Accueil›Blog›Test technique Web Analytics : GA4, Tag Management, server-side tracking

Guide recrutement data

Test technique Web Analytics : GA4, Tag Management, server-side tracking

Le web analytics est un metier a part dans la data. GA4 et le tracking server-side ont transforme les pratiques depuis 2023. En entretien, on evalue la capacite a implementer, mesurer et optimiser la collecte de donnees web.

Data Builder·Juin 2025·6 min de lecture·Data Analyst · Analytics Engineer

Sommaire

GA4 vs Universal Analytics
Google Tag Manager
Tracking server-side
Evenements et parametres
Data quality
CRO et A/B testing
Grille

GA4 a change la philosophie du web analytics : exit les sessions, bienvenue les evenements. Le tracking server-side devient le standard pour contourner les bloqueurs et respecter la confidentialite. Ce sont les deux sujets cles en entretien Digital Analyst.

1GA4 vs Universal Analytics : les differences fondamentales

Question discriminante

Quelles sont les 3 differences fondamentales entre GA4 et Universal Analytics ?

	Universal Analytics (fin 2023)	GA4 (actuel)
Modele de donnees	Sessions et pages vues	Evenements et parametres (tout est event)
Identite	Cookie-based, un Device = un User	User ID + Google Signals, cross-device
Integration	Google Ads, Data Studio	BigQuery natif (export gratuit), Looker Studio
Modele d attribution	Last click par defaut	Data-driven par defaut (si volume suffisant)

2Google Tag Manager : architecture et bonnes pratiques

Question discriminante

Comment organisez-vous vos tags, triggers et variables dans GTM pour un site e-commerce ?

Nomenclature — convention de nommage stricte : [type]-[outil]-[action] ex : tag-ga4-purchase, trigger-clic-cta-hero, var-datalayer-product-id
DataLayer — centrale de communication entre le site et GTM. Le developpeur pousse les donnees, GTM les lit
Preview mode — tester chaque modification avant publication. Utiliser GA4 DebugView pour valider les evenements
Versioning — publier avec des messages de version clairs, ne jamais ecraser une version sans la noter

// DataLayer e-commerce - exemple purchase
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3
    }]
  }
});

3Tracking server-side : pourquoi et comment

Question discriminante

Quels sont les avantages du tracking server-side par rapport au client-side ? Quels sont ses inconvenients ?

Avantages — contourne les bloqueurs de publicite et les ITP (Intelligent Tracking Prevention), ameliore les performances (moins de JS cote client), meilleure confidentialite (filtrer les donnees sensibles avant envoi), first-party data
Architecture — GTM Server Container (SGTM) deploye sur App Engine / Cloud Run. Le client envoie vers SGTM, SGTM redistribue vers GA4, Meta, Criteo, etc.
Inconvenients — cout d infrastructure, complexity accrue, necessite des competences techniques plus avancees

4Evenements GA4 : standard vs custom

Evenements auto-collectes — page_view, session_start, first_visit (collectes automatiquement)
Evenements recommandes — purchase, add_to_cart, begin_checkout, login (nommage GA4 standard a respecter)
Evenements custom — tout ce qui est specifique a votre business. Limites : 500 types d evenements, 25 parametres par evenement
Parametres d evenement — donnees contextuelles associees a l evenement (product_id, category, user_segment). Certains doivent etre marques comme 'custom dimensions'

5Data quality : verifier que votre tracking est fiable

GA4 DebugView — voir les evenements en temps reel depuis votre session de navigation
BigQuery export — requeter les evenements bruts pour verifier l absence de doublons, valider les transactions
Monitoring automatise — Looker Studio de surveillance, alertes sur les ecarts de volume
Recette de tracking — plan de recette formel avant toute mise en production d un nouveau tag

6CRO et A/B testing

CRO — Conversion Rate Optimisation : identifier et supprimer les frictions dans le parcours utilisateur
Methode Double Diamond — decouvrir (analytics + UX research) → definir (hypotheses) → developper (variations) → livrer (A/B test)
Significativite statistique — 95% minimum, volume suffisant (plusieurs centaines de conversions par variante)
Outils — AB Tasty, VWO, Google Optimize (arrete), Kameleoon. Analyse dans GA4 ou BigQuery

// DataLayer e-commerce GA4
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3,
      item_category: 'Vetements'
    }]
  }
});

// Validation GA4 DebugView en JavaScript
console.log(window.dataLayer);
// Ou dans Chrome DevTools : Network > filter 'collect' pour voir les hits GA4

Server-side GTM (SGTM) - deployer un container GTM sur Cloud Run ou App Engine. Les hits passent par votre serveur avant d aller vers GA4, Meta, etc. Contourne les bloqueurs, ameliore la confidentialite
Consent Mode v2 - obligatoire en Europe depuis mars 2024. Modeler le comportement des tags selon le consentement cookie. Conversion modeling pour recuperer les donnees perdues
BigQuery export GA4 - gratuit depuis GA4. Requeter les evenements bruts pour valider les transactions, detecter les doublons, construire des entonnoirs custom
Recette de tracking - plan de recette formel avant chaque mise en production d un nouveau tag. Verifier sur staging avec DebugView avant de publier en prod

// DataLayer e-commerce GA4
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3
    }]
  }
});

// Requete BigQuery sur les donnees GA4 brutes
SELECT
    event_name,
    COUNT(*) AS event_count,
    COUNT(DISTINCT user_pseudo_id) AS unique_users
FROM `mon-projet.analytics_123.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20250101' AND '20250131'
  AND event_name IN ('purchase', 'add_to_cart', 'begin_checkout')
GROUP BY 1 ORDER BY 2 DESC;

Server-side GTM (sGTM) - deployer un container GTM sur Cloud Run. Les hits passent par votre serveur. Contourne les bloqueurs, ameliore la confidentialite
Consent Mode v2 - obligatoire en Europe depuis mars 2024. Modeler le comportement des tags selon le consentement. Conversion modeling pour recuperer les donnees perdues
BigQuery export GA4 - gratuit depuis GA4. Requeter les evenements bruts pour valider les transactions, detecter les doublons, construire des entonnoirs custom
Recette de tracking - plan de recette formel avant chaque mise en production. Verifier sur staging avec DebugView avant de publier en prod

7Grille par niveau

Niveau	Maitrise	Signal GO	NO-GO
Junior	GA4 basique, GTM pour les tags standards, events recommandes	Sait creer un tag GA4 dans GTM, comprend le DataLayer	Ne sait pas ce qu est le DataLayer
Confirme	e-commerce GA4, DataLayer avance, BigQuery export, recette tracking	A implemente le e-commerce GA4 complet, a fait une recette de tracking	N a jamais utilise l export BigQuery de GA4
Senior	Tracking server-side SGTM, architecture cross-channel, CRO	A deploye un SGTM, a concu un plan de tagging pour une refonte	Ne sait pas expliquer le server-side tracking

GA4 has changed the philosophy of web analytics: goodbye sessions, welcome events. Server-side tracking is becoming the standard to bypass blockers and respect privacy. These are the two key topics in Digital Analyst interviews.

1GA4 vs Universal Analytics: the fundamental differences

Discriminating question

What are the 3 fundamental differences between GA4 and Universal Analytics?

	Universal Analytics (end 2023)	GA4 (current)
Data model	Sessions and pageviews	Events and parameters (everything is an event)
Identity	Cookie-based, one Device = one User	User ID + Google Signals, cross-device
Integration	Google Ads, Data Studio	Native BigQuery (free export), Looker Studio
Attribution model	Last click by default	Data-driven by default (if sufficient volume)

2Google Tag Manager: architecture and best practices

Discriminating question

How do you organize your tags, triggers and variables in GTM for an e-commerce site?

Naming convention — strict naming convention: [type]-[tool]-[action] e.g.: tag-ga4-purchase, trigger-click-cta-hero, var-datalayer-product-id
DataLayer — communication hub between the site and GTM. The developer pushes the data, GTM reads it
Preview mode — test every change before publishing. Use GA4 DebugView to validate events
Versioning — publish with clear version messages, never overwrite a version without noting it

// DataLayer e-commerce - purchase example
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3
    }]
  }
});

3Server-side tracking: why and how

Discriminating question

What are the advantages of server-side tracking over client-side? What are its disadvantages?

Advantages — bypasses ad blockers and ITP (Intelligent Tracking Prevention), improves performance (less client-side JS), better privacy (filter sensitive data before sending), first-party data
Architecture — GTM Server Container (SGTM) deployed on App Engine / Cloud Run. The client sends to SGTM, SGTM redistributes to GA4, Meta, Criteo, etc.
Disadvantages — infrastructure cost, increased complexity, requires more advanced technical skills

4GA4 Events: standard vs custom

Auto-collected events — page_view, session_start, first_visit (collected automatically)
Recommended events — purchase, add_to_cart, begin_checkout, login (standard GA4 naming to follow)
Custom events — anything specific to your business. Limits: 500 event types, 25 parameters per event
Event parameters — contextual data associated with the event (product_id, category, user_segment). Some must be marked as 'custom dimensions'

5Data quality: verifying your tracking is reliable

GA4 DebugView — see events in real time from your browsing session
BigQuery export — query raw events to check for duplicates, validate transactions
Automated monitoring — Looker Studio dashboard, alerts on volume discrepancies
Tracking test plan — formal test plan before any new tag goes into production

6CRO and A/B testing

CRO — Conversion Rate Optimisation: identify and remove friction points in the user journey
Double Diamond method — discover (analytics + UX research) → define (hypotheses) → develop (variations) → deliver (A/B test)
Statistical significance — 95% minimum, sufficient volume (several hundred conversions per variant)
Tools — AB Tasty, VWO, Google Optimize (discontinued), Kameleoon. Analysis in GA4 or BigQuery

// DataLayer e-commerce GA4
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3,
      item_category: 'Vetements'
    }]
  }
});

// GA4 DebugView validation in JavaScript
console.log(window.dataLayer);
// Or in Chrome DevTools: Network > filter 'collect' to see GA4 hits

Server-side GTM (SGTM) - deploy a GTM container on Cloud Run or App Engine. Hits go through your server before reaching GA4, Meta, etc. Bypasses blockers, improves privacy
Consent Mode v2 - mandatory in Europe since March 2024. Model tag behavior based on cookie consent. Conversion modeling to recover lost data
BigQuery export GA4 - free since GA4. Query raw events to validate transactions, detect duplicates, build custom funnels
Tracking test plan - formal test plan before each new tag goes into production. Verify on staging with DebugView before publishing to prod

// DataLayer e-commerce GA4
window.dataLayer = window.dataLayer || [];
dataLayer.push({
  event: 'purchase',
  ecommerce: {
    transaction_id: 'T_12345',
    value: 89.90,
    currency: 'EUR',
    items: [{
      item_id: 'SKU_001',
      item_name: 'T-shirt bleu',
      price: 29.90,
      quantity: 3
    }]
  }
});

// BigQuery query on raw GA4 data
SELECT
    event_name,
    COUNT(*) AS event_count,
    COUNT(DISTINCT user_pseudo_id) AS unique_users
FROM `mon-projet.analytics_123.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20250101' AND '20250131'
  AND event_name IN ('purchase', 'add_to_cart', 'begin_checkout')
GROUP BY 1 ORDER BY 2 DESC;

Server-side GTM (sGTM) - deploy a GTM container on Cloud Run. Hits go through your server. Bypasses blockers, improves privacy
Consent Mode v2 - mandatory in Europe since March 2024. Model tag behavior based on consent. Conversion modeling to recover lost data
BigQuery export GA4 - free since GA4. Query raw events to validate transactions, detect duplicates, build custom funnels
Tracking test plan - formal test plan before each release to production. Verify on staging with DebugView before publishing to prod

7Level grid

Level	Proficiency	GO signal	NO-GO
Junior	Basic GA4, GTM for standard tags, recommended events	Can create a GA4 tag in GTM, understands the DataLayer	Does not know what the DataLayer is
Mid-level	GA4 e-commerce, advanced DataLayer, BigQuery export, tracking test plan	Has implemented full GA4 e-commerce, has run a tracking test plan	Has never used the GA4 BigQuery export
Senior	Server-side tracking SGTM, cross-channel architecture, CRO	Has deployed a SGTM, has designed a tagging plan for a site redesign	Cannot explain server-side tracking

Vous recrutez un Digital Analyst ou Analytics Engineer ?

Premier entretien gratuit. Rapport GO/NO-GO sous 48h.

Tester gratuitement Reserver un appel