<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Pierre Bazoge on Medium]]></title>
        <description><![CDATA[Stories by Pierre Bazoge on Medium]]></description>
        <link>https://medium.com/@pierrebaz?source=rss-e74e20c0e2a7------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*CWE3qO9XoRJCRedY.png</url>
            <title>Stories by Pierre Bazoge on Medium</title>
            <link>https://medium.com/@pierrebaz?source=rss-e74e20c0e2a7------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 19:58:53 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@pierrebaz/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Secrets of the only trustworthy web analytics]]></title>
            <link>https://medium.com/rimdian/secrets-of-the-only-trustworthy-web-analytics-727b65a75f80?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/727b65a75f80</guid>
            <category><![CDATA[cookie-less]]></category>
            <category><![CDATA[google-analytics]]></category>
            <category><![CDATA[customer-data-platform]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[modern-data-stack]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Tue, 07 Mar 2023 22:25:49 GMT</pubDate>
            <atom:updated>2023-03-08T00:01:40.684Z</atom:updated>
            <content:encoded><![CDATA[<p>With the implementations of new regulations (GDPR…), browser restrictions around cookies &amp; cross-device consumer behaviors, can we still trust our Web Analytics?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ie7tdo8k893b6KKNAY4pKA.png" /></figure><blockquote>Long story short: you can’t trust numbers from Google Analytics, <a href="http://segment.com/">Segment.com</a>, Rudderstack… and we will explain why. But solutions exist since the beginning.</blockquote><p>The metrics (KPIs) you are looking at in your web analytics are computed from data. But what happens when the data is missing or incomplete? <strong>What’s the impact of a business decision taken on wrong numbers?</strong></p><p>Let’s list all the issues and see how we can fix them. Some points might be a bit technical, feel free to stick to the conclusions.</p><h3>Cookies lifetime &amp; AdBlockers</h3><p>Since the implementation of <a href="https://webkit.org/blog/9521/intelligent-tracking-prevention-2-3/">Safari ITP</a>, the cookies created by JavaScript trackers (Google Analytics, Facebook pixel…) have a lifetime of 24h for incoming marketing traffic (Adwords…), and 7 days for organic traffic.</p><h4><strong>Impact</strong></h4><p>It’s impossible to properly track/count visitors over time &amp; attribute conversions on iPhones &amp; iPads.</p><p>AdBlockers, which had around <a href="https://www.statista.com/statistics/351862/adblocking-usage/">40% of penetration in 2021</a>, also prevent major JavaScript trackers (GA, Facebook…) to collect data.</p><h4><strong>Impact</strong></h4><p>A big chunk of your users are actually invisible to your web analytics. Your Facebook/Adwords conversions won’t be recorded and you can’t optimize your campaigns.</p><h4><strong>The solution</strong></h4><p>The web analytics data should be collected via <strong>a first-party tracking subdomain</strong> and the cookies lifetime created by JavaScript trackers should be extended server-side to go beyond the 24h.</p><p>Implementing this requires to create a DNS entry that points your tracking subdomain to your web analytics provider.</p><blockquote>Anyone who claims to provide solid stats without touching your DNS is lying to you. “cookieless” technologies are also a lie.</blockquote><p>Example of first-party tracking domain: https://data-collector.your-website.com</p><p>You will also need to push your conversions with server-side integration to the Facebook and Google Ads APIs in order to let them optimize your campaigns, as they became blind because of AdBlockers.</p><h3>Cross-device tracking &amp; user identity reconciliation</h3><p>On an average eCommerce website, 60% of the traffic comes from mobile phones while 60% of conversions happen on desktop computers.</p><p>This demonstrates that most conversion paths trigger many sessions over many devices.</p><p>However, unless the customer is logged on your website on all its devices from the first session, your web analytics will see and count different users.</p><h4><strong>Impact</strong></h4><p>The count of unique users is wrong, and the performance of your marketing channels is only attributed to the last sessions (because the first sessions happened on other devices).</p><h4><strong>Side-effect</strong></h4><p>You will compute your ROI/ROAS only on the “end of the funnel” channels, cutting off the marketing budget that wasn’t tracked and destroying your marketing mix. At this point, your Web Analytics is your worst enemy, as you took very bad decisions because of it.</p><h4><strong>The Solution</strong></h4><p>The JS cookies should survive long enough (thanks to the tracking subdomain explained above) to be able to track the user until he authenticates himself on your website, and let your web analytics merge the previous anonymous activity with the authenticated user profile.</p><p>Merging user identities is the job of a Customer Data Platform, if your Web Analytics is not a CDP you will be stuck.</p><p>You also need to pass the authenticated “User ID” in the URLs of all your emails / SMS communication (post conversion), hoping that they will be opened at some point on the anonymous devices, and allowing your JS tracker to merge those identities.</p><p>Another aspect of doing cross-device reconciliation, is that your multi-touch attribution algorithm should recompute the conversion paths every time an identity/device is merged. It also means that you have to wait for some weeks before having stable stats regarding your campaigns performance.</p><blockquote>Unfortunately, Google Analytics and many others are “append-only” databases (aka Modern Data Stack on Snowflake/BigQuery) that never rewrite the past, and their JS trackers don’t natively detect User IDs passed in URLs for cross-device reconciliation.</blockquote><h3>Session filtering</h3><p>A web analytics will trigger a new session after 30 minutes of inactivity OR if a page is loaded from a different origin (utm_source / utm_medium / utm_campaign).</p><p>Most eCommerce website redirect their customers to a payment gateway that is hosted on a different domain (Paypal, Stripe…), and redirect the customer back to the “confirmation page” when the transaction is complete.</p><p>Redirecting the user to the confirmation page from a payment gateway will trigger a new session, as the web analytics will see a visit coming from a new origin (i.e: www.paypal.com / referral).</p><h4><strong>Impact</strong></h4><p>Triggering a new session will add another touch point in your conversion path, and “dilute” the performance attributed to the other marketing channels. It will also prevent you from knowing what was the last channel that converted, as it will be replaced by the payment gateway domain.</p><h4><strong>The solution</strong></h4><p>Your JS tracker should be able to receive a list of web traffic origins to ignore. Unfortunately, Google Analytics, <a href="http://segment.com/">Segment.com</a> etc… don’t provide such feature.</p><h3>Cross-domain reconciliation</h3><p>Many eCommerce website have their shop hosted at www.my-shop.com and a blog at blog.my-shop.comand some content at magazine.my-shop.com etc…</p><p>Every time a visitor switches from the shop to the blog, and vice-versa, your JS trackers will trigger a new session and count a new visitor, because the cookies are not shared between domains.</p><h4><strong>Impact</strong></h4><p>The count of visitors and sessions is wrong, and the marketing channels attributed performance is “diluted” by the cross-domain sessions.</p><h4><strong>The solution</strong></h4><p>The JS tracker should be able to accept a list of “friends” domains, in order to pass in the URLs the current User ID + Session ID to the other domains, and read those values first when it’s initializing itself instead of counting new visitors and new sessions.</p><h3>Is implementing all of this legal?</h3><p><strong>Yes it is.</strong></p><p>The regulations allow you to collect web traffic activity without user consent in order to produce aggregated anonymous statistics.</p><p>However, as soon as you build a user-centric view of your customers activity you need to gather user consent.</p><p>Technically we can only reconcile authenticated users (with their customer User ID), but the good news is they already accepted the Terms &amp; Conditions of your business upon registration.</p><p>So it’s your responsibility to describe in your Terms &amp; Conditions the purpose of such user-centric data view and request additional consents for marketing communications etc… as usual.</p><blockquote>All the solutions described above are implemented in Rimdian. And we’d like to promote our technology via sharing knowledge.<br><a href="https://www.rimdian.com/">Rimdian.com</a> is the #1 Digital Marketing OS, that brings the real-time algorithms of a Customer Data Platform + the flexibility of a Modern Data Stack with an OS approach through “apps”.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=727b65a75f80" width="1" height="1" alt=""><hr><p><a href="https://medium.com/rimdian/secrets-of-the-only-trustworthy-web-analytics-727b65a75f80">Secrets of the only trustworthy web analytics</a> was originally published in <a href="https://medium.com/rimdian">Rimdian</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Votre attribution marketing est menacée par Google Analytics]]></title>
            <link>https://medium.com/captain-metrics-fr/votre-attribution-marketing-est-menac%C3%A9e-par-google-analytics-fbd0f5a0a869?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/fbd0f5a0a869</guid>
            <category><![CDATA[attribution-modeling]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[tracking]]></category>
            <category><![CDATA[google-analytics]]></category>
            <category><![CDATA[adwords]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Wed, 01 Jul 2020 14:01:15 GMT</pubDate>
            <atom:updated>2020-07-06T12:19:59.438Z</atom:updated>
            <content:encoded><![CDATA[<p>Tout ce que l’on ne vous a jamais dit sur Google Analytics. Et c’est du lourd !</p><blockquote>L’attribution est l’art de calculer l’impact d’une source de trafic, le plus souvent issue d’une campagne marketing, sur les conversions, qui sont le plus souvent des ventes e-commerce.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*S5kfD_yKwHP9vyHunI00eA.png" /></figure><p>Dans un contexte où les utilisateurs utilisent plusieurs terminaux (mobile, laptop, tablette…), à différents endroits et à différents moments de la journée, il devient difficile de pouvoir capter l’ensemble du parcours client sans perte.</p><h3>Google Analytics et la réconciliation cross-device</h3><blockquote><strong>Google Analytics track des terminaux (= “device”) et non des utilisateurs.</strong></blockquote><p>Ce qui signifie que par défaut, il verra 2 visiteurs différents si un même internaute visite votre site avec son mobile et son laptop.</p><p>Pour palier à cette fragmentation des parcours clients, le tracker de Google Analytics accepte un User ID, qui permet d’unifier les sessions d’un internaute identifié (grâce à son User ID) lorsqu’il utilise plusieurs devices.</p><p>Mais, <strong>Google ne modifiera JAMAIS les données collectées dans le passé</strong>, et tiendra compte du User ID uniquement pour les sessions post-identification je cite (<a href="https://support.google.com/analytics/answer/4574780?hl=fr&amp;ref_topic=3123660">source ici</a>) :<em> Avec cette option, les appels collectés avant l’attribution du User ID peuvent être associés à l’identifiant correspondant, pourvu que ces appels aient lieu au cours de la session dans laquelle une valeur d’ID spécifique est attribuée pour la première fois.</em></p><p>Donc GA sera uniquement pertinent pour l’attribution des conversions en rétention mais complètement hors-sujet pour l’acquisition. Et à condition que votre internaute soit identifié sur tous ses devices… ce qui est loin d’être évident !</p><h3>Google Analytics face aux mesures anti-tracking</h3><p>Depuis Octobre 2019 tous <strong>les navigateurs Apple Safari (iPhone, iPad, Mac…) limitent la durée des cookies posés par les trackers JavaScript présents sur votre site web tels que Google Analytics à 24h</strong> pour le trafic issu du marketing (Adwords, FB Ads…) et 7 jours pour le trafic organique.</p><p>Si l’internaute revient sur votre site après 24h il sera comptabilisé comme un nouvel utilisateur et son parcours client sera perdu.</p><p>Seul un sous-domaine de tracking (ex : tracking.votre-site.com) permet de poser des cookies à long-terme, et Google Analytics ne fournit pas cette solution.</p><p>Nous avons par ailleurs lancé le service <a href="https://cookiefreezer.com/">https://cookiefreezer.com</a> pour palier à ce problème.</p><h3>Google Analytics face aux WebViews de Facebook Ads</h3><p>Vous avez surement remarqué que lorsque vous cliquez sur une publicité dans l’univers de Facebook (Messenger, Instagram, Facebook…), celle-ci s’ouvre dans un navigateur embarqué, qui s’appelle une WebView.</p><p>La particularité d’une WebView est quelle ne maintient pas ses cookies et n’a pas accès aux cookies des sessions précédentes.</p><blockquote>Ce qui rend la réconciliation cross-device impossible si l’internaute n’est pas identifié sur votre site.</blockquote><p>Seul Facebook Analytics est en mesure de réconcilier ces sessions WebViews grâce à un paramètre <em>fbclid</em> injecté dans les URLs de votre landing-page lors du clic sur votre publicité.</p><h3>Google Analytics et les modèles d’attribution</h3><p>L’attribution de vos conversions, disponible dans les colonnes “Conversions” de vos rapports d’acquisition, page vues etc… utilise un modèle au “Dernier clic, non-direct”. Il est possible d’utiliser l’outil d’attribution multi-touch complémentaire de Google Analytics pour comparer d’autres modèles.</p><p>Cependant nous vous conseillons de ne jamais utiliser de modèle d’attribution, dont <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-1-9806b8a21d94">nous avons démontré la dangerosité ici</a>. Privilégiez plutôt une approche d’analyse des contributions.</p><h3>Conclusion</h3><p>S’appuyer sur des rapports d’attribution complètement faux, puisqu’une partie significative des parcours clients ont été fragmentés, peut vous amener à prendre de mauvaises décisions d’investissement de vos budgets marketing.</p><p>Il existe de nombreuses méthodes pour réduire la fragmentation des parcours clients, mais cela demanderait un effort considérable à Google Analytics : repenser toute son infrastructure pour permettre à sa base de données de modifier les données du passé afin d’unifier les parcours clients fragmentés, que l’on aurait identifié post-conversion.</p><p>Si vous êtes intéressé à changer de paradigme en déployant une meilleure alternative dans votre système d’information, nous vous invitons à évaluer <a href="https://captainmetrics.com/fr">Captain Metrics</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fbd0f5a0a869" width="1" height="1" alt=""><hr><p><a href="https://medium.com/captain-metrics-fr/votre-attribution-marketing-est-menac%C3%A9e-par-google-analytics-fbd0f5a0a869">Votre attribution marketing est menacée par Google Analytics</a> was originally published in <a href="https://medium.com/captain-metrics-fr">Captain Metrics FR</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Les modèles d’attribution étaient une fausse bonne idée — Partie 3]]></title>
            <link>https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-3-c8a2f1fdaeeb?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8a2f1fdaeeb</guid>
            <category><![CDATA[web-analytics]]></category>
            <category><![CDATA[attribution]]></category>
            <category><![CDATA[google-analytics]]></category>
            <category><![CDATA[adtech]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Thu, 12 Sep 2019 13:44:42 GMT</pubDate>
            <atom:updated>2020-04-21T12:31:01.771Z</atom:updated>
            <content:encoded><![CDATA[<h3>Les modèles d’attribution étaient une fausse bonne idée — Partie 3</h3><p>Cet article s’adresse aux annonceurs désireux d’améliorer les performances de leurs investissements publicitaires via des modèles d’attribution multi-canal.</p><p>Les sujets abordés sont :<br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-1-9806b8a21d94">Partie 1 : L’origine de l’attribution multi-canal</a><br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-2-9059d73fe4e7">Partie 2 : La démonstration de sa dangerosité</a><br>- Partie 3 : Une nouvelle approche</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/636/1*z5hpjprrFkazXqslEXSkKg.png" /></figure><h3>Une nouvelle approche d’attribution multi-canal</h3><p>Dans les parties précédentes, nous avons vu l’origine de l’attribution multi-canal et la dangerosité des modèles d’attributions.</p><p>Cette dernière partie propose une nouvelle approche d’attribution multi-canal sans modèle d’attribution, qui va se concentrer sur les contributions et leurs inter-dépendances.</p><p>Pour bien saisir la teneur de cette approche, nous vous invitons à vider votre tête des anciens concepts des modèles d’attribution, pendant un instant, pour reprendre le problème à zéro.</p><h4>Objectif</h4><blockquote>L’objectif final est la découverte du meilleur mix marketing (canaux, campagnes, créa…).</blockquote><p>Le meilleur mix marketing est celui qui vous procure les meilleurs retours sur investissements pour obtenir des clients fidèles à forte LTV (Life Time Value).</p><p>Pour arriver à obtenir le meilleur mix possible, nous allons devoir analyser les performances de nos investissements marketing, en reconnaissant les contributions ayant menées aux conversions : c’est la phase d’attribution.</p><p>Il faudra ensuite rééquilibrer les budgets publicitaires en fonction des résultats pour optimiser le mix.</p><h4>Changement de point de focus</h4><p>Dans cette nouvelle approche nous considérons que les parcours clients doivent être considérés tels quels, et que les internautes visitent toujours la page qui leur semble la plus pertinente pour avancer dans leur processus d’achat.</p><p>Ce qui signifie qu’<strong>il n’y a pas de contribution plus importante qu’une autre</strong>, mais qu’il y a <strong>des parcours clients plus empruntés que d’autres</strong> et plus rentables aussi.</p><blockquote>Là où les modèles d’attribution multi-canal “découpent” les parcours clients, et mettent en concurrence les canaux pour leur attribuer plus ou moins de poids dans le processus de conversion; notre méthode les considère inviolables et se concentre sur la performance de chaque scénario.</blockquote><p><em>PS: Un scénario, parcours d’achat, parcours client… sont ici des synonymes.</em></p><h4>La phase de découverte</h4><blockquote>La première phase consiste à découvrir un maximum de scénarios possibles, et comment les canaux dépendent les uns des autres.</blockquote><p>Pour cela, il faudra activer un maximum de canaux en parallèle, si possible avec un budget plus large que d’habitude, et tester un maximum de messages (bannières, offres etc…).</p><p>L’objectif est que vos messages publicitaires soient visibles à chaque étape des éventuels processus de conversion des internautes. Pas question d’être à court de budget !</p><p>La phase de découverte devra durer au moins 2x la durée d’un cycle de conversion long.</p><p>Exemple : si certains clients mettent jusqu’à 2 semaines, entre leur première visite et l’achat, alors la phase de découverte devra durer un mois pour recueillir suffisamment de données.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/770/1*oORQVJ0K249qTjePZav4ug.png" /><figcaption>Exemple d’analyse des performances des scénarios</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/769/1*2BSP-uTbXjpX4ie0H-2qfQ.png" /><figcaption>Exemple de matrice de dépendances des canaux</figcaption></figure><h4>Se concentrer sur le principal</h4><blockquote>La seconde phase consiste à optimiser les principaux scénarios de conversion et à identifier les signaux faibles pour optimiser l’ensemble.</blockquote><p>Vouloir satisfaire tous les parcours clients peut complexifier votre mix marketing et devenir chronophage, il est donc recommandé de se concentrer en premier sur les plus grosses sources de conversion.</p><p>Maintenant que vous avez identifié comment les canaux interagissent entre eux et dépendent les uns des autres, nous allons passer à la phase d’optimisation qui devra être exécutée régulièrement.</p><p>L’objectif de l’analyse des signaux faibles, est d’identifier les canaux qui n’ont pas reçu assez de budget et ceux qui en ont reçu trop.</p><p>Lorsqu’un canal reçoit un budget trop conséquent, il va surexposer les internautes à la publicité (notamment dans le display) ou étendre le ciblage pour toucher du trafic moins qualifié.</p><p>La surexposition ne génèrera pas de nouveaux clics, on peut la détecter si CTR est plus bas que la moyenne.</p><p>Le trafic moins qualifié aura un plus fort taux de rebond, une durée moyenne de session plus basse ou un taux de conversion plus faible.</p><p>A l’inverse, un canal qui ne reçoit pas assez de budget aura tendance à être plus performant que les autres: faible taux de bounce, meilleur taux de conversion, meilleur CTR en display.</p><blockquote>Attention, il est possible qu’un canal performant ne convertisse pas bien, s’il a besoin d’un autre canal en aval du parcours client, et que cet autre canal ne reçoit pas assez de budget ou que le message (bannière) n’est pas adapté.</blockquote><p>On peut détecter cette situation en analysant la dépendance d’un canal vis à vis d’un autre, et mesurer l’engagement d’une session (temps passé, nombre de pages vues…) qui n’a poursuivi son parcours client.</p><h4>La cerise sur le gâteau</h4><blockquote>L’analyse de la performance des scénarios, des signaux faibles et du ROI qui en découle devra se faire en Acquisition / Rétention, pour chaque segment de clients et tout autre dimension importante (campagnes, créas…).</blockquote><p>Il sera alors indispensable que votre solution de reporting ne mélange pas les parcours des nouveaux clients (= Acquisition) avec ceux des clients déjà existants (= Rétention).</p><h4>Calcul du ROI</h4><p>Le ROI doit être calculé pour chaque scénario (addition des CPC/CPM etc… de chaque session présente dans le tunnel), et par segment.</p><p>Il est possible que certains parcours client aient un coût d’acquisition trop élevé, mais rapporté à la LTV ce n’est peut être plus vrai.</p><p>Ci-dessous, quelques aperçus d’outils concrets développés pour cette nouvelle approche par <a href="https://captainmetrics.com/">Captain Metrics</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Io22C95dYj_Iev7NideMBQ.png" /><figcaption>Exemple d’analyse de la performance des scénarios — Captain Metrics</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cuf07eOdixpXyfspGTCw3Q.png" /><figcaption>Matrice des dépendances des canaux — Captain Metrics</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JOvXYIRpY1UFXgAZ77I7JA.png" /><figcaption>Revenu concerné et rôles des canaux — Captain Metrics</figcaption></figure><p>Dans l’aperçu ci-dessus, le “Revenu concerné” représente le montant que vous allez perdre si vous coupez le budget de ce canal, puisque les scénarios ayant généré ce chiffre d’affaire ne seront plus correctement alimentés.</p><p>PS : Captain Metrics est une Customer Data Platform, qui permet d’unifier vos profils clients “First-party”, les segmenter, analyser le traffic et attribuer les conversions, afin de pouvoir automatiser vos actions marketing via des intégrations natives (Google Ads, Facebook Ads, Email marketing etc…). N’hésitez pas à nous contacter, nous serons ravis de répondre à vos questions.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8a2f1fdaeeb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-3-c8a2f1fdaeeb">Les modèles d’attribution étaient une fausse bonne idée — Partie 3</a> was originally published in <a href="https://medium.com/captain-metrics-fr">Captain Metrics FR</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Les modèles d’attribution étaient une fausse bonne idée — Partie 2]]></title>
            <link>https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-2-9059d73fe4e7?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/9059d73fe4e7</guid>
            <category><![CDATA[adtech]]></category>
            <category><![CDATA[attribution]]></category>
            <category><![CDATA[google-analytics]]></category>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[web-analytics]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Wed, 11 Sep 2019 16:32:13 GMT</pubDate>
            <atom:updated>2019-09-24T17:06:34.472Z</atom:updated>
            <content:encoded><![CDATA[<h3>Les modèles d’attribution étaient une fausse bonne idée — Partie 2</h3><p>Cet article s’adresse aux annonceurs désireux d’améliorer les performances de leurs investissements publicitaires via des modèles d’attribution multi-canal.</p><p>Les sujets abordés sont :<br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-1-9806b8a21d94">Partie 1 : L’origine de l’attribution multi-canal</a><br>- Partie 2 : La démonstration de sa dangerosité<br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-3-c8a2f1fdaeeb">Partie 3 : Une nouvelle approche</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/636/1*_WVFXYBOzn1b70WLUfRDZg.png" /></figure><h3>Démonstration de la dangerosité des modèles d’attribution multi-canal</h3><p>La logique de l’attribution multi-canal consiste à attribuer une partie du chiffre d’affaire généré par une conversion, à tous les canaux qui y ont contribué.</p><p>La part attribuée à chaque canal est calculée par le modèle d’attribution, qui va appliquer des règles prédéfinies. L’objectif étant d’évaluer l’impact du canal dans le processus de conversion.</p><p>Pour comprendre la dangerosité d’une telle approche, nous allons utiliser le scénario suivant :</p><p>Admettons que 100% des conversions ont un parcours client composé de 2 sessions : Google Adwords puis Retargeting. Et que le modèle attribue 75% à la première session et 25% à la deuxième.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/891/1*PTdNO-KFNpKnzGPnrciJ6Q.png" /><figcaption>Exemple de scénario d’attribution multi-canal</figcaption></figure><p>Puisque seulement 25% des revenus ont été attribués au canal de Retargeting, il est très probable que son <strong>retour sur investissement soit négatif.</strong></p><p>L’annonceur va alors vouloir réduire ou couper son budget de Retargeting pour éviter de perdre plus d’argent, et le résultat sera : <strong>il n’y aura plus aucune conversions</strong>. 0.</p><p><strong>Pourquoi ?</strong></p><p>Parce que les internautes qui avaient besoin du Retargeting pour avancer dans leur processus d’achat ne seront plus ciblés; le budget aura été coupé.</p><p>Les régies pubs ayant horreur du vide, vont afficher des offres concurrentes.</p><blockquote>Le vice des modèles d’attribution multi-canal est qu‘ils ne tiennent pas compte de l’inter-dépendance des canaux les uns vis-avis des autres.</blockquote><p>Alors si l’on ne peut pas attribuer un fragment de chiffre d’affaire à chaque canal ayant contribué à une conversion, quel outil permet de calculer les retours sur investissement publicitaires dans un parcours multi-canal ?</p><p>La réponse dans la partie 3 : <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-3-c8a2f1fdaeeb">une nouvelle approche</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9059d73fe4e7" width="1" height="1" alt=""><hr><p><a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-2-9059d73fe4e7">Les modèles d’attribution étaient une fausse bonne idée — Partie 2</a> was originally published in <a href="https://medium.com/captain-metrics-fr">Captain Metrics FR</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Les modèles d’attribution étaient une fausse bonne idée — Partie 1]]></title>
            <link>https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-1-9806b8a21d94?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/9806b8a21d94</guid>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[advertising]]></category>
            <category><![CDATA[google-analytics]]></category>
            <category><![CDATA[web-analytics]]></category>
            <category><![CDATA[adtech]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Wed, 11 Sep 2019 16:00:36 GMT</pubDate>
            <atom:updated>2019-09-12T13:45:21.514Z</atom:updated>
            <content:encoded><![CDATA[<h3>Les modèles d’attribution étaient une fausse bonne idée — Partie 1</h3><p>Cet article s’adresse aux annonceurs désireux d’améliorer les performances de leurs investissements publicitaires via des modèles d’attribution multi-canal.</p><p>Les sujets abordés sont :<br>- Partie 1 : L’origine de l’attribution multi-canal<br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-2-9059d73fe4e7">Partie 2 : La démonstration de sa dangerosité</a><br>- <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-3-c8a2f1fdaeeb">Partie 3 : Une nouvelle approche</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/636/1*LHpvO_ipj-NKn7lhgdNg5A.png" /></figure><h3>L’origine de l’attribution multi-canal</h3><p>La question primordiale que se pose un annonceur lorsqu’il dépense un budget dans un canal publicitaire est :</p><blockquote>Quel est le retour sur investissement de ce canal ?</blockquote><p>Pour répondre à cette question, les<em> </em>Web Analytics tel que Google Analytics, ont ajouté une colonne <strong>Conversions</strong> dont les stats sont calculées avec un modèle d’attribution “dernier clic, non-direct” : <strong>taux de conversion</strong>, <strong>nombre de conversions</strong> et <strong>revenu généré</strong>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_23I4b5lNdmlFbodeWnhUg.png" /><figcaption>Attribution des conversions — Google Analytics</figcaption></figure><h4>Pourquoi Google Analytics utilise un modèle “Dernier click, non direct” ?</h4><p>Le modèle “dernier clic, non-direct” découle tout simplement du mode de fonctionnement du “tracker JavaScript” de Google Analytics présent sur le site annonceur.</p><p>En effet, tous les “hits” (pages vues, conversions…) transmis par le tracker sont accompagnés des méta-données de “session”. Une session permet de regrouper toutes les actions effectuées par un internaute lors de sa visite sur le site web.</p><p>Une session est initiée si :<br>- c’est la première visite de l’internaute<br>- une session précédente a expiré après 30 minutes d’inactivité<br>- l’internaute atterrit sur le site annonceur via un autre site (= visite non-directe)</p><p>Donc lorsque Google Analytics attribue l’origine des conversions, il s’appuie uniquement sur l’origine (= site web de provenance de l’internaute) de la session courante lorsque la conversion a eu lieu.</p><blockquote>C’est une approche techniquement simple, mais naïve !</blockquote><p>Sauf qu’un internaute averti ne se contente pas d’acheter bêtement le premier produit qu’il trouve… il va d’abord s’informer, puis comparer les offres avant d’acheter (ou d’abandonner son panier), et génère ainsi plusieurs sessions.</p><h4>L’attribution multi-canal, pour enfin y voir plus clair… ou pas !</h4><p>Avec la multiplication des canaux marketing et la complexité croissante des parcours clients, il était important de prendre en compte tous les canaux ayant contribué à une même conversion, et pas uniquement le dernier.</p><p>Voici un exemple de parcours client, où chaque étape va déclencher une nouvelle “session” dans le Web Analytics :</p><ol><li>l’internaute clic sur un produit découvert via un blog spécialisé et arrive sur le site annonceur</li><li>il va comparer les offres via Google Shopping, et revient sur le site annonceur en ayant cliqué sur une offre</li><li>il n’a pas le temps de passer commande, et revient le lendemain sur le site annonceur via une annonce Retargeting de Facebook</li><li>avant de payer, il va chercher un bon de réduction sur un autre site qui va ouvrir une pop-up du site annonceur</li></ol><blockquote>La logique voudrait que l’on attribue une partie du chiffre d’affaire généré par cette vente aux 4 canaux qui y ont contribué. Et c’est une fausse bonne idée.</blockquote><p>Ci-dessous, l’outil d’attribution multi-canal de Google Analytics, qui a été ajouté en plus des rapports d’acquisition au “Dernier clic, non-direct”, dans le menu Conversions &gt; Attribution :</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FCe4imizq9fJaR3Fn1Xa2Q.png" /><figcaption>Attribution multi-canal — Google Analytics</figcaption></figure><p>Rendez-vous sur la deuxième partie de l’article pour la <a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-2-9059d73fe4e7">démonstration de la dangerosité de cet outil</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9806b8a21d94" width="1" height="1" alt=""><hr><p><a href="https://medium.com/captain-metrics-fr/les-mod%C3%A8les-dattribution-%C3%A9taient-une-fausse-bonne-id%C3%A9e-partie-1-9806b8a21d94">Les modèles d’attribution étaient une fausse bonne idée — Partie 1</a> was originally published in <a href="https://medium.com/captain-metrics-fr">Captain Metrics FR</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The secret of successful multiplayer games]]></title>
            <link>https://medium.com/@pierrebaz/the-secret-of-successful-multiplayer-games-a01e4654ac88?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/a01e4654ac88</guid>
            <category><![CDATA[gaming]]></category>
            <category><![CDATA[blizzard]]></category>
            <category><![CDATA[esport]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Fri, 14 Aug 2015 15:20:37 GMT</pubDate>
            <atom:updated>2015-08-14T15:20:37.345Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>The best memories I have about a multiplayer game are from a game called Natural-Selection created in 2002. Ten years later they released Natural-Selection2 and the new version did not take off.<br>Here is why.</blockquote><p>I started playing online in 1997 at the age of 11, with the beta version of Counter-Strike, Quake, Unreal-Tournament, Day of Defeat… only on PC and in various teams.</p><p>So this article will focus on multiplayer games, which the final purpose is esport. It means games that last for many years, are played by a lot of people, have line-ups in all multigaming teams like <a href="http://www.millenium.org/">Millenium</a> and are streamed on <a href="http://www.twitch.tv/directory">Twitch.tv</a>.</p><p>At the end it turns out that ingredients to create a successful game are not far from a successful app/startup.</p><h3>Graphics: less is more</h3><blockquote>Try to play Crysis 3 online, and you get a headache in 5 minutes.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*k7NacikcLyFAGYC8SdlBkw.jpeg" /><figcaption>Crysis3</figcaption></figure><p>Gamers don’t play online to live an immersive experience but to compete and fight against other people, so every graphic effects that won’t help them win should be used carefully. The moves have to be as fluid as possible.</p><p>The first Half-Life mods were almost all successful games (Counter-Strike, Day of Defeat, Team Fortress, Natural Selection…) but their second version haven’t all been as successful as the first one.</p><p>Your ennemies should take their bullets when you want to, there is no room for a FX lag. That’s why pro gamers turn off every shadow effects and needed 5 years to switch to Counter-Strike Source.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*BW1AcnvaSVZLGcBLNb-Fjg.jpeg" /></figure><blockquote>The first who understood this is <a href="http://www.twitch.tv/directory/game/Team%20Fortress%202">Team Fortress 2</a>, instead of adding realism and crazy light effects, they used a cartoon style and only kept the minimum.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*WYV1ecxdaPvQy0jEgiz9gA.jpeg" /></figure><p>This has been prooved by Trackmania over Need for Speed, by <a href="http://www.twitch.tv/directory/game/Borderlands%202$">Borderlands,</a> and now <a href="http://www.twitch.tv/directory/game/Minecraft">Minecraft</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*colue_caaqReYgQG_rD3gw.jpeg" /></figure><p>I already hear you saying “Realistic games like Call of Duty and Battlefield are successful too !”. In fact, not really, they need to release a new game every 2 years with always new stuff in it (cars, weapons…) to keep their active gamers and are not played in esport.</p><p>The last huge fail is Battlefield Hardline, it’s Counter-Strike with the realism of Battlefield ! They surely did not understand the CS lesson…. that leads us to the identity ingredient.</p><h3>The identity</h3><p>The identity is what makes your game different from others, it can be the game play, the number of players on the map, the mechanics, the atmosphere…</p><blockquote>The reason why World of Warcraft is still here, it’s because it’s World of Warcraft !</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*lm76DqfwdxrujQYEMtW5Zw.jpeg" /></figure><p>The reason why Battlefield 4 is not dead yet, it’s because it’s the only one realistic FPS that can handle 64 gamers on the same map without much problems. This is the identity of BF4.</p><p>The reason why Starcraft 2 is not dead yet, it’s because it’s the only one RTS that has 3 races perfectly balanced, so it’s viable for competition.</p><blockquote>The reason Minecraft is successful it’s because what matters is not what you see but what you do.</blockquote><h3>The first takes it all</h3><p>Like every new disruptive ideas, the first on the market takes it all.</p><blockquote>League of Legends made a game out of the Warcraft3 “tower attack” mod DotA and killed DotA2, Heroes of the Storm and Smite even before they were finished.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*cbRFnBPMvSR6CvoDBKGIZg.jpeg" /></figure><p>Another example is how everyone failed to compete with World of Warcraft: Warhammer Online, Guild Wars… even with a strong identity.</p><p>Having fluid experience, a unique identity and being first on the market is not enough to stay on the top. The last ingredient is bringing changes.</p><h3>Little changes, every year</h3><p>To keep your gamers active, they need new things often, most games bring changes by adding maps, but the trend is now by adding almost everything even in FPS.</p><p>The business model of League of Legend is just that, selling you new and cheated characters, then they get nerfed, so you buy the new ones.</p><blockquote>Be careful ! Team Fortress 2 added too much stuff: random items loot, crafting, skins… and killed the game by selling their souls instead of focusing on esport maps.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*6I6qArebrd19SbBM3IeX_Q.jpeg" /></figure><p>This is also the Call of Duty business model, every game brings a lot of new experiences, cars, boats, planes, weapons… now that you have jumped of an orbital station they should wonder what to add !</p><h3>The perfect game ?</h3><p>Blizzard is working on a new game called <a href="http://eu.battle.net/overwatch">Overwatch,</a> which looks like having all ingredients to be an esport star:</p><ul><li>simple and colorful graphics</li><li>a new concept: a mix between League of Legend and Team Fortress 2</li><li>you can expect new maps and characters often because the characters list is already long</li></ul><p>Potential issues:</p><ul><li>the identity of characters is too close to TF2 and LoL</li><li>adding too much characters won’t help to balance the game, which is required to be played in esport</li><li>selling new characters every weeks like LoL makes me furious: if you stop playing for 2 months you don’t understand the game anymore</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*AgBNBFjrfq3P-ZHItO7MoA.jpeg" /></figure><h3>My wishlist</h3><blockquote>Ok now I will ask Santa what I would like to see: World of GTA.</blockquote><p>World of Warcraft is very old, it’s time to launch a serious new MMO.</p><p>GTA 5 is a good solo game but horrible online: many bugs due to latency between players, too long to load, hackers…</p><p>Here is what ingredients should have World of GTA:</p><ul><li>incredibly fluid experience, less effects if you can’t handle it</li><li>lot people in the street, not 5 like today, guys I know this is one of the hardest thing to sync online</li><li>giant new city every 2 years (NYC, Paris, Bali ? …)</li><li>FPS style only, except in cars</li><li>hard to get money, expensive weapons and cars…</li><li>4 teams: cops, business people, gangsta, politics</li><li>doing bad things in the game should be punished very hard, so you get adrenalin when things go wrong</li></ul><p><strong>Special message to the Natural Selection dev team: </strong>Thank you for the first version of the game, if you could please turn the v2 into a cartoon style like TF2 it should help the mechanics of the game and attract gamers. It’s a tactical game, not a RPG.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a01e4654ac88" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why I won’t recommend Meteor anymore]]></title>
            <link>https://medium.com/@pierrebaz/why-i-won-t-recommend-meteor-anymore-fcec4e478d78?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/fcec4e478d78</guid>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[reactjs]]></category>
            <category><![CDATA[meteor]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Tue, 02 Jun 2015 05:55:47 GMT</pubDate>
            <atom:updated>2015-06-08T09:04:37.908Z</atom:updated>
            <content:encoded><![CDATA[<h3>A little bit about Meteor</h3><p><a href="http://meteor.com/">Meteor</a> is a full-stack JS framework (client+server) which is composed of cutting-edge libraries like:</p><ul><li>Blaze: rendering (HTML with Handlebars syntax) &amp; reactive engine (with Tracker) with virtual DOM</li><li>DDP: socket based messaging protocol to make your app talk live with your server in both directions and avoid writing API endpoints</li><li>minimongo: client-side database with a MongoDB API to cache and query data for your views</li></ul><p>It also allows you to write sync NodeJS code on the server-side thanks to Fibers (avoid callback hell) and update your view in live when the database state changes.</p><blockquote>Meteor brings the best UX with minimum efforts, and it’s like magic.</blockquote><h3>The downsides of the magic</h3><p>The easier it is to start with a full-stack framework, the faster you hit a wall when your app becomes more specific.</p><h4>Authentication</h4><p>To authenticate a user on your app, Meteor provides an “Account” package that handles things over DDP, security, sessions revocation, reset password, links etc…</p><p>This lib is well done, but:</p><blockquote>You can’t authenticate a user if the client is not talking over DDP which is really bad when you want to provide API endpoints to your customers.</blockquote><h4>App shutdown</h4><p>There is also no mechanism to do a graceful shutdown of the server.</p><blockquote>If you have transactions running when you restart your server, you are dead.</blockquote><h4>Subscriptions scalability</h4><p>To access your data on the client-side, you have to subscribe to your MongoDB collection. The Meteor server then tails the MongoDB oplog to detect changes and pushes the new data to the client.</p><p>It works, but it’s not clear what would happen if you observe too much collections/fields at the same time and you have a heavy write load.</p><blockquote>How much CPU would it takes to observe thousands of writes per second ?</blockquote><h4>Server-side rendering for SEO</h4><p>Actually server-side rendering is not officially supported, you can use a community package but it’s still a hacky way.</p><h3>Do we need that much magic ?</h3><p>What makes Meteor cool is the way the view updates itself live and you get rid of HTTP roundtrips to receive fresh data.</p><p>But few applications components really need live updates via socket, if you don’t write a messaging app it’s ok to refresh your view every 15 secs via an HTTP roundtrip.</p><p>Meteor is sitting on the fence. When you don’t need live updates, Meteor is overkill. But if you build a messaging app that needs live update, then Meteor is too constrained and has too much overhead, and you need a more specific stack.</p><h3>Similar UX without Meteor</h3><h4>Rendering</h4><p>The Blaze rendering engine can be replaced by <a href="https://facebook.github.io/react/">React,</a> which also is a better way to tie logic with the view, is going native on iOS and can render HTML on server-side for SEO.</p><h4>Live updates</h4><p>Components of your app which really need live updates can use a <a href="https://www.npmjs.com/package/sockjs">Sockjs</a> implementation with a scalable pubsub engine like Kafka, Google Pubsub, AWS SQS etc…</p><h4>Advantages</h4><ul><li>You choose your authentication strategy</li><li>Your server can gracefully restart</li><li>You are not tied to MongoDB</li><li>You can change your frontend or backend technology easily in the future</li></ul><h3>Conclusion</h3><p>Meteor is a great framework for web developpers who want a state-of-the-art UX but don’t care much about data integrity and don’t want to create their own stack.</p><p>But if your app needs to expose API endpoints with authentication, don’t mess running transactions when restarting the server and be more resilient to embrace incoming technologies (like React Native, new databases…), you will be stuck with Meteor.</p><p><em>Special thanks to @ngrilly for typos verification and writting recommendations.</em></p><p><em>Thanks to @arunoda some reactions to this article are discussed in the </em><a href="https://forums.meteor.com/t/why-i-won-t-recommend-meteor-anymore/5285"><em>Meteor forum</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fcec4e478d78" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Lessons learnt from running a social network]]></title>
            <link>https://medium.com/@pierrebaz/lessons-learnt-from-running-a-social-network-76b6153c6adc?source=rss-e74e20c0e2a7------2</link>
            <guid isPermaLink="false">https://medium.com/p/76b6153c6adc</guid>
            <category><![CDATA[social-network]]></category>
            <category><![CDATA[strategy]]></category>
            <category><![CDATA[monetization]]></category>
            <dc:creator><![CDATA[Pierre Bazoge]]></dc:creator>
            <pubDate>Tue, 02 Jun 2015 03:37:58 GMT</pubDate>
            <atom:updated>2015-06-02T03:45:04.459Z</atom:updated>
            <content:encoded><![CDATA[<p>Back in 2006, I was in university and looking for a local airsoft team (similar to paintball) to discover this new sport.</p><p>Unfortunately there was no directory and just a massive and messy web forum, so I decided to create a social network where people could list their teams, contact each others, sell their second-hand stuff etc…</p><p>Few months later (May 2007) I launched <a href="http://www.airsoft-project.com/">airsoft-project.com</a> and got <strong>2000 signups</strong> in a week. At this time there was about 30 000 active players in France.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/889/0*utZgm2tknRpWIlDH.png" /></figure><blockquote>After getting about 1500€/m from advertising, launching worldwide, selling stuff, and having 10 000 members, I had enough and I sold it. Here is my story.</blockquote><h2>Signup or get lost</h2><p>When you launch a social network (or marketplace) you need a critical mass of users quickly to leverage it’s power and make it useful.</p><p>So I choosed to show the content only to signed-in users to force new visitors to register.</p><blockquote>People are nosey, so they signup.</blockquote><h2>Reserve your nickname</h2><p>Airsoft is like gaming, players get a nickname online and offline (on the battlefield).</p><p>However, lots of them use the same ones, like “snake”, “hitman”, “ghost”… they all saw same movies !</p><p>So I choosed to put this punchline on all pages:</p><blockquote>Reserve your nickname now before someone else takes it !</blockquote><p>And guess what ? Even people who didn’t care about my social network, did care about their <em>airsoft-project.com/nickname</em> !</p><h2>Create great content</h2><p>You need content to get SEO, page-views, and monetize your website, here is what worked, and what failed.</p><h3>User content</h3><p>The best content, is the one you don’t make, because we are lazy right ?</p><p><strong>Classifieds ads:</strong></p><blockquote>50% of my traffic was comming from classifieds ads</blockquote><p>When people sell their second-hand stuff, they write keywords, put photos of it, and send messages to make deals, so you get a long-long-long tail SEO and notifications-based traffic from messages.</p><p><strong>Everything else:</strong></p><p>Photos (you host), videos (youtube embed), teams profiles, events… is easy to provide, and keep your visitors on your website.</p><blockquote>If they don’t pay attention, they will spend hours watching photos and videos… and ads ☺</blockquote><h3>Aggregated content</h3><p><strong>News</strong>:</p><blockquote>Because people want to know everything, even if it’s meaningless</blockquote><p>I started to display news from an RSS feed of another news website, to get free content, but one day I thought “<em>Why am I sending my traffic to this website ?</em>”, so I started writing news by myself with my cousin.</p><p>What a fail ! After few weeks we had enough spending hours everyday looking for news to write.</p><blockquote>If you are lazy, share your traffic with aggregated content, it’s better than nothing</blockquote><h3>Your content</h3><p><strong>Products reviews:</strong></p><p>Because of the reputation of the website, brands started sending me their products to get articles. It’s what we call products reviews, you test the product, take photos, and write what you like or… what you like !</p><blockquote>If you want to preserve your relations with the brands don’t write bad things ☺</blockquote><p>Again, after few months I had enough unpacking products for reviews, so I sent them all to a friend, the deal was: you write about the product, you keep the product.</p><blockquote>One day, after an exhibition, I sent him 10 guns for reviews, worth 2000€</blockquote><h2>Making money</h2><p>Few days after the launch, shops started contacting me to put their banners on it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/351/0*k2rllT8ujTjzjK66.jpg" /></figure><p><strong>Sponsorship:</strong></p><p>A big shop wanted to sponsor me in exchange of a banner slot, I was supposed to receive loads of stuff, I accepted. But after few months I received 5% of what was promised, so I stopped. They were not used to send promotion stuff.</p><blockquote>Accept products sponsorship from brands, not shops. Shops have to pay your banners.</blockquote><p><strong>Banners:</strong></p><p>Because main airsoft medias was paper, shops used to pay ads on a monthly base, or for a year.</p><blockquote>In the paper business, the price of a fullpage ad was around 1000€/m. So I sold banners slots 1000€/m.</blockquote><p>The argument was, you pay 1k€ per month for a paper page read 5000 times, you can pay 1k€ per month on a banner displayed 1 000 000 times (one million).</p><p><strong>Online shop:</strong></p><p>One day a guy living not so far phones me to show me his military gear line, he was the exclusive French retailer of an unknown Italian brand, and a very kind man.</p><blockquote>The product line was awesome, but he had no website, so I opened a little shop on my social network, and got 15% of the sales.</blockquote><p>At this time, it was the best move I’ve ever done, 500€ of commissions per month, but it was “luck” !</p><blockquote>I did not open the online shop for money but because I loved his products and wanted to make them available to everyone</blockquote><p>It was luck because when he had enough money I sold him a full online shop and closed my ephemeral one. I was too good, I naively cut off my recurring revenue… ☺</p><blockquote>I had opportunity to run a marketplace where retailers could promote their products, and give me commissions back and I did not see it !</blockquote><p><strong>Shopping engine / comparision:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/400/0*UwnUGZzAHPLl3tvP.jpg" /></figure><p>I finally decided to create a shopping engine with the idea of:</p><blockquote>If people do their shopping through my engine, and I take a commission on every sales, I get rich !</blockquote><p>In 2008, retailers were on a pricing war, and try to convince me to not launch a comparison engine.</p><p>So I built a web crawler to index all the products of their shops, and then started to setup affiliate marketing programs.</p><p>Because they were all small shops, affiliate markeeting platforms refused them all, and I had to build my own tracking platform.</p><p>That’s when I started thinking about Affilae.com, and got bored of running a social network</p><p><strong>I Love blablabla:</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/140/0*Sdy6DcuK9ypjjpXu.jpg" /></figure><p>Tried to sell “I love Airsoft” tee-shirts, but unless you print them yourself, it’s not worth it.</p><h2>Conclusion</h2><p>The worldwide launch was epic, with no ambassadors abroad and no budget, it totally failed !</p><p>The internationalisation layer of the software was really a pain to maintain and after few months I completely switched it back to a French only network.</p><blockquote>If you launch worldwide, try to start focusing a country with ambassadors.</blockquote><p>With time, the fact I was promoting guns became more and more embarassing to me and I didn’t want to resume my life as guns promotion.</p><p>So I sold the social network to a paper media, to launch <a href="https://affilae.com/">Affilae.com</a> : the affiliate marketing platform as-a-service !</p><p>Before I sell it, the traffic was 2000 unique users per month, with more than one million page-views per month and 10 000 registered users.</p><p>To conclude, Airsoft-Project was a giant lab where I tried every monetization methods (plans, affiliate marketing, CPC, CPM…), and discovered what people really want online.</p><blockquote>When you wake up in the morning seeing you got hundreds of signups in the night, it changes your life.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=76b6153c6adc" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>