[{"data":1,"prerenderedAt":408},["ShallowReactive",2],{"blog-incrementality-testing-for-ads-zh-CN":3,"blog-ad-banners-zh-CN":380},{"id":4,"title":5,"excerpt":6,"content":7,"coverImage":322,"meta":332,"site":335,"status":356,"slug":357,"author":358,"category":365,"publishDate":374,"featured":81,"updatedAt":375,"createdAt":376,"contentHtml":377,"previewUrl":378,"localeSlugs":379},779,"广告增量测试：衡量真实 lift（2026）","广告增量测试如何超越平台 ROAS：holdout 设计、何时拆分测试够用，以及放量前如何读结果。",{"root":8},{"children":9,"direction":19,"format":15,"indent":13,"type":321,"version":18},[10,22,34,38,42,46,70,104,108,113,117,121,125,129,133,137,141,145,162,183,187,229,233,259,263,277,281,285,289,293,297,301,305,309,313,317],{"children":11,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[12],{"detail":13,"format":13,"mode":14,"style":15,"text":16,"type":17,"version":18},0,"normal","","广告增量测试，衡量「没有广告就不会发生」的提升","text",1,null,"heading","h2",{"children":23,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":18,"textStyle":15},[24,26,28,31],{"detail":13,"format":18,"mode":14,"style":15,"text":25,"type":17,"version":18},"Incrementality testing for ads（广告增量测试）",{"detail":13,"format":13,"mode":14,"style":15,"text":27,"type":17,"version":18}," 用来估计：有多少转化、收入或安装是",{"detail":13,"format":29,"mode":14,"style":15,"text":30,"type":17,"version":18},2,"因为",{"detail":13,"format":13,"mode":14,"style":15,"text":32,"type":17,"version":18},"付费投放才发生的，而不是没有这些广告也会发生。平台回报的转化和末次点击 ROAS 可以很好看，但增量可能是平的——尤其品牌搜索、自然流量或再营销本来就会转化时。","paragraph",{"children":35,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[36],{"detail":13,"format":13,"mode":14,"style":15,"text":37,"type":17,"version":18},"本文帮你选设计（地域对照组、PSA 类、转化提升类或受众分割），读结果时别自欺，并判断何时便宜的站内 A/B 就够用。",{"children":39,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[40],{"detail":13,"format":13,"mode":14,"style":15,"text":41,"type":17,"version":18},"为什么广告增量测试强过虚荣归因",{"children":43,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[44],{"detail":13,"format":13,"mode":14,"style":15,"text":45,"type":17,"version":18},"归因回答「功劳记给谁」。增量回答「因为我们花钱，世界改变了多少」。两者会分叉，当：",{"children":47,"direction":19,"format":15,"indent":13,"type":67,"version":18,"listType":68,"start":18,"tag":69},[48,53,57,62],{"children":49,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":18},[50],{"detail":13,"format":13,"mode":14,"style":15,"text":51,"type":17,"version":18},"拉新与强品牌需求重叠；","listitem",{"children":54,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":29},[55],{"detail":13,"format":13,"mode":14,"style":15,"text":56,"type":17,"version":18},"再营销收割本已在市场中的人；",{"children":58,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":61},[59],{"detail":13,"format":13,"mode":14,"style":15,"text":60,"type":17,"version":18},"多渠道争抢同一笔购买；",3,{"children":63,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":66},[64],{"detail":13,"format":13,"mode":14,"style":15,"text":65,"type":17,"version":18},"隐私与模型抬高了平台转化。",4,"list","bullet","ul",{"children":71,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[72,74,84,86,93,95,102],{"detail":13,"format":13,"mode":14,"style":15,"text":73,"type":17,"version":18},"若只优化 ",{"children":75,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":79,"id":83},[76],{"detail":13,"format":13,"mode":14,"style":15,"text":77,"type":17,"version":18},"ROAS","link",{"linkType":80,"newTab":81,"url":82},"custom",false,"https://deepclick.com/zh-CN/resources/blog/what-is-roas-in-advertising/","6aebf7cbdd368a00c8387d6a",{"detail":13,"format":13,"mode":14,"style":15,"text":85,"type":17,"version":18},"，可能放大「偷走自然转化功劳」的花费。放量前用增量做校验——配合 ",{"children":87,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":90,"id":92},[88],{"detail":13,"format":13,"mode":14,"style":15,"text":89,"type":17,"version":18},"广告花费预算分配",{"linkType":80,"newTab":81,"url":91},"https://deepclick.com/zh-CN/resources/blog/ad-spend-budget-allocation/","6aebf7cbdd368a00c8387d6b",{"detail":13,"format":13,"mode":14,"style":15,"text":94,"type":17,"version":18}," 与 ",{"children":96,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":99,"id":101},[97],{"detail":13,"format":13,"mode":14,"style":15,"text":98,"type":17,"version":18},"盈利放量门槛",{"linkType":80,"newTab":81,"url":100},"https://deepclick.com/zh-CN/resources/blog/scaling-facebook-ads-profitably/","6aebf7cbdd368a00c8387d6c",{"detail":13,"format":13,"mode":14,"style":15,"text":103,"type":17,"version":18},"。",{"children":105,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[106],{"detail":13,"format":13,"mode":14,"style":15,"text":107,"type":17,"version":18},"常见设计（按风险与成熟度选）",{"children":109,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[110],{"detail":13,"format":13,"mode":14,"style":15,"text":111,"type":17,"version":18},"1. 地域 / 市场对照组","h3",{"children":114,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[115],{"detail":13,"format":13,"mode":14,"style":15,"text":116,"type":17,"version":18},"在匹配区域降量或停投，比较人均转化或收入。适合大体量、相似市场；注意溢出（跨区用户、全国品牌效应）。",{"children":118,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[119],{"detail":13,"format":13,"mode":14,"style":15,"text":120,"type":17,"version":18},"2. 受众 / 用户级对照组",{"children":122,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[123],{"detail":13,"format":13,"mode":14,"style":15,"text":124,"type":17,"version":18},"随机切出一刀不看广告（或不看某渠道）。平台若支持真 holdout 更干净；身份不完整时易碎。",{"children":126,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[127],{"detail":13,"format":13,"mode":14,"style":15,"text":128,"type":17,"version":18},"3. PSA / 公益创意对照",{"children":130,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[131],{"detail":13,"format":13,"mode":14,"style":15,"text":132,"type":17,"version":18},"给对照组看不转化创意，让拍卖与频次条件更接近。「直接关广告」会改变谁赢拍卖时有用。",{"children":134,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[135],{"detail":13,"format":13,"mode":14,"style":15,"text":136,"type":17,"version":18},"4. 站内转化 / 品牌提升产品",{"children":138,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[139],{"detail":13,"format":13,"mode":14,"style":15,"text":140,"type":17,"version":18},"厂商 lift 研究方便但黑盒。当作证据之一，别当圣经——尤其卖媒体的是同一家。",{"children":142,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[143],{"detail":13,"format":13,"mode":14,"style":15,"text":144,"type":17,"version":18},"5. 何时拆分测试就够",{"children":146,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[147,149,156,158,160],{"detail":13,"format":13,"mode":14,"style":15,"text":148,"type":17,"version":18},"创意或落地页变化往往先做 ",{"children":150,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":153,"id":155},[151],{"detail":13,"format":13,"mode":14,"style":15,"text":152,"type":17,"version":18},"Facebook 广告拆分测试",{"linkType":80,"newTab":81,"url":154},"https://deepclick.com/zh-CN/resources/blog/facebook-ads-split-testing/","6aebf7cbdd368a00c8387d6d",{"detail":13,"format":13,"mode":14,"style":15,"text":157,"type":17,"version":18},"。它比较的是拍卖",{"detail":13,"format":29,"mode":14,"style":15,"text":159,"type":17,"version":18},"内部",{"detail":13,"format":13,"mode":14,"style":15,"text":161,"type":17,"version":18},"相对赢家，不是真实增量需求。拆分测试用来改进机器；增量测试决定要不要加燃料。",{"children":163,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[164,166,173,175,182],{"detail":13,"format":13,"mode":14,"style":15,"text":165,"type":17,"version":18},"渠道组合也会改变问题：Search 与 Performance Max 的增量问法不同——见 ",{"children":167,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":170,"id":172},[168],{"detail":13,"format":13,"mode":14,"style":15,"text":169,"type":17,"version":18},"PMax vs Search",{"linkType":80,"newTab":81,"url":171},"https://deepclick.com/zh-CN/resources/blog/pmax-vs-search-campaigns/","6aebf7cbdd368a00c8387d6e",{"detail":13,"format":13,"mode":14,"style":15,"text":174,"type":17,"version":18},"。出价卫生仍重要——",{"children":176,"direction":19,"format":15,"indent":13,"type":78,"version":61,"fields":179,"id":181},[177],{"detail":13,"format":13,"mode":14,"style":15,"text":178,"type":17,"version":18},"有官方出处的 Google Ads 技巧",{"linkType":80,"newTab":81,"url":180},"https://deepclick.com/zh-CN/resources/blog/google-ads-tips/","6aebf7cbdd368a00c8387d6f",{"detail":13,"format":13,"mode":14,"style":15,"text":103,"type":17,"version":18},{"children":184,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[185],{"detail":13,"format":13,"mode":14,"style":15,"text":186,"type":17,"version":18},"怎样跑得可信",{"children":188,"direction":19,"format":15,"indent":13,"type":67,"version":18,"listType":227,"start":18,"tag":228},[189,195,201,207,213,220],{"children":190,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":18},[191,193],{"detail":13,"format":18,"mode":14,"style":15,"text":192,"type":17,"version":18},"预先登记决策",{"detail":13,"format":13,"mode":14,"style":15,"text":194,"type":17,"version":18},"：多少 lift % 或增量 CPA 会改预算。",{"children":196,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":29},[197,199],{"detail":13,"format":18,"mode":14,"style":15,"text":198,"type":17,"version":18},"对齐单元",{"detail":13,"format":13,"mode":14,"style":15,"text":200,"type":17,"version":18},"：季节、库存、促销日历在各组一致。",{"children":202,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":61},[203,205],{"detail":13,"format":18,"mode":14,"style":15,"text":204,"type":17,"version":18},"样本量够力",{"detail":13,"format":13,"mode":14,"style":15,"text":206,"type":17,"version":18},"：小账户常常测不出温和 lift；拉长窗口胜过噪音周。",{"children":208,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":66},[209,211],{"detail":13,"format":18,"mode":14,"style":15,"text":210,"type":17,"version":18},"冻结非测试改动",{"detail":13,"format":13,"mode":14,"style":15,"text":212,"type":17,"version":18},"：别同时改版、降价或重写追踪。",{"children":214,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":219},[215,217],{"detail":13,"format":18,"mode":14,"style":15,"text":216,"type":17,"version":18},"主指标 + 护栏",{"detail":13,"format":13,"mode":14,"style":15,"text":218,"type":17,"version":18},"：增量转化，以及扣广告后的贡献、退款与客服负荷。",5,{"children":221,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":226},[222,224],{"detail":13,"format":18,"mode":14,"style":15,"text":223,"type":17,"version":18},"读置信区间",{"detail":13,"format":13,"mode":14,"style":15,"text":225,"type":17,"version":18},"，不要只看一个点估计。",6,"number","ol",{"children":230,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[231],{"detail":13,"format":13,"mode":14,"style":15,"text":232,"type":17,"version":18},"读结果时别骗自己",{"children":234,"direction":19,"format":15,"indent":13,"type":67,"version":18,"listType":68,"start":18,"tag":69},[235,241,247,253],{"children":236,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":18},[237,239],{"detail":13,"format":18,"mode":14,"style":15,"text":238,"type":17,"version":18},"平台 ROAS 好、增量平",{"detail":13,"format":13,"mode":14,"style":15,"text":240,"type":17,"version":18}," → 削减或重塑拉新；收割型渠道说实话。",{"children":242,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":29},[243,245],{"detail":13,"format":18,"mode":14,"style":15,"text":244,"type":17,"version":18},"增量正、ROAS 弱",{"detail":13,"format":13,"mode":14,"style":15,"text":246,"type":17,"version":18}," → 归因低估；别只靠末次点击饿死渠道。",{"children":248,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":61},[249,251],{"detail":13,"format":18,"mode":14,"style":15,"text":250,"type":17,"version":18},"lift 只在品牌词",{"detail":13,"format":13,"mode":14,"style":15,"text":252,"type":17,"version":18}," → 可能在买自己的需求；品牌与冷流量分开看。",{"children":254,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":66},[255,257],{"detail":13,"format":18,"mode":14,"style":15,"text":256,"type":17,"version":18},"周周 lift 不稳",{"detail":13,"format":13,"mode":14,"style":15,"text":258,"type":17,"version":18}," → 功效不足；放大或延长再行动。",{"children":260,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[261],{"detail":13,"format":13,"mode":14,"style":15,"text":262,"type":17,"version":18},"轻量运营节奏",{"children":264,"direction":19,"format":15,"indent":13,"type":67,"version":18,"listType":68,"start":18,"tag":69},[265,269,273],{"children":266,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":18},[267],{"detail":13,"format":13,"mode":14,"style":15,"text":268,"type":17,"version":18},"季度（或大额加预算前）：对最大付费渠道做一次 holdout 类增量检查。",{"children":270,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":29},[271],{"detail":13,"format":13,"mode":14,"style":15,"text":272,"type":17,"version":18},"月度：创意/报价的站内拆分测试。",{"children":274,"direction":19,"format":15,"indent":13,"type":52,"version":18,"value":61},[275],{"detail":13,"format":13,"mode":14,"style":15,"text":276,"type":17,"version":18},"每周：效率与贡献看板——不是每周都新开因果实验。",{"children":278,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[279],{"detail":13,"format":13,"mode":14,"style":15,"text":280,"type":17,"version":18},"小结",{"children":282,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[283],{"detail":13,"format":13,"mode":14,"style":15,"text":284,"type":17,"version":18},"广告增量测试让你少为「本来就会来的转化」付钱。从你能供得起功效的最简单 holdout 开始，预先承诺决策规则，并让拆分测试留在优化工具的位置——而不是因果 lift 的替身。",{"children":286,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":21},[287],{"detail":13,"format":13,"mode":14,"style":15,"text":288,"type":17,"version":18},"常见问题",{"children":290,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[291],{"detail":13,"format":13,"mode":14,"style":15,"text":292,"type":17,"version":18},"什么是广告增量测试？",{"children":294,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[295],{"detail":13,"format":13,"mode":14,"style":15,"text":296,"type":17,"version":18},"一种因果测试：估计付费媒体相对「未触达广告（或安慰剂广告）」对照组带来的提升，而不只是平台归因功劳。",{"children":298,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[299],{"detail":13,"format":13,"mode":14,"style":15,"text":300,"type":17,"version":18},"A/B 测试等于增量测试吗？",{"children":302,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[303],{"detail":13,"format":13,"mode":14,"style":15,"text":304,"type":17,"version":18},"不等于。多数站内 A/B 是在「已经会看到你广告」的人里比版本。增量问的是：相对无广告基线，广告有没有创造出需求。",{"children":306,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[307],{"detail":13,"format":13,"mode":14,"style":15,"text":308,"type":17,"version":18},"增量测试该跑多久？",{"children":310,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[311],{"detail":13,"format":13,"mode":14,"style":15,"text":312,"type":17,"version":18},"要够统计功效，并覆盖完整购买周期——常常数周。因为某一周「看起来好」就提前结束，是常见翻车点。",{"children":314,"direction":19,"format":15,"indent":13,"type":20,"version":18,"tag":112},[315],{"detail":13,"format":13,"mode":14,"style":15,"text":316,"type":17,"version":18},"小广告主值得做吗？",{"children":318,"direction":19,"format":15,"indent":13,"type":33,"version":18,"textFormat":13,"textStyle":15},[319],{"detail":13,"format":13,"mode":14,"style":15,"text":320,"type":17,"version":18},"当付费已是增长的重要一块、且你准备放量时值得。花费很小且噪音大时，先修追踪与创意；因果测试需要量。","root",{"id":323,"alt":324,"updatedAt":325,"createdAt":325,"url":326,"thumbnailURL":19,"filename":327,"mimeType":328,"filesize":329,"width":330,"height":331},1291,"muvloom inspiration r26b-43","2026-10-20T10:06:28.421Z","https://cms-r2.deepclick.com/r26b-43-e552100c0631.jpg","r26b-43-e552100c0631.jpg","image/jpeg",862728,1536,1024,{"title":5,"description":333,"image":334},"用 holdout / PSA 类对照做广告增量测试，定决策规则，放大因果 lift 而非虚荣 ROAS。",{"id":323,"alt":324,"updatedAt":325,"createdAt":325,"url":326,"thumbnailURL":19,"filename":327,"mimeType":328,"filesize":329,"width":330,"height":331},{"id":18,"key":336,"name":337,"prodHost":338,"testHost":339,"blogPath":340,"docPath":341,"zhPrefix":342,"deployHookTest":343,"deployHookProd":344,"brandAuthor":345,"brandKit":349,"enabled":353,"updatedAt":354,"createdAt":355},"deepclick","DeepClick","https://deepclick.com","https://www-test-deepclick.qiliangjia.one","/resources/blog/{slug}","/docs/{slug}","/zh-CN","https://api.cloudflare.com/client/v4/pages/webhooks/deploy_hooks/60a9adef-153b-4c89-8d07-7118e91e9522","https://api.cloudflare.com/client/v4/pages/webhooks/deploy_hooks/05323321-f694-4ce8-a5af-173c507b8bae",{"id":29,"name":337,"avatar":346,"updatedAt":347,"createdAt":348},25,"2026-04-22T08:09:35.299Z","2026-04-22T06:42:49.116Z",{"logoLight":19,"logoDark":19,"primaryColor":350,"accentColor":351,"tagline":352,"wechatName":19,"xhsHandle":19,"wechatQr":19},"#1a73e8","#ff6600","一次点击，多重价值",true,"2026-08-25T06:40:57.663Z","2026-07-14T06:26:38.962Z","published","incrementality-testing-for-ads",{"id":29,"name":337,"avatar":359,"updatedAt":347,"createdAt":348},{"id":346,"alt":337,"updatedAt":360,"createdAt":360,"url":361,"thumbnailURL":19,"filename":362,"mimeType":363,"filesize":364,"width":331,"height":331},"2026-04-22T08:09:22.606Z","https://cms-r2.deepclick.com/头像-白.png","头像-白.png","image/png",26626,{"id":366,"site":367,"titleZh":369,"titleEn":370,"slug":371,"order":219,"updatedAt":372,"createdAt":373},7,{"id":18,"key":336,"name":337,"prodHost":338,"testHost":339,"blogPath":340,"docPath":341,"zhPrefix":342,"deployHookTest":343,"deployHookProd":344,"brandAuthor":29,"brandKit":368,"enabled":353,"updatedAt":354,"createdAt":355},{"logoLight":19,"logoDark":19,"primaryColor":350,"accentColor":351,"tagline":352,"wechatName":19,"xhsHandle":19,"wechatQr":19},"技术导航","Tech Guides","tech-guides","2026-04-27T08:37:10.576Z","2026-04-23T02:59:13.436Z","2026-11-05T03:28:17.945Z","2026-11-05T03:28:19.463Z","2026-11-05T03:28:07.717Z","\u003Cdiv class=\"payload-richtext\">\u003Ch2>广告增量测试，衡量「没有广告就不会发生」的提升\u003C/h2>\u003Cp>\u003Cstrong>Incrementality testing for ads（广告增量测试）\u003C/strong> 用来估计：有多少转化、收入或安装是\u003Cem>因为\u003C/em>付费投放才发生的，而不是没有这些广告也会发生。平台回报的转化和末次点击 ROAS 可以很好看，但增量可能是平的——尤其品牌搜索、自然流量或再营销本来就会转化时。\u003C/p>\u003Cp>本文帮你选设计（地域对照组、PSA 类、转化提升类或受众分割），读结果时别自欺，并判断何时便宜的站内 A/B 就够用。\u003C/p>\u003Ch2>为什么广告增量测试强过虚荣归因\u003C/h2>\u003Cp>归因回答「功劳记给谁」。增量回答「因为我们花钱，世界改变了多少」。两者会分叉，当：\u003C/p>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >拉新与强品牌需求重叠；\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >再营销收割本已在市场中的人；\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >多渠道争抢同一笔购买；\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >隐私与模型抬高了平台转化。\u003C/li>\u003C/ul>\u003Cp>若只优化 \u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/what-is-roas-in-advertising/\">ROAS\u003C/a>，可能放大「偷走自然转化功劳」的花费。放量前用增量做校验——配合 \u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/ad-spend-budget-allocation/\">广告花费预算分配\u003C/a> 与 \u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/scaling-facebook-ads-profitably/\">盈利放量门槛\u003C/a>。\u003C/p>\u003Ch2>常见设计（按风险与成熟度选）\u003C/h2>\u003Ch3>1. 地域 / 市场对照组\u003C/h3>\u003Cp>在匹配区域降量或停投，比较人均转化或收入。适合大体量、相似市场；注意溢出（跨区用户、全国品牌效应）。\u003C/p>\u003Ch3>2. 受众 / 用户级对照组\u003C/h3>\u003Cp>随机切出一刀不看广告（或不看某渠道）。平台若支持真 holdout 更干净；身份不完整时易碎。\u003C/p>\u003Ch3>3. PSA / 公益创意对照\u003C/h3>\u003Cp>给对照组看不转化创意，让拍卖与频次条件更接近。「直接关广告」会改变谁赢拍卖时有用。\u003C/p>\u003Ch3>4. 站内转化 / 品牌提升产品\u003C/h3>\u003Cp>厂商 lift 研究方便但黑盒。当作证据之一，别当圣经——尤其卖媒体的是同一家。\u003C/p>\u003Ch3>5. 何时拆分测试就够\u003C/h3>\u003Cp>创意或落地页变化往往先做 \u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/facebook-ads-split-testing/\">Facebook 广告拆分测试\u003C/a>。它比较的是拍卖\u003Cem>内部\u003C/em>相对赢家，不是真实增量需求。拆分测试用来改进机器；增量测试决定要不要加燃料。\u003C/p>\u003Cp>渠道组合也会改变问题：Search 与 Performance Max 的增量问法不同——见 \u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/pmax-vs-search-campaigns/\">PMax vs Search\u003C/a>。出价卫生仍重要——\u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/google-ads-tips/\">有官方出处的 Google Ads 技巧\u003C/a>。\u003C/p>\u003Ch2>怎样跑得可信\u003C/h2>\u003Col class=\"list-number\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cstrong>预先登记决策\u003C/strong>：多少 lift % 或增量 CPA 会改预算。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cstrong>对齐单元\u003C/strong>：季节、库存、促销日历在各组一致。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cstrong>样本量够力\u003C/strong>：小账户常常测不出温和 lift；拉长窗口胜过噪音周。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cstrong>冻结非测试改动\u003C/strong>：别同时改版、降价或重写追踪。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"5\"\n        >\u003Cstrong>主指标 + 护栏\u003C/strong>：增量转化，以及扣广告后的贡献、退款与客服负荷。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"6\"\n        >\u003Cstrong>读置信区间\u003C/strong>，不要只看一个点估计。\u003C/li>\u003C/ol>\u003Ch2>读结果时别骗自己\u003C/h2>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cstrong>平台 ROAS 好、增量平\u003C/strong> → 削减或重塑拉新；收割型渠道说实话。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cstrong>增量正、ROAS 弱\u003C/strong> → 归因低估；别只靠末次点击饿死渠道。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cstrong>lift 只在品牌词\u003C/strong> → 可能在买自己的需求；品牌与冷流量分开看。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cstrong>周周 lift 不稳\u003C/strong> → 功效不足；放大或延长再行动。\u003C/li>\u003C/ul>\u003Ch2>轻量运营节奏\u003C/h2>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >季度（或大额加预算前）：对最大付费渠道做一次 holdout 类增量检查。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >月度：创意/报价的站内拆分测试。\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >每周：效率与贡献看板——不是每周都新开因果实验。\u003C/li>\u003C/ul>\u003Ch2>小结\u003C/h2>\u003Cp>广告增量测试让你少为「本来就会来的转化」付钱。从你能供得起功效的最简单 holdout 开始，预先承诺决策规则，并让拆分测试留在优化工具的位置——而不是因果 lift 的替身。\u003C/p>\u003Ch2>常见问题\u003C/h2>\u003Ch3>什么是广告增量测试？\u003C/h3>\u003Cp>一种因果测试：估计付费媒体相对「未触达广告（或安慰剂广告）」对照组带来的提升，而不只是平台归因功劳。\u003C/p>\u003Ch3>A/B 测试等于增量测试吗？\u003C/h3>\u003Cp>不等于。多数站内 A/B 是在「已经会看到你广告」的人里比版本。增量问的是：相对无广告基线，广告有没有创造出需求。\u003C/p>\u003Ch3>增量测试该跑多久？\u003C/h3>\u003Cp>要够统计功效，并覆盖完整购买周期——常常数周。因为某一周「看起来好」就提前结束，是常见翻车点。\u003C/p>\u003Ch3>小广告主值得做吗？\u003C/h3>\u003Cp>当付费已是增长的重要一块、且你准备放量时值得。花费很小且噪音大时，先修追踪与创意；因果测试需要量。\u003C/p>\u003C/div>","https://deepclick.com/zh-CN/resources/blog/incrementality-testing-for-ads",{"zh-CN":357,"en":357},[381],{"id":61,"title":382,"site":383,"image":385,"mobileImage":394,"targetUrl":403,"enabled":353,"postDateFrom":404,"postDateTo":405,"updatedAt":406,"createdAt":407},"2026.10.20 Jakarta summit (deepclick)",{"id":18,"key":336,"name":337,"prodHost":338,"testHost":339,"blogPath":340,"docPath":341,"zhPrefix":342,"deployHookTest":343,"deployHookProd":344,"brandAuthor":29,"brandKit":384,"enabled":353,"updatedAt":354,"createdAt":355},{"logoLight":19,"logoDark":19,"primaryColor":350,"accentColor":351,"tagline":352,"wechatName":19,"xhsHandle":19,"wechatQr":19},{"id":386,"alt":387,"updatedAt":388,"createdAt":388,"url":389,"thumbnailURL":19,"filename":390,"mimeType":363,"filesize":391,"width":392,"height":393},1069,"TrafficTalking Jakarta summit 2026.10.20 (zh)","2026-09-04T10:44:19.548Z","https://cms-r2.deepclick.com/traffictalking-jakarta-banner-zh-aa032503f69b.png","traffictalking-jakarta-banner-zh-aa032503f69b.png",1128551,2000,377,{"id":395,"alt":396,"updatedAt":397,"createdAt":397,"url":398,"thumbnailURL":19,"filename":399,"mimeType":363,"filesize":400,"width":401,"height":402},1071,"TrafficTalking Jakarta summit 2026.10.20 (mobile-zh)","2026-09-04T10:44:29.391Z","https://cms-r2.deepclick.com/traffictalking-jakarta-banner-mobile-zh-372fe26c1a2d.png","traffictalking-jakarta-banner-mobile-zh-372fe26c1a2d.png",649961,1200,400,"https://luma.com/g4tnhetc","2026-09-03T00:00:00.000Z","2026-10-04T00:00:00.000Z","2026-09-22T01:42:15.402Z","2026-09-04T10:45:03.031Z",1793933267817]