The Replenishment Gap

Four frameworks we've been running inside high-spend accounts. A scatter chart that cut creative review time in half. A fix for Meta's age skew problem. The flywheel behind every account that actually scales. And the pipeline health metric most agencies never look at.

This week:

01 — The scatter chart that changed how we read creative

02 — Meta is aging your account — here's the fix

03 — Creative × Data × Capital: the flywheel

04 — Pipeline health vs. performance metrics

05 — Why AI systems win, not AI tools

It's not a table. It's not a ROAS trend line. It's a scatter chart — and once you start reading your account this way, you can't go back.

Plot every ad as a dot. CPA on the Y-axis. Total spend on the X-axis, log scale. Then draw two threshold lines.

The red line is your cut threshold. At $250 in spend you might tolerate a $120 CPA. At $5,000, anything above $58 is a confirmed loser. The confidence interval tightens as spend accumulates. At some point, the math does the work your gut used to do.

The yellow line is your scale trigger. When a dot falls below it, move quickly. Everything between the lines is still learning — don't touch it.

Top-right — High spend, high CPA You aren't cutting fast enough. These are confirmed losers eating budget.

Bottom-left — Low spend, low CPA You found winners and didn't scale them. Opportunity left on the table.

Far right, flat — High spend, consistent CPA This is what a well-run account looks like. A few ads doing the heavy lifting.

Most brands never look at their accounts this way. Most agencies track average outcomes instead of operational signals. The standard Meta UI leaves you to figure out the rest.

We're seeing this across almost every high-spend account right now — especially for brands pushing 100+ ads per week. Meta has heavily indexed toward older demographics, particularly when you're running click attribution.

The result: you think your targeting is broad, but delivery is quietly concentrating in the 45–64 bucket. Higher CPAs. Lower LTV. And you're not even sure when it started.

01 — Export and segment Pull full performance broken out by age and gender. Map CPA and ROAS by segment specifically. Account-level averages hide this. You need segment-level evidence before you touch anything.

02 — Apply value rules — don't restrict Use Meta's value rules to adjust bids by segment: increase bids 15% for ages 25–44, decrease 15% for 55+. You're not locking Meta out of older audiences. You're correcting the anchor point it bids from.

03 — Validate and rebalance Monitor the value rules breakdown at ad set level. Watch delivery shift over 7–14 days. Adjust multipliers based on what the data tells you, not what you assumed going in.

We've seen blended CPA drop from $84 to $51 in accounts where age skew was the primary drag. Not because we changed the creative. Because we changed who was seeing it.

The brands that scale long-term are building repeatable systems around three inputs. Everything else is tactics layered on top.

More spend generates more data. More data enables better creative decisions. Better creative drives better performance. Better performance earns the ability to spend more. Simple in theory. Hard in practice.

It's easy to set the plan proactively. Stopping yourself from making reactive decisions is something only experience can teach you.

Stall A — Great creative, no budget The flywheel stalls at conversion. Winners can't scale because capital isn't there to amplify them. Momentum dies shortly after launch.

Stall B — Big budget, no creative engine Capital is scaling losers. Without a volume-and-velocity creative process, spend just confirms what doesn't work faster — and you end up over-relying on blended MER while your fundamentals erode.

Stall C — Tons of data, no process Most common at 8-figure brands at scale plateau. Dashboards show boilerplate metrics that can't power proper ops decisions.

Most brands in Stall B don't know they're in it. They're spending aggressively. Seeing some results. But the return on every dollar is capped because the creative engine is running at 20% capacity.

They think they need a better media buyer. They need a better creative pipeline.

Before you hire, audit. Before you scale, find the stall.

Most brands track ad performance. The brands that actually scale track creative pipeline health. Those are not the same thing.

Here's the five-stage framework we use to manage every high-spend account:

LAUNCH → Volume enters testing. No judgment yet. TEST → Budget-capped. Signal collection. Avg lifespan: 7 days. HIT → ROAS clears threshold. Creative earns its way forward. [20.4% hit rate] SCALE → Budget opens. Delivery broadens. Avg age at churn: 8–9 days. CHURN → Scale ads die. Always. ~69 per month in high-spend accounts.

20.4% lifetime hit rate 7% hit but never moved to scale 1/14 graduation-to-depletion ratio

That gap between hit rate and graduation rate is opportunity left on the table. In high-spend accounts, creatives graduate to scale at roughly 1/14th the rate the pipeline is being depleted.

The replenishment gap is the silent killer of account performance. Most teams never see it because they're watching performance metrics instead of pipeline health.

This is not a creative quality problem. This is a volume problem. The fix is launching more, consistently, systematically. Every downstream pillar of your account depends on what you feed into Stage 1.

They're building systems where the tools connect.

Most teams use ChatGPT for copy, Claude for analysis, Gemini when they're working out of Google Docs. Each tool is good at its job. But jumping between them with a human copy-pasting every output isn't a system — it's a workflow held together with tabs.

Old workflow: Pull data manually → Paste into AI → Write brief from output → Revision loops → Upload to Meta manually 5 human touchpoints · 5 days

AdSignl: Auto-ingest via Meta API → AI analyzes + patterns → Brief generated → Strategist reviews → Live via API 2 human touchpoints · hours

The humans aren't doing logistics anymore. They're making judgment calls — which is what we should have been doing the whole time.

Other teams have access to the same models we do. What's different is what's behind the recommendations: a pattern library built from $150M in real spend. We're opening access to AdSignl to the public soon.