel have very different research landscapes. The sequence is the same. The emphasis is not.

Home Services · Lead Gen

High-ticket, long consideration, category skepticism

Lead with the reputation audit

In roofing, windows, and bath remodeling, the category has an entrenched contractor-trust problem. Google results surface complaint forums and horror stories before the brand appears. If the ad doesn't acknowledge this context, it's competing on the wrong level. The reputation audit determines the awareness layer the hook needs to enter through.

Mine for triggering events, not features

In home services, the buyer has usually been sitting on the idea for months. What finally makes them call is a specific event: a storm, a structural crack, a family circumstance. The hook that names that trigger outperforms the hook that describes the product every time.

Read performance through to close rate

CPL is the most misleading metric in lead gen. An angle producing cheap leads that don't set appointments is invisible in CPL data. Audit past performance all the way through the sales funnel, set rate and close rate, before briefing anything new. An angle with a high set rate and a higher CPL is almost always the better brief.

DTC eCommerce

Short cycle, high volume, faster signal decay

Lead with review mining

Higher review volume, faster to process. Post-purchase survey language is the richest source. It captures the exact moment a buyer decided to commit, not their satisfaction six months later. Mine for purchase trigger language specifically: what made them add to cart on that day, at that moment, after potentially seeing the product before.

Add social comment scanning

Viral posts in the category reveal real-time format signals. When a hook style is being widely distributed by the algorithm, that is a brief direction. Not a cue to copy the content, but a signal about what opening frame the platform is currently rewarding.

Run performance audits more frequently

At meaningful DTC spend levels, winning angles decay faster and the gap between what was working last quarter and what is working now is larger. The weekly rhythm below is not optional at this scale. New angles need to be entering the test queue before performance drops, not after.

📋

Case Study · Basement Renovation

This process run end-to-end before a dollar moved

The basement renovation account had no prior data, no pixel history, no creative to inherit. The pre-launch research phase produced a ranked buyer map, four hook types briefed against specific buying situations, and a 27% set rate by month two.

Operating cadence

Research is a weekly habit, not a launch task.

The accounts that keep finding new creative territory don't run deeper research at launch. They run lighter, more consistent research every week. The signal stays current because someone is checking it. Angles don't emerge from quarterly planning sessions. They come from the person who read Tuesday's comment section and noticed something that rewrote Friday's brief.

Monday

Review last 7 days of account data. Update the angle leaderboard. Note what is gaining spend, what has stalled, what was cut. Brief the next batch from signal, not from memory or last week's meeting notes.

Tuesday

20 minutes in comment sections: the ad account's own comments, plus comments on 3 to 5 viral posts in the category on Instagram and TikTok this week. Log any new objection language, questions, or emotional triggers appearing for the first time.

Wednesday

Pull new customer reviews or post-purchase responses from the last 30 days. Add standout language to the angle bank. Flag anything that contradicts what current creative is saying. That contradiction is usually a brief.

Thursday

Check competitor ad libraries. Note any new ads gaining spend that weren't running last week. Look for angles entering the category, not formats to copy. Identify what they're still not saying. That gap remains yours to take.

Friday

Finalize briefs for the following week. Every brief must cite its source: the review phrase, the comment, the data point. If the angle cannot be sourced, it does not ship as a brief. This is the quality gate.

The brief standard

If you can't point to where it came from, it isn't research.

The brief is the quality gate for the entire process. Every angle should trace back to a specific source. That traceability is not process overhead. It is what separates a brief that produces a winning creative from one that produces a plausible-sounding guess. Brainstormed angles can look identical to sourced ones in a document. The difference shows up in hook rate within the first five days of testing.

Brief is ready when:

✓

The angle traces to a specific source: a review phrase, a comment, or a data point

✓

The persona is named with a decision context, not just a demographic

✓

The awareness level is stated — where in their journey does this buyer sit?

✓

The hook opens at the buyer's situation, not the brand's product

Brief is not ready when:

The angle came from a team brainstorm with no external source

The persona is a demographic ("women 35–54") without a decision context

The hook describes what the product does rather than what the buyer feels

Next in the system

Research fills the territory. Coverage maps it.

Once the angles, personas, and hook language exist, creative coverage determines how they get distributed across the account, and how to audit what is missing before briefing anything new.

Creative Coverage Playbook