Build a Customer Avatar That Actually Helps You Write Better Ads.

Less than you'd think. Meta no longer offers homeowner targeting, and if your ads fall in the Housing special ad category (its definition includes housing repairs) or Financial products, you can't target by age, gender or ZIP code either; location and your own customer lists are what's left, and not 'people with broken AC at 3am.' The targeting is broader; the AVATAR informs the CREATIVE. Broad audience + avatar-specific creative is the winning combo.

Run separate campaigns per service line, not mixed. Roofing, HVAC and plumbing buyers respond to different offers, and one campaign optimising for all of them gives Meta a muddy signal about who converts. Separate campaigns also let you read cost per booked job for each line.

Yearly, after major service additions, or after 3+ months of poor CPL. Avatars drift over time — especially if you've added service lines or shifted neighborhoods. A stale avatar is worse than no avatar.

Usually means your OFFER is too broad, not that your customers are too diverse. If you serve 'anyone with a roof,' your avatar can't be precise. Narrow the offer ('storm-damaged roofs in Texas,' 'aging-out roofs over 15 years') and a cleaner avatar emerges. Vague customer = vague offer = both problems compound.

Build 2-3 avatars, not 1. As a starting split (a rule of thumb), the primary avatar drives about 70% of campaigns and secondary avatars get 20-30% as test budget so you don't miss adjacent customer types. A common way to get stuck is doing the avatar exercise once and never revisiting it; markets shift over 12-24 months. Refresh avatars annually using last 90 days of actual closed-customer data, not memory.

Skip gut feel. Three concrete methods, in priority order: (1) Phone-call interviews with 5-10 of your best closed customers — ask 'what was happening when you decided to call us?', 'who else did you consider?', 'what would have made you NOT hire us?' Real customers reveal language patterns you'd never invent. (2) Audit your last 50 form-fills for repeated phrases — 'leaking', 'storm damage', 'old system' — those exact words go directly into ad copy. (3) Read 1-star competitor reviews to find unmet pain ('took 3 weeks to schedule,' 'never returned my call') — your avatar's pain is the gap competitors are leaving. Total time investment: 4-6 hours every 6 months. Your avatar will be far more accurate than any AI-generated 'persona' template.

Steal language directly from customer interviews — don't invent it. Three replicable patterns: (1) PROBLEM-FIRST opener — repeat the exact phrase customers used to describe the issue ('Your AC drips water on the carpet again? You're not crazy — it's almost always one of three things...'); (2) SPECIFIC fear acknowledgment — 'You're worried about getting ripped off by a contractor you can't vet, right? Let's solve that...'; (3) CONCRETE outcome — exact numbers + timeline ('Most repairs done same day, with an up-front price before we start'). Avoid generic phrases anyone could use: 'quality service,' 'experienced team,' 'family-owned for 30 years,' '100% satisfaction guaranteed.' These sound the same to a customer scrolling past — your competitor uses identical phrases. The differentiation comes from specific, customer-language details that prove you actually understand their problem.

Separate the creative and landing pages, not the audience: in a Housing-category campaign you can't split ad sets by age, so both ads go to the same broad service-area audience and the creative does the sorting. The mistake is trying to write one piece of copy that 'speaks to everyone' — it speaks to no one. New homeowner avatar: emphasize 'protect your investment' messaging, modern aesthetics, financing language, fast turnaround. Retiree avatar: emphasize 'we'll explain everything,' phone-first contact (avoid form-only), trust badges + reviews + insurance language, slower decision timelines. Same trade, two completely different ads. It doubles the number of ad variations to manage. Test it against blended messaging on cost per booked job. Don't over-segment to 5 avatars; 2-3 is the practical maximum for a typical contractor budget (rule of thumb). Anything more splits a small budget across too many ads to learn anything.

Run a quarterly avatar audit using these 3 tests: (1) MIRROR TEST — pull your last 50 closed customers from your CRM. Read their names + addresses + project notes. Do they look like the avatar you've been writing copy to? If most of them don't match (say 60%+, as a rule of thumb), your avatar is wrong; (2) WIN-LOSS analysis — interview 5 won customers + 5 lost prospects. Ask 'what was the deciding factor for/against hiring us?' If your avatar's deciding factors come up consistently, you're right. If different factors emerge, update the avatar; (3) AD PERFORMANCE — your top-performing creative IS your avatar talking back to you. Read the highest-CTR ad copy + the messaging that converts best. Does it match the avatar you wrote down? If not, the data is telling you the actual avatar. Revalidate quarterly (or after major life events: new service line, geo expansion, post-recession). Stale avatars quietly drag down ad performance — your copy still 'sounds right' but doesn't match who's actually buying.

Force yourself to use customer DATA, not customer ASSUMPTIONS. Three reality checks: (1) AGE/INCOME alignment — pull your last 100 customers' ZIP codes and cross-reference them with US Census income data for each ZIP. Are they actually in the income bracket you assumed? It's easy to picture customers who look like your own household; (2) DEMOGRAPHIC mirror — owners tend to build avatars that look like themselves. Your closed customers may skew female (homeowners often delegate maintenance decisions to spouses), older (60s + retirees with property maintenance needs), or different family structures than you. Pull the actual gender + age data from your CRM; let the data override your gut; (3) PAIN-POINT calibration — your assumptions about 'what worries homeowners' come from your own home experience. Customer interview questions reveal different pains. Contractor avatars are often off because they're written from the contractor's assumptions, not customer data. Audit annually.

A critical distinction that's easy to blur. AVATAR = the customer in your head (psychographic + demographic detail used for creative + offer creation). AUDIENCE SETTING in Meta = the targeting parameters (location, plus age, interests and lookalikes where the ad's category allows them). They DON'T match 1:1. Example: your avatar is 'first-time homeowner, 32-38, anxious about hiring contractors, just bought a $500K house, wants validation from someone trustworthy.' But your Meta audience setting might be '[your service area], broad targeting' (Meta no longer offers homeowner targeting, and a housing-category ad can't target age) — much wider than the avatar. Why the gap is intentional: Meta's algorithm performs better with broader targeting + relies on your CREATIVE to filter for the avatar. Narrow targeting kills algorithm performance; broad targeting + avatar-specific creative wins. The avatar guides what you SAY in the ad; the audience setting determines who Meta SHOWS it to. A common mistake is over-narrowing audience settings to match the avatar — wasting algorithm signal. Wide audience + sharp creative = the formula.

Three reality anchors that prevent aspirational drift: (1) USE LAST 90 DAYS of CLOSED customer data — not 'who I imagine I'll attract once we scale.' Pull the names + ZIP codes + project values; that IS your avatar. The mythical 'higher-tier customer' you wish you had is a different exercise (market expansion strategy), not your current avatar; (2) ANCHOR to lead source — if, hypothetically, 80% of closed customers came from referrals and 20% from Meta, your Meta avatar should reflect MORE of who's actually buying through Meta specifically, not your overall mix. The two might look quite different; (3) WRITE the avatar in customer's WORDS, not yours — re-read 10 actual lead-form text submissions; the language in those captures who's actually finding you. Aspirational avatars produce ads that under-convert because they speak to people who aren't shopping. Real avatars produce ads that resonate because they speak to people who ARE shopping. The 90-day audit forces a calibration; do it quarterly.

Same person, different awareness state — your messaging shifts but the underlying avatar is similar. COLD avatar: someone who's been thinking about a problem but hasn't researched solutions yet. They need: (a) education + problem-validation language ('You're not crazy if you noticed your AC running 24/7'); (b) low-commitment CTA (free inspection); (c) heavy trust signals (license, reviews, founder face); (d) wider age range + softer interest targeting. WARM RETARGETING avatar: same person, now actively researching. They've visited your site or watched your video. They need: (a) reduced education (skip the 'why this matters'); (b) higher-commitment CTA (book consultation, get exact quote); (c) social proof shifted to specifics ('see Mike's Houston roof project from last month'); (d) urgency-driven messaging ('financing rates locked in this week'). The avatar doesn't change between cold + warm; the awareness STAGE changes. Match content to where they are in the buyer journey, not to a different person.

1-page brief, not 5-page persona deck. Three sections that production teams actually use: (1) WHO (3-5 sentences) — 'A 45-year-old homeowner in Houston, married with kids, owns a $400K home, has noticed their AC running 24/7 + worried it's about to fail'; (2) WHAT THEY'RE FEELING (3-5 specific phrases customers actually said in interviews) — 'I'm worried about getting overcharged,' 'I called 3 plumbers + only one called back,' 'I don't want my kids in a hot house this summer'; (3) WHAT WOULD MAKE THEM TRUST YOU (3-5 specific elements) — 'License number visible,' 'real customer photos not stock,' 'specific pricing not vague,' 'fast response promise.' One page. Production teams glance + understand within 2 minutes. AVOID: long demographic profiles, fictional 'a day in the life' stories, marketing-personality archetypes ('the practical Pat'). Five-page persona decks tend to get skimmed and ignored; the 1-page customer-voice brief is what actually informs production decisions.