---
title: "Tuning ads with funnel data"
description: "Ads Manager tells you which ads are cheap. Only the funnel tells you which ads are worth it — and those are different lists more often than not."
canonical_url: https://chatfuel.com/docs/for-agencies/tuning-ads-with-funnel-data
markdown_url: https://chatfuel.com/docs/for-agencies/tuning-ads-with-funnel-data.md
last_updated: 2026-08-12
lang: en
site: https://chatfuel.com/docs
llms_txt: https://chatfuel.com/llms.txt
---

# Tuning ads with funnel data

Ads Manager tells you which ads are cheap. Only the funnel tells you which ads are worth it — and those are different lists more often than not.

Ads Manager tells you which ads are cheap. Only the funnel tells you which ads are *worth it* — and those are different lists more often than not. Tuning ads with funnel data means making every kill, scale, and refresh decision on what happened after the click: the qualification rates, the junk share, the deals that closed. Agencies that run this loop don't just perform better; they can *explain* their decisions to clients, which is half of what a retainer buys.

## The signal: CPQL, not CTR [#the-signal-cpql-not-ctr]

CTR and cost per conversation are early diagnostics — useful for spotting a broken creative, useless for judging a campaign. The decision metric is **[cost per qualified lead](https://chatfuel.com/docs/for-agencies/lead-quality-and-cost-per-qualified-lead)**: spend from Ads Manager divided by the qualified count from Chatfuel. An ad set that Meta's auto-optimization loves for its cheap conversations can be your worst performer per real lead, and vice versa. The rule that follows: never kill or scale on Ads Manager numbers alone; always join them with the funnel first.

## Kill, scale, reallocate [#kill-scale-reallocate]

The playbook seasoned media buyers converge on, with the funnel supplying the verdicts:

* **Reallocate weekly.** Once a week, rank ad sets by CPQL and shift budget toward the cheapest qualified leads. Weekly is the sweet spot — daily reshuffling chases noise, monthly ships a full budget cycle to a known loser.
* **Scale in steps of 20–30%.** Doubling a winner's budget overnight resets Meta's learning and often kills the very performance you're scaling. Raise in increments, let delivery stabilize, repeat.
* **Kill cleanly.** When an ad set's CPQL runs well above its siblings for a full week — or its leads qualify but [never close](https://chatfuel.com/docs/for-agencies/conversation-analytics) — cut it and move the budget the same day. Starving a loser slowly just buys worse results on a longer schedule.
* **Diagnose before you kill.** A suddenly expensive ad set isn't always a bad ad — check the [troubleshooting map](https://chatfuel.com/docs/for-agencies/troubleshooting-ads) first; sometimes the leak is the CTA, the number linked, or the bot, not the creative.

## Creative fatigue: the refresh trigger [#creative-fatigue-the-refresh-trigger]

Every creative wears out — the audience has simply seen it. The signals, in the order they usually appear:

| Signal            | Threshold to act                             |
| ----------------- | -------------------------------------------- |
| Frequency (7-day) | Above \~3 on cold audiences                  |
| CTR               | Down 20–30% from the creative's own baseline |
| CPM               | Creeping up while CTR falls                  |
| Creative age      | 30+ days on meaningful spend                 |

The traditional problem with fatigue was that *knowing* you need fresh creatives didn't make them exist — a designer round-trip per variation meant agencies knowingly ran tired ads for weeks. That constraint is gone: tell the AI Co-Worker what angle to try and [a fresh batch of variants](https://chatfuel.com/docs/for-agencies/ai-generation-of-ad-creatives) is rendered in minutes, grounded in the client's actual offers. When refreshing costs minutes instead of days, you rotate on the data's schedule, not the design queue's.

## The weekly loop [#the-weekly-loop]

The whole system, as a repeatable circuit per client:

1. **Pull the funnel** — *"business report"*, *"show conversion funnel"*, this month's qualified counts and junk breakdown from the [Co-Worker](https://chatfuel.com/docs/for-agencies/reports-via-the-ai-co-worker).
2. **Compute CPQL per ad set** — spend from Ads Manager over qualified from Chatfuel.
3. **Decide in Ads Manager** — reallocate to the CPQL winners, scale in 20–30% steps, kill the confirmed losers.
4. **Refresh what's fatiguing** — new variants from the Co-Worker, informed by what leads actually said this week.

Forty minutes per client, weekly. Each pass makes the next one better — the [copy](https://chatfuel.com/docs/for-agencies/ai-copy-and-hooks-for-the-niche) learns from the objections, the budget concentrates on proven angles, and the client report starts writing itself from the decisions log.

> **Note:** Keep a one-line log of every change. "June 12: killed ad set C, CPQL $31 vs $12 sibling; budget moved to B." A month of these lines is the "what we did and why" section of your client report, pre-written — and the difference between an agency that tweaks and an agency that manages.