Practical guide · free to read
Before spending, prove the journey works.
A sound campaign is prepared before its first impression. Offer, destination, events, consent, spend and sales follow-up should align. Every later change answers a written diagnosis.
Use after the economics foundation and channel guide. The checklist helps review; it does not certify complete compliance or profitability.
- NeedA person to help
- AdA verifiable promise
- PageA consistent answer
- ActionMeasurable success
- ContributionEconomics that work
Measure without bypassing visitor choice
Advertising trackers generally require prior consent under the framework described by CNIL. Provide clear information, accessible refusal and withdrawal of choice. Test with no choice, refusal and acceptance: affected trackers should not activate as though agreement had been obtained.
An audience-measurement exemption relies on strict conditions; it does not automatically authorize retargeting, advertising matching or platform transmission. Consent mode and server tracking are technical mechanisms, not blanket permission. Never put email, phone, diagnosis or personal free text into URLs or UTM parameters.
Section sources : CNIL: cookies and consent · CNIL: conditions for exempt audience measurement
One real sale, several possible readings
Two platforms may claim one sale after multiple interactions. Do not add their attributed revenue together as business revenue. Compare consistent windows, distinguish click and view where reporting allows, then reconcile unique orders and new customers with sales records.
Write a simple UTM convention: utm_source=meta, utm_medium=paid_social, utm_campaign=audit_octobre, utm_content=demo_a. Keep case and names consistent. For Google Search, for example use google / cpc and respect any auto-tagging. UTM parameters classify visits; they do not recreate missing data or prove causal effect.
Find the blocking step
Diagnose in order. No delivery suggests checking approval, payment, dates, bids and eligibility. Impressions without useful clicks suggest audience, messaging and offer review. Clicks without action suggest testing destination, consistency, speed and measurement. Inquiries without sales suggest qualification and follow-up.
Uncertainty remains high after a few conversions. Avoid declaring a winner from one day or one sale. Record period, maturity, spend and tested change. An automatic stop rule can limit risk, but timing and threshold should account for report and sales delays.
Increase with verified capacity
Before increasing spend, check contribution, volume, retention where relevant, response and delivery capacity. Make a documented change, then observe its consequence. No increase percentage guarantees risk-free growth.
The launch plan
The launch plan
D−3: offer and threshold
Write the brief, costs, target CPA and total ceiling. Day markers illustrate organization, not a requirement.
D−2: page and measurement
Test on a phone, refusal and acceptance of consent, success, value and duplicates.
D−1: campaign review
Review objective, event, spend, dates, geography, networks, creative and links with another person when possible.
D0: monitored activation
Check approval, incoming visits, incidents and actual spend. Fix a failure without waiting for learning.
Review: mature data
On the planned date, examine sales lag, qualification and contribution. Stop, fix or continue with a written reason.
Avoid
See mistakes before paying for them
Fictional examples. Corrections illustrate a method without guaranteeing an outcome.
Avoid
The banner records refusal, but advertising pixels still load.
Why it blocks progress
The interface says one thing while behavior does another. Technical refusal testing is missing.
Build instead
Test requests and events in all three states, then correct activation and choice transmission.
What it changes
The measurement journey respects the applicable choice.
Avoid
Meta claims €100, Google claims €100: report €200 sales for one order.
Why it blocks progress
Attribution models and windows may claim the same sale.
Build instead
One €100 order in the accounts; separate advertising readings and overall contribution review.
What it changes
Management does not artificially create revenue.
Avoid
Change audience, offer, budget, bid and video together after no sales yesterday.
Why it blocks progress
Lag may explain the day, while simultaneous changes make diagnosis unreadable.
Build instead
Check failures, spend and maturity. Then test one documented hypothesis within the planned ceiling.
What it changes
You can link the decision to an observed problem.
Read these indicators together
- Mature data
- State reading date, reporting delays and conversion window.
- Complete funnel
- Impressions, useful visits, inquiries, qualification, appointments, sales and returns. Relevant steps depend on the offer.
- Overall contribution
- Unique sales and contribution compared with total spend. This checks against duplicate attribution.
Put it into practice
Your ad gets clicks, but useful outcomes remain low.
Which check should come before increasing spend?
See the reasoning
First verify destination and measurement success, then intent and inquiry quality. Next compare actual costs with the economic threshold using mature data. Increasing spend on a broken journey amplifies the problem.
Pre-launch checklist
My campaign brief
This brief stays on this page. Download it to retain it. Do not enter customer personal data.
Your brief has been downloaded.Frequently asked questions
When should I stop a campaign immediately?
When an error invalidates the test or exceeds accepted exposure: broken destination, wrong geography, false conversion, spending beyond limits or another identified incident. For poor performance, first examine result maturity and planned criteria.
Can I add ROAS across platforms?
No. Ratios cannot be added, and platforms may attribute the same sale. Calculate an overall view using unique revenue and comparable spend, then retain separate channel readings.
Official sources
Sources checked: 11 October 2026.
VIFLY editorial method. Examples and test plans are recommendations to adapt to your offer. Interfaces and rules can evolve.