Buy the answer, not the click.
Paid Social Services — A Field Guide
The algorithm took over targeting. Bidding is a slider now. The only lever still big enough to move the number is the idea inside the ad — so we stopped running paid social as media-buying and started running it as the fastest customer research on earth. Every impression is a test. Every test is a fact about your buyer you can spend again — on the next ad, and on keeping them.
What’s inside
Creative is the whole game now.
Since iOS 14.5 broke precise targeting, the machine decides who sees your ad — and it decides based on the ad. Nielsen puts creative at 56% of what drives sales; everything you used to obsess over — targeting, placements, bidding — splits the rest. The ad isn’t the message anymore. It’s the strategy.
Share of digital sales ROI by lever — Nielsen meta-analysis of digital campaigns; creative quality is the single largest driver.1
Make something worth sending.
Instagram’s own boss says the fastest way to grow is to be shared, not liked — the metric he watches is sends per reach. We’ve always believed the same thing: the best distribution is earned by making content people forward. The trick is doing it on purpose, every time.
“The most important thing you can do is make content people want to send to a friend. A send is worth far more than a like — it’s a real recommendation.”
— Adam Mosseri, Head of Instagram
Sends per reach — shares into DMs ÷ accounts reached — is among the signals Instagram weights most for reaching non-followers, per Head of Instagram Adam Mosseri (2024–2025).2
The feed is a laboratory.
Run paid social as a grid of experiments — each audience against each angle — and the feed tells you which idea actually lands. The winning cell isn’t just a cheaper ad. It’s a fact about your customer you can use everywhere. Press run.
Illustrative scores for teaching; real grids use your creative and conversion data. The method — angle × audience, then act on the winner — is RGM’s core practice.2
One insight, paid five times.
The reason to chase the insight, not the click: the message that wins a cold audience is usually the same one that keeps a customer, earns the repurchase, and gets the referral. Find why people buy and you spend it five times, not once. Drag the slider.
This is also why we refuse to judge paid social on first-purchase ROAS alone. A message that wins the cold click and then powers retention emails, win-back ads, and referral asks is worth multiples of one that only converts once. The insight is the asset; the first sale is just where you find out it works.
Illustrative model: resonance lifts each stage with diminishing transfer downstream; LTV is the compounded result — RGM analysis.3
The creative engine.
Good paid social isn’t a campaign you launch — it’s an engine you run. Concept in, winners out, on repeat. Each round’s winners fund and inform the next, so the system compounds instead of starting from zero every quarter. Watch it turn.
The discipline is in the reads, not the launches. You judge a concept on leading indicators — hook rate and hold rate — long before the conversions finish landing, so you can kill a dud in days instead of weeks. Winners earn more budget and spawn variations; losers free up spend for the next angle. Run that loop for a year and your account is not a pile of ads, it is a compounding library of things you know work on your audience.
RGM’s creative operating loop, from the Paid Social Blueprint: concept → produce native → test at significance → read hook + hold → scale or kill → winners feed the next round.4
Test concepts, not colors.
Most “creative testing” is button shades and headline tweaks that move nothing. The wins come from distinct angles — pain, proof, identity, mechanism — the big swings. Test angle against angle first; polish only inside the winner.
There are only a handful of angles worth fighting over, and they map to why people buy: pain (name the problem), proof (let customers sell), identity (who this is for), and mechanism (why it works). Test those against each other first. Once one wins, express it across formats — UGC video, a static split, a carousel — and only then fuss over headlines and CTAs. Concept decides whether you win; craft decides by how much.
We test concepts, not colors. Order matters: concept, then craft. Tweak-testing feels productive and almost never changes the curve.
“Nobody counts the number of ads you run; they just remember the impression you make.”Bill Bernbach, co-founder, DDB
Budget to significance, not fairness.
Splitting $20 a day evenly across ten ads tests nothing — none gets enough data to read. A concept needs roughly 50 results before you can trust it. Concentrate budget so each angle earns a readable sample. See how many you can really run:
~50 results to judge a concept (cost = 50 × target CPA); concepts/month = budget ÷ that cost. A planning model — RGM analysis, per the Blueprint volume math.5
The three-second test.
Before any learning happens, the ad has to be watched — and almost three in four ecommerce video ads lose the viewer in three seconds. The hook is the gate. A strong one stops 50%+ of scrolls; 30% is the floor. Tap a framework:
73% of ecommerce video ads fail in the first three seconds; hook (3-second view) rate of 50%+ is excellent, 30%+ acceptable; silent-friendly captions lift completion ~40%.6
Steal your own reviews.
The best-performing hook is almost never a copywriter’s clever line. It’s a phrase a real customer already used. Mine your reviews, support tickets, and comments — the winning words are usually already written. Tap a raw review to see the ad it becomes:
We build creative from the customer’s own words — objections and phrases pulled straight from reviews and comments. Resonance we didn’t have to invent.
Attention has a half-life.
Even a winning ad wears out. Show the same person the same thing too often and performance slips — slowly, then off a cliff. That’s not failure; it’s the signal to refresh the concept, not just feed it more budget.
A strong opening, a slow decline as the same people see it again, a cliff once the feeling wears off. Refreshing resets the curve; doing nothing pays more for less.
Two kinds of fatigue, two fixes. Creative-level fatigue is one ad tiring — swap the hook or the cut and the concept keeps earning. Concept-level fatigue is the whole idea going stale, and only a genuinely new angle resets it. Watch the first-time-impression ratio: when it falls, you’re paying to re-reach the already-convinced.
Prospecting fatigue tends to set in above ~2.5–3 impressions per person per 7 days; CTR can fall ~35% and CPC rise ~20% once an ad fatigues.7
The Creative Half-Life calculator.
Three numbers you already know — daily spend, CPM, and the audience the algorithm can really reach — tell you the day your creative tires and how many fresh concepts a month that takes.
Reach via a saturation curve; fatigue = the day average lifetime frequency crosses your threshold. Simplified planning model — RGM analysis.8
Your audience is a pyramid.
You don’t pick interests anymore; you build a stack and feed the machine. It rests on data you own. Tap a tier to see its job:
We score first-party lists by recency and value, not size. Fifty thousand recent, high-LTV buyers seed a better lookalike than two million cold leads. Purer beats bigger.
Formats, not tribes.
Meta, TikTok, YouTube, LinkedIn — rooms with different moods and prices, not religions. Same discipline everywhere; what changes is the room and the cost of entry. Tap one to compare:
Map platforms to the funnel, too. TikTok and Reels are demand-creation engines — cheap, native, top of funnel. Meta is the all-rounder that carries every stage. YouTube bridges discovery and intent. LinkedIn is a bottom-heavy B2B room where one closed deal pays for a lot of impressions. Run the mix where your buyer pays attention and where the unit economics survive the CPM — not where competitors happen to be loud.
2026 CPM ranges: TikTok ~$4–8, Meta ~$8–14, YouTube ~$10–18, LinkedIn ~$20–45; Meta CPM rose ~20% year over year.9
Proving it worked.
Paid social creates demand upstream, so last-click reporting flatters it — the platform claims conversions it merely influenced. A geo holdout settles it: go dark in matched regions and watch what really changes. Drag the real-lift reading:
The gap between reported and real isn’t a reason to stop spending — it’s the number you manage to. Set your true target off the holdout, not the platform’s self-graded ROAS, and budget to that. If going dark barely dents revenue, you were paying for buyers you’d have won anyway; if revenue sags where you went quiet, the lift is real and you scale with confidence. Either way you decide on evidence, not on a dashboard grading its own homework.
Last-click over-credits social because it works upstream; geo holdouts and incrementality tests measure true lift — what would not have happened without the spend.10
TOFU, MOFU, BOFU.
One budget, three jobs. The top creates demand, the middle nurtures it, the bottom captures it — and each stage wants different channels, creative, and metrics. Run them as one system, not three campaigns that ignore each other. Tap a stage:
The classic mistake is running these as three disconnected campaigns — a brand team chasing reach, a performance team chasing ROAS, and nothing connecting them. Treated as one system, TOFU’s job is to refill the warm audiences that BOFU harvests, and BOFU’s ROAS is only as durable as the demand TOFU keeps creating. Starve either end and the whole machine seizes. The planner below shows exactly where yours is leaking.
Stage roles, channels, and metrics — RGM Paid Social Blueprint; figures and placements per 2026 platform practice.12
The Full-Funnel Planner.
Most plans over-fund the top and starve the bottom — reach looks great, revenue doesn’t move. Give it your budget and economics; it splits the spend, forecasts customers, CAC, and profit, and finds the balance where the next dollar pays best.
A planning model — RGM analysis. Reach via cold CPM ÷ frequency; warm prospects via blended create cost; customers = the lesser of demand created and demand you can capture (harvest cost scales with order value).13 Full method on the standalone tool page.
Right unit, right job.
Placements aren’t interchangeable. Reels buy cheap reach but the lowest click intent; Feed costs more but converts; Stories suit urgency; carousels and DPA do the closing. Cast the format to the funnel job. Tap one:
The cheap-CPM trap catches most teams: they pour budget into Reels because the reach looks efficient, then wonder why sales stay flat. Cheap reach in entertainment mode is not the same as cheap customers. The fix is to match the unit to the funnel job — Reels and short video to create demand at the top, Feed and carousels to convert it, Stories to add urgency at the bottom — and to judge each placement by the job it was given, not by one blended number.
Format tendencies (2026): Reels ~15–25% lower CPM, lowest CTR; Feed highest CTR; carousel ~2.08% CTR / 3.7× ROAS / ~$16 CPM; 15s video ~2.31% CTR / 3.6× ROAS, ~34% lower CPA than static.14
Your numbers aren’t everyone’s.
A “good” CPM or CTR only means something against your own industry. Apparel clicks cheap; B2B clicks dear and converts slow. Judge against the right benchmark, then beat it. Tap your category:
Benchmarks are a sanity check, not a scoreboard. A $0.45 click is a triumph in B2B and a disappointment in apparel; an 8× ROAS is routine for a mature ecommerce account and a fantasy for a six-week-old one. The number that matters is your own trend against your own category — pick the right comparison, then beat last quarter. We use these ranges to set expectations before a launch, not to grade a campaign after one.
2026 Meta benchmarks, illustrative ranges: all-industry CPM ~$14, CTR ~1.5%, CPA ~$38; apparel CPC ~$0.45 (lowest), food & bev CPC ~$0.52, B2B CPC ~$2.52; ecommerce ROAS 3–5× (8×+ top).9
The numbers that set the rules.
Six figures we’d staple to every media plan. Learn them and you’ll out-argue most agencies in the room — and spend like the house, not the gambler.
Straight answers.
How much should we spend to start?
How often do we need new creative?
Is targeting dead now that the algorithms run it?
Why does the platform say it worked but revenue didn’t move?
What does paid social cost to run with an agency?
Keep reading.
Learn your audience faster than anyone else, then spend that knowledge across the whole lifecycle. The platforms keep changing the buttons; the discipline doesn’t.
No juniors. Ever.
Every engagement is reviewed by hand — twelve a year. We don’t chase logos; the work chooses us.
Apply for an engagementThe market moved again. Here’s the read.
- Creative = 56% of digital sales ROI: Nielsen meta-analysis of digital campaigns (media ~30%, other ~14%); high-quality creative tested ~12% more effective on Meta. Via RGM Paid Social Blueprint.
- Audience × creative testing as core practice: RGM methodology; matrix scores illustrative for teaching.
- Resonance-to-LTV compounding: an illustrative planning model — a sharp message lifts acquisition and carries, with diminishing transfer, into retention, repurchase, referral, and LTV. RGM analysis.
- The creative engine loop: RGM Paid Social Mastery, Module 4 (Creative Testing Protocol).
- Volume math: ~50 results to read a concept (cost = 50 × target CPA); “budget to significance, not fairness.” RGM Paid Social Mastery, Module 4.
- Three-second hook, hook-rate bands (50%+ excellent / 30%+ acceptable) & captions: RGM Module 4; 2025–26 short-form analyses (Predictive Marketing, Brandwatch, Zeely); 73% of ecommerce video ads fail in first 3s.
- Ad-fatigue thresholds & decay: Shno ad-fatigue statistics (2026); Singular (2025); Search Engine Land via Lionelz (2025). Frequency above ~2.5–3 per 7 days signals prospecting saturation; CTR ~−35% / CPC ~+20% at fatigue (AdEspresso).
- Half-life model: reach via saturation curve R = A(1 − e−I/A); fatigue = day average lifetime frequency crosses threshold. RGM analysis.
- CPM benchmarks (2026): WebFX Meta benchmarks; Improvado social benchmarks; Jonas Agency paid-social & video benchmarks. Meta all-industry CPM ~$14, up ~20% YoY.
- Incrementality & holdouts: last-click over-credits upstream channels; geo holdouts and incrementality tests measure true lift. RGM Attribution & Measurement.
- Delivery signals: Meta Advantage+ 2026 guides (Birch, Adverge, Adligator); Andromeda weighs 10,000+ signals per impression; first-party audience tiers per RGM Module 2 (Audiences in the Cookieless Era).
- Funnel stage roles, channels & metrics: RGM Paid Social Blueprint (TOFU/MOFU/BOFU); placements per 2026 Meta practice.
- Full-Funnel Planner model: RGM analysis. Reach = TOFU budget ÷ (cold CPM/1000 × frequency); warm prospects = (TOFU+MOFU) ÷ blended create cost; customers = min(demand created × warm CVR, BOFU budget ÷ harvest CPA), harvest CPA scales with order value. Illustrative planning model.
- Ad-format tendencies (2026): Dataslayer Meta ad formats; Benly placement performance; AdAmigo creative-format benchmarks. Reels lower CPM/lowest CTR; Feed highest CTR; carousel ~2.08% CTR/3.7× ROAS; 15s video ~2.31% CTR/3.6× ROAS, ~34% lower CPA than static.
- Meta Platforms, Inc. “Meta Reports First Quarter 2026 Results” (29 Apr 2026). Average price per ad +12% YoY; ad impressions +19% YoY; ad revenue $55.02B, +33% YoY. investor.atmeta.com (accessed 6 Jul 2026).
- IAB. “Outlook Study Forecasts 9.5% Growth in U.S. Ad Spend” (Jan 2026). US social media ad spend projected +14.6% in 2026 vs +9.5% for total US ad spend. iab.com (accessed 6 Jul 2026).
- WordStream/LocaliQ. “Facebook Ads Benchmarks 2025” (2025). Average cost per lead $27.66 across industries, up from $22.87 the prior year; traffic-campaign CPC $0.70. wordstream.com (accessed 6 Jul 2026).
- eMarketer. “US Social Network Ad Spending Forecast Report H2 2025” (12 Dec 2025). Advertisers will spend over $121B on US social networks in 2026 — close to 32% of US digital ad spending. emarketer.com (accessed 6 Jul 2026).