[{"data":1,"prerenderedAt":887},["ShallowReactive",2],{"blog-mobile-advertising-attribution-guide-zh-zh-CN":3},{"id":4,"title":5,"excerpt":6,"content":7,"coverImage":850,"meta":860,"status":864,"slug":865,"author":866,"category":876,"publishDate":881,"featured":38,"updatedAt":882,"createdAt":883,"contentHtml":884,"previewUrl":885,"localeSlugs":886},125,"移动广告归因完全指南：原理拆解与 AppsFlyer 实操（2026）","移动广告归因到底怎么跑通？本文拆解确定性与概率归因、AppsFlyer/Adjust 等 MMP 平台、多触点归因模型与 iOS 隐私变化，并附 AppsFlyer 归因链接实操步骤，帮你把每一笔广告预算追到结果。",{"root":8},{"type":9,"format":10,"indent":11,"version":12,"direction":13,"children":14},"root","",0,1,"ltr",[15,23,30,46,50,54,94,98,102,108,114,242,246,250,256,364,372,377,383,449,453,457,469,591,595,599,607,615,623,627,635,686,690,694,736,740,744,748,752,763,767,771,775,779,783,787,791],{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":18},"heading","h2",[19],{"type":20,"text":21,"format":11,"detail":11,"mode":22,"style":10,"version":12},"text","什么是移动广告归因","normal",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":25},"paragraph",[26,28],{"type":20,"text":27,"format":12,"detail":11,"mode":22,"style":10,"version":12},"移动广告归因（mobile advertising attribution）",{"type":20,"text":29,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，是把一次 App 安装或应用内事件，追溯回究竟是哪条广告、哪个 campaign、哪个渠道、哪套素材带来的过程。用户点了广告、随后装了你的 App，归因要回答的就是「这次安装该记在谁头上」——算清这一笔，你才能衡量广告投入产出比、优化出价、砍掉不出转化的渠道。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":31},[32,34,44],{"type":20,"text":33,"format":11,"detail":11,"mode":22,"style":10,"version":12},"没有归因，移动买量就是凭感觉花钱；有了归因，每块预算都能追到结果。本文是一篇「原理拆解」型指南：讲清归因是什么、匹配怎么发生、MMP 扮演什么角色、多触点模型如何分配功劳，以及苹果隐私变化如何重塑整套体系。如果你想要的是各家工具的横向对比排名，请看姊妹篇",{"type":35,"fields":36,"format":10,"indent":11,"version":40,"direction":13,"children":41},"link",{"url":37,"newTab":38,"linkType":39},"https://deepclick.com/zh-CN/resources/blog/best-attribution-software-zh/",false,"custom",3,[42],{"type":20,"text":43,"format":11,"detail":11,"mode":22,"style":10,"version":12},"最佳归因软件盘点",{"type":20,"text":45,"format":11,"detail":11,"mode":22,"style":10,"version":12},"——本文讲「怎么做」，那篇讲「选哪个」。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":47},[48],{"type":20,"text":49,"format":11,"detail":11,"mode":22,"style":10,"version":12},"移动广告归因是如何工作的：基本链路",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":51},[52],{"type":20,"text":53,"format":11,"detail":11,"mode":22,"style":10,"version":12},"无论用哪家平台，归因系统的骨架都一样：",{"type":55,"listType":56,"start":12,"tag":57,"format":10,"indent":11,"version":12,"direction":13,"children":58},"list","number","ol",[59,68,77,85],{"type":60,"value":12,"format":10,"indent":11,"version":12,"direction":13,"children":61},"listitem",[62],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":63},[64,66],{"type":20,"text":65,"format":12,"detail":11,"mode":22,"style":10,"version":12},"记录广告互动。",{"type":20,"text":67,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 用户在某个广告网络（Meta、Google、TikTok、DSP）上点击广告，网络记下点击 ID、时间戳、设备信号和 campaign 元数据。",{"type":60,"value":69,"format":10,"indent":11,"version":12,"direction":13,"children":70},2,[71],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":72},[73,75],{"type":20,"text":74,"format":12,"detail":11,"mode":22,"style":10,"version":12},"记录安装。",{"type":20,"text":76,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 用户完成安装并打开 App，你的测量 SDK 触发首次打开（first-open）事件。",{"type":60,"value":40,"format":10,"indent":11,"version":12,"direction":13,"children":78},[79],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":80},[81,83],{"type":20,"text":82,"format":12,"detail":11,"mode":22,"style":10,"version":12},"尝试匹配。",{"type":20,"text":84,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 归因服务把这次安装信号与近期广告互动池逐一比对，判断该把功劳记给哪一次。",{"type":60,"value":86,"format":10,"indent":11,"version":12,"direction":13,"children":87},4,[88],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":89},[90,92],{"type":20,"text":91,"format":12,"detail":11,"mode":22,"style":10,"version":12},"分配功劳并回传。",{"type":20,"text":93,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 匹配到的来源进入数据看板，并通过 postback 回传给网络的优化引擎。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":95},[96],{"type":20,"text":97,"format":11,"detail":11,"mode":22,"style":10,"version":12},"最关键、也最有讲究的，是第 3 步的「匹配」。匹配本质上有两条完全不同的路。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":99},[100],{"type":20,"text":101,"format":11,"detail":11,"mode":22,"style":10,"version":12},"确定性归因 vs 概率归因",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":103},[104,106],{"type":20,"text":105,"format":12,"detail":11,"mode":22,"style":10,"version":12},"确定性归因（deterministic）",{"type":20,"text":107,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，靠的是广告点击和安装两端共享的同一个唯一、持久标识——历史上是设备 ID（iOS 的 IDFA、Android 的 GAID），或经深度链接透传的点击 ID。两端出现同一 ID，匹配几乎板上钉钉，精确且可审计。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":109},[110,112],{"type":20,"text":111,"format":12,"detail":11,"mode":22,"style":10,"version":12},"概率归因（probabilistic，又叫指纹匹配）",{"type":20,"text":113,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，在没有共享确定性 ID 时使用。它用一组非唯一信号——IP、设备型号、系统版本、时间临近度——加上统计建模「估」出一个匹配。它是有依据的猜测而非确定结论；当隐私限制把信号一层层剥掉，准确率随之下降。",{"type":115,"direction":116,"format":10,"indent":11,"version":12,"colWidths":117,"children":119},"table",null,[118,118,118],280,[120,142,162,182,202,222],{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":122},"tablerow",[123,130,136],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":125},"tablecell",[126],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":127},[128],{"type":20,"text":129,"format":11,"detail":11,"mode":22,"style":10,"version":12},"维度",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":131},[132],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":133},[134],{"type":20,"text":135,"format":11,"detail":11,"mode":22,"style":10,"version":12},"确定性归因",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":137},[138],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":139},[140],{"type":20,"text":141,"format":11,"detail":11,"mode":22,"style":10,"version":12},"概率归因",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":143},[144,150,156],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":145},[146],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":147},[148],{"type":20,"text":149,"format":11,"detail":11,"mode":22,"style":10,"version":12},"匹配依据",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":151},[152],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":153},[154],{"type":20,"text":155,"format":11,"detail":11,"mode":22,"style":10,"version":12},"共享唯一 ID（设备 ID / 点击 ID）",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":157},[158],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":159},[160],{"type":20,"text":161,"format":11,"detail":11,"mode":22,"style":10,"version":12},"统计信号（IP、设备、时间）",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":163},[164,170,176],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":165},[166],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":167},[168],{"type":20,"text":169,"format":11,"detail":11,"mode":22,"style":10,"version":12},"准确率",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":171},[172],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":173},[174],{"type":20,"text":175,"format":11,"detail":11,"mode":22,"style":10,"version":12},"非常高",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":177},[178],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":179},[180],{"type":20,"text":181,"format":11,"detail":11,"mode":22,"style":10,"version":12},"中等，取决于模型",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":183},[184,190,196],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":185},[186],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":187},[188],{"type":20,"text":189,"format":11,"detail":11,"mode":22,"style":10,"version":12},"隐私敏感度",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":191},[192],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":193},[194],{"type":20,"text":195,"format":11,"detail":11,"mode":22,"style":10,"version":12},"较高（用到标识符）",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":197},[198],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":199},[200],{"type":20,"text":201,"format":11,"detail":11,"mode":22,"style":10,"version":12},"较低（无持久 ID）",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":203},[204,210,216],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":205},[206],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":207},[208],{"type":20,"text":209,"format":11,"detail":11,"mode":22,"style":10,"version":12},"iOS（ATT 后）可用性",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":211},[212],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":213},[214],{"type":20,"text":215,"format":11,"detail":11,"mode":22,"style":10,"version":12},"无授权则受限",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":217},[218],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":219},[220],{"type":20,"text":221,"format":11,"detail":11,"mode":22,"style":10,"version":12},"常受苹果政策限制",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":223},[224,230,236],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":225},[226],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":227},[228],{"type":20,"text":229,"format":11,"detail":11,"mode":22,"style":10,"version":12},"适用场景",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":231},[232],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":233},[234],{"type":20,"text":235,"format":11,"detail":11,"mode":22,"style":10,"version":12},"自有渠道、已授权用户",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":237},[238],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":239},[240],{"type":20,"text":241,"format":11,"detail":11,"mode":22,"style":10,"version":12},"ID 不可用时的兜底",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":243},[244],{"type":20,"text":245,"format":11,"detail":11,"mode":22,"style":10,"version":12},"实际落地中，现代归因两者混合：授权与标识符允许时走确定性，缺失时用概率（或隐私保护的聚合方法）兜底。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":247},[248],{"type":20,"text":249,"format":11,"detail":11,"mode":22,"style":10,"version":12},"移动测量合作伙伴（MMP）：中立的裁判",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":251},[252,254],{"type":20,"text":253,"format":12,"detail":11,"mode":22,"style":10,"version":12},"移动测量合作伙伴（mobile measurement partner，MMP）",{"type":20,"text":255,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，是一个独立第三方，跨你所有广告网络汇总归因数据，充当唯一、中立的事实来源。与其相信每个网络自报的数字——每家都会把同一次安装算成自己的——不如让 MMP 统一去重、一致分配功劳。MMP 还会把数据归一化成同一套结构、在 iOS 上托管 SKAdNetwork、运行反作弊，并把 postback 回传给各网络让其算法优化，是任何认真做移动买量的团队的中枢。",{"type":115,"direction":116,"format":10,"indent":11,"version":12,"colWidths":257,"children":259},[258,258],420,[260,274,292,310,328,346],{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":261},[262,268],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":263},[264],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":265},[266],{"type":20,"text":267,"format":11,"detail":11,"mode":22,"style":10,"version":12},"MMP 平台",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":269},[270],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":271},[272],{"type":20,"text":273,"format":11,"detail":11,"mode":22,"style":10,"version":12},"突出优势",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":275},[276,286],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":277},[278],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":279},[280],{"type":35,"fields":281,"format":10,"indent":11,"version":40,"direction":13,"children":283},{"url":282,"newTab":38,"linkType":39},"https://www.appsflyer.com/",[284],{"type":20,"text":285,"format":11,"detail":11,"mode":22,"style":10,"version":12},"AppsFlyer",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":287},[288],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":289},[290],{"type":20,"text":291,"format":11,"detail":11,"mode":22,"style":10,"version":12},"集成网络目录最大；看板深入；反作弊工具强",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":293},[294,304],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":295},[296],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":297},[298],{"type":35,"fields":299,"format":10,"indent":11,"version":40,"direction":13,"children":301},{"url":300,"newTab":38,"linkType":39},"https://www.adjust.com/",[302],{"type":20,"text":303,"format":11,"detail":11,"mode":22,"style":10,"version":12},"Adjust",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":305},[306],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":307},[308],{"type":20,"text":309,"format":11,"detail":11,"mode":22,"style":10,"version":12},"数据干净、自动化、反作弊；已并入 AppLovin",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":311},[312,322],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":313},[314],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":315},[316],{"type":35,"fields":317,"format":10,"indent":11,"version":40,"direction":13,"children":319},{"url":318,"newTab":38,"linkType":39},"https://www.branch.io/",[320],{"type":20,"text":321,"format":11,"detail":11,"mode":22,"style":10,"version":12},"Branch",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":323},[324],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":325},[326],{"type":20,"text":327,"format":11,"detail":11,"mode":22,"style":10,"version":12},"深度链接与 web-to-app 起家；跨平台旅程强",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":329},[330,340],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":331},[332],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":333},[334],{"type":35,"fields":335,"format":10,"indent":11,"version":40,"direction":13,"children":337},{"url":336,"newTab":38,"linkType":39},"https://www.kochava.com/",[338],{"type":20,"text":339,"format":11,"detail":11,"mode":22,"style":10,"version":12},"Kochava",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":341},[342],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":343},[344],{"type":20,"text":345,"format":11,"detail":11,"mode":22,"style":10,"version":12},"灵活、隐私优先、数据管控可配置",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":347},[348,358],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":349},[350],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":351},[352],{"type":35,"fields":353,"format":10,"indent":11,"version":40,"direction":13,"children":355},{"url":354,"newTab":38,"linkType":39},"https://www.singular.net/",[356],{"type":20,"text":357,"format":11,"detail":11,"mode":22,"style":10,"version":12},"Singular",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":359},[360],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":361},[362],{"type":20,"text":363,"format":11,"detail":11,"mode":22,"style":10,"version":12},"营销分析 + 归因；成本/ROI 聚合见长",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":365},[366,368,370],{"type":20,"text":367,"format":11,"detail":11,"mode":22,"style":10,"version":12},"DeepClick ",{"type":20,"text":369,"format":12,"detail":11,"mode":22,"style":10,"version":12},"不是",{"type":20,"text":371,"format":11,"detail":11,"mode":22,"style":10,"version":12}," MMP，也不与上述平台竞争。它处在更靠前的一层——流量与安装基建侧——让流入你 MMP 的数据更干净。下文展开。",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":374},"h3",[375],{"type":20,"text":376,"format":11,"detail":11,"mode":22,"style":10,"version":12},"如何配置 AppsFlyer 归因链接（实操）",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":378},[379,381],{"type":20,"text":380,"format":12,"detail":11,"mode":22,"style":10,"version":12},"AppsFlyer 归因链接（attribution link，又称 OneLink）",{"type":20,"text":382,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，就是你交给广告网络、用来给安装记功的追踪 URL。一套可落地的配置：",{"type":55,"listType":56,"start":12,"tag":57,"format":10,"indent":11,"version":12,"direction":13,"children":384},[385,393,401,426,434],{"type":60,"value":12,"format":10,"indent":11,"version":12,"direction":13,"children":386},[387],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":388},[389,391],{"type":20,"text":390,"format":12,"detail":11,"mode":22,"style":10,"version":12},"接入 SDK。",{"type":20,"text":392,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 把 AppsFlyer SDK 加进 iOS 和 Android 包并配好 dev key，让 App 能触发首次打开与应用内事件。",{"type":60,"value":69,"format":10,"indent":11,"version":12,"direction":13,"children":394},[395],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":396},[397,399],{"type":20,"text":398,"format":12,"detail":11,"mode":22,"style":10,"version":12},"注册广告网络为 partner。",{"type":20,"text":400,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 在后台启用对应的集成 partner，让 postback 自动配好。",{"type":60,"value":40,"format":10,"indent":11,"version":12,"direction":13,"children":402},[403],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":404},[405,407,409,412,414,416,418,420,422,424],{"type":20,"text":406,"format":12,"detail":11,"mode":22,"style":10,"version":12},"生成归因链接。",{"type":20,"text":408,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 挂上 ",{"type":20,"text":410,"format":411,"detail":11,"mode":22,"style":10,"version":12},"pid",16,{"type":20,"text":413,"format":11,"detail":11,"mode":22,"style":10,"version":12},"（媒体来源）、",{"type":20,"text":415,"format":411,"detail":11,"mode":22,"style":10,"version":12},"c",{"type":20,"text":417,"format":11,"detail":11,"mode":22,"style":10,"version":12},"（campaign）、",{"type":20,"text":419,"format":411,"detail":11,"mode":22,"style":10,"version":12},"af_adset",{"type":20,"text":421,"format":11,"detail":11,"mode":22,"style":10,"version":12},"、",{"type":20,"text":423,"format":411,"detail":11,"mode":22,"style":10,"version":12},"af_ad",{"type":20,"text":425,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 等参数，让报表能按来源、campaign、广告组、素材拆分。",{"type":60,"value":86,"format":10,"indent":11,"version":12,"direction":13,"children":427},[428],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":429},[430,432],{"type":20,"text":431,"format":12,"detail":11,"mode":22,"style":10,"version":12},"加深度链接跳转。",{"type":20,"text":433,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 用 OneLink 让同一条 URL 跳转到 App Store、Google Play，或在已安装时直接进入某页面，减少流失。",{"type":60,"value":435,"format":10,"indent":11,"version":12,"direction":13,"children":436},5,[437],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":438},[439,441,443,444,446,447],{"type":20,"text":440,"format":12,"detail":11,"mode":22,"style":10,"version":12},"放量前 QA。",{"type":20,"text":442,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 跑一次测试安装，在原始数据里确认它被正确归到对应的 ",{"type":20,"text":410,"format":411,"detail":11,"mode":22,"style":10,"version":12},{"type":20,"text":445,"format":11,"detail":11,"mode":22,"style":10,"version":12},"/",{"type":20,"text":415,"format":411,"detail":11,"mode":22,"style":10,"version":12},{"type":20,"text":448,"format":11,"detail":11,"mode":22,"style":10,"version":12},"。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":450},[451],{"type":20,"text":452,"format":11,"detail":11,"mode":22,"style":10,"version":12},"而以上每一次匹配的质量，都取决于链接本身控制不了的一件事：打到它上面的流量是不是真的。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":454},[455],{"type":20,"text":456,"format":11,"detail":11,"mode":22,"style":10,"version":12},"多触点归因：功劳到底算谁的",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":458},[459,461,463,465,467],{"type":20,"text":460,"format":11,"detail":11,"mode":22,"style":10,"version":12},"现实中的转化路径往往不止一次触达——用户可能先看了视频广告，过几天点了搜索广告，然后才安装。",{"type":20,"text":462,"format":12,"detail":11,"mode":22,"style":10,"version":12},"多触点归因（multi-touch attribution）",{"type":20,"text":464,"format":11,"detail":11,"mode":22,"style":10,"version":12},"决定这些触点之间如何分配功劳，一个",{"type":20,"text":466,"format":12,"detail":11,"mode":22,"style":10,"version":12},"多触点归因平台（multi touch attribution platform）",{"type":20,"text":468,"format":11,"detail":11,"mode":22,"style":10,"version":12},"会套用模型，而不是一股脑把功劳全记给最后一次点击。",{"type":115,"direction":116,"format":10,"indent":11,"version":12,"colWidths":470,"children":471},[118,118,118],[472,491,511,531,551,571],{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":473},[474,480,486],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":475},[476],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":477},[478],{"type":20,"text":479,"format":11,"detail":11,"mode":22,"style":10,"version":12},"模型",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":481},[482],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":483},[484],{"type":20,"text":485,"format":11,"detail":11,"mode":22,"style":10,"version":12},"功劳如何分配",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":12,"children":487},[488],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":489},[490],{"type":20,"text":229,"format":11,"detail":11,"mode":22,"style":10,"version":12},{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":492},[493,499,505],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":494},[495],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":496},[497],{"type":20,"text":498,"format":11,"detail":11,"mode":22,"style":10,"version":12},"末次点击",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":500},[501],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":502},[503],{"type":20,"text":504,"format":11,"detail":11,"mode":22,"style":10,"version":12},"100% 给安装前最后一次触点",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":506},[507],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":508},[509],{"type":20,"text":510,"format":11,"detail":11,"mode":22,"style":10,"version":12},"简单、效果直投",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":512},[513,519,525],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":514},[515],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":516},[517],{"type":20,"text":518,"format":11,"detail":11,"mode":22,"style":10,"version":12},"首次点击",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":520},[521],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":522},[523],{"type":20,"text":524,"format":11,"detail":11,"mode":22,"style":10,"version":12},"100% 给第一次触点",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":526},[527],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":528},[529],{"type":20,"text":530,"format":11,"detail":11,"mode":22,"style":10,"version":12},"衡量漏斗顶端的发现",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":532},[533,539,545],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":534},[535],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":536},[537],{"type":20,"text":538,"format":11,"detail":11,"mode":22,"style":10,"version":12},"线性",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":540},[541],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":542},[543],{"type":20,"text":544,"format":11,"detail":11,"mode":22,"style":10,"version":12},"所有触点均分",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":546},[547],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":548},[549],{"type":20,"text":550,"format":11,"detail":11,"mode":22,"style":10,"version":12},"每次交互同等重要",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":552},[553,559,565],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":554},[555],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":556},[557],{"type":20,"text":558,"format":11,"detail":11,"mode":22,"style":10,"version":12},"时间衰减",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":560},[561],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":562},[563],{"type":20,"text":564,"format":11,"detail":11,"mode":22,"style":10,"version":12},"越靠近转化的触点分得越多",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":566},[567],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":568},[569],{"type":20,"text":570,"format":11,"detail":11,"mode":22,"style":10,"version":12},"短、强意图的周期",{"type":121,"direction":116,"format":10,"indent":11,"version":12,"children":572},[573,579,585],{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":574},[575],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":576},[577],{"type":20,"text":578,"format":11,"detail":11,"mode":22,"style":10,"version":12},"数据驱动",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":580},[581],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":582},[583],{"type":20,"text":584,"format":11,"detail":11,"mode":22,"style":10,"version":12},"算法按实际贡献分配",{"type":124,"direction":116,"format":10,"indent":11,"version":12,"backgroundColor":116,"colSpan":12,"rowSpan":12,"headerState":11,"children":586},[587],{"type":24,"direction":116,"format":10,"indent":11,"version":12,"textFormat":11,"textStyle":10,"children":588},[589],{"type":20,"text":590,"format":11,"detail":11,"mode":22,"style":10,"version":12},"大体量、多渠道项目",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":592},[593],{"type":20,"text":594,"format":11,"detail":11,"mode":22,"style":10,"version":12},"末次点击仍是移动端默认，因为 App 安装窗口短；但当预算大、渠道相互重叠时，数据驱动模型能给出更公平的图景。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":596},[597],{"type":20,"text":598,"format":11,"detail":11,"mode":22,"style":10,"version":12},"iOS 隐私冲击：ATT 与 SKAdNetwork",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":600},[601,603,605],{"type":20,"text":602,"format":11,"detail":11,"mode":22,"style":10,"version":12},"苹果的 ",{"type":20,"text":604,"format":12,"detail":11,"mode":22,"style":10,"version":12},"App 跟踪透明度（ATT）",{"type":20,"text":606,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 框架，要求 App 在访问 IDFA 前必须征得用户同意。大多数用户会拒绝，这就抽走了支撑 iOS 上经典确定性归因的那个确定性标识符。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":608},[609,611,613],{"type":20,"text":610,"format":11,"detail":11,"mode":22,"style":10,"version":12},"为填补空缺，苹果提供 ",{"type":20,"text":612,"format":12,"detail":11,"mode":22,"style":10,"version":12},"SKAdNetwork（SKAN）",{"type":20,"text":614,"format":11,"detail":11,"mode":22,"style":10,"version":12},"——一个隐私保护框架：由苹果自己校验安装，并通过「转化值（conversion value）」回传聚合、延迟的转化数据。代价是实打实的：数据是聚合的（非用户级）、被随机计时器延迟、粒度有限。较新版本（SKAN 4）引入多次 postback 和粗粒度数值，苹果的 AdAttributionKit 又进一步扩展了该模型。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":616},[617,619,621],{"type":20,"text":618,"format":11,"detail":11,"mode":22,"style":10,"version":12},"实际后果是：在 iOS 上你要把 SKAN、已授权的确定性数据和建模估算混着用，并接受更多噪声。这让",{"type":20,"text":620,"format":12,"detail":11,"mode":22,"style":10,"version":12},"流量质量与干净的安装链路比以往更重要",{"type":20,"text":622,"format":11,"detail":11,"mode":22,"style":10,"version":12},"——当信号本就聚合且稀缺，每一份作弊或浪费的曝光，都会污染那本就不多的数据中更大的比例。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":624},[625],{"type":20,"text":626,"format":11,"detail":11,"mode":22,"style":10,"version":12},"干净流量与安装基建，落在哪一环",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":628},[629,631,633],{"type":20,"text":630,"format":11,"detail":11,"mode":22,"style":10,"version":12},"归因再准，也只能和喂给它的数据一样准。如果你相当一部分点击是机器人、点击注入或低意图垃圾，MMP 会忠实地把这些垃圾也归因进去——虚高某些渠道、饿死另一些，并教会算法去追错误的人群。DeepClick 正作用在这一层：它优化归因的",{"type":20,"text":632,"format":12,"detail":11,"mode":22,"style":10,"version":12},"输入",{"type":20,"text":634,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，而不取代 MMP。",{"type":55,"listType":636,"start":12,"tag":637,"format":10,"indent":11,"version":12,"direction":13,"children":638},"bullet","ul",[639,655,671],{"type":60,"value":12,"format":10,"indent":11,"version":12,"direction":13,"children":640},[641],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":642},[643,645,647,653],{"type":20,"text":644,"format":12,"detail":11,"mode":22,"style":10,"version":12},"干净流量分离。",{"type":20,"text":646,"format":11,"detail":11,"mode":22,"style":10,"version":12}," ",{"type":35,"fields":648,"format":10,"indent":11,"version":40,"direction":13,"children":650},{"url":649,"newTab":38,"linkType":39},"https://deepclick.com/product/shield",[651],{"type":20,"text":652,"format":11,"detail":11,"mode":22,"style":10,"version":12},"DeepClick 绿盾（Shield）",{"type":20,"text":654,"format":11,"detail":11,"mode":22,"style":10,"version":12},"在流量进入安装链路之前，就把真实合格的访客与机器人、无效流量分开。更少的垃圾会话污染数据池，意味着归因信号更干净、MMP 的匹配建立在真人之上。",{"type":60,"value":69,"format":10,"indent":11,"version":12,"direction":13,"children":656},[657],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":658},[659,661,663,669],{"type":20,"text":660,"format":12,"detail":11,"mode":22,"style":10,"version":12},"降低安装流失。",{"type":20,"text":662,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 一套快速无摩擦的 ",{"type":35,"fields":664,"format":10,"indent":11,"version":40,"direction":13,"children":666},{"url":665,"newTab":38,"linkType":39},"https://deepclick.com/product/pwa-install",[667],{"type":20,"text":668,"format":11,"detail":11,"mode":22,"style":10,"version":12},"PWA 安装链路",{"type":20,"text":670,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，让用户不在「点广告」和「首次打开」之间放弃——那道缝隙正是归因丢失、安装无法记功的地方。",{"type":60,"value":40,"format":10,"indent":11,"version":12,"direction":13,"children":672},[673],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":674},[675,677,678,684],{"type":20,"text":676,"format":12,"detail":11,"mode":22,"style":10,"version":12},"再触达找回。",{"type":20,"text":646,"format":11,"detail":11,"mode":22,"style":10,"version":12},{"type":35,"fields":679,"format":10,"indent":11,"version":40,"direction":13,"children":681},{"url":680,"newTab":38,"linkType":39},"https://deepclick.com/product/re-engagement",[682],{"type":20,"text":683,"format":11,"detail":11,"mode":22,"style":10,"version":12},"再触达（Re-engagement）",{"type":20,"text":685,"format":11,"detail":11,"mode":22,"style":10,"version":12},"把流失用户拉回来，找回那些本会从漏斗漏掉、永不被计量的安装与转化。",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":687},[688],{"type":20,"text":689,"format":11,"detail":11,"mode":22,"style":10,"version":12},"这一切都不会让 DeepClick 变成归因厂商——做归因的依然是你的 MMP。DeepClick 只是确保它所测量的流量与安装是真实且完整的。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":691},[692],{"type":20,"text":693,"format":11,"detail":11,"mode":22,"style":10,"version":12},"移动广告归因最佳实践",{"type":55,"listType":636,"start":12,"tag":637,"format":10,"indent":11,"version":12,"direction":13,"children":695},[696,704,712,720,728],{"type":60,"value":12,"format":10,"indent":11,"version":12,"direction":13,"children":697},[698],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":699},[700,702],{"type":20,"text":701,"format":12,"detail":11,"mode":22,"style":10,"version":12},"用 MMP 作为中立的事实来源。",{"type":20,"text":703,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 别信各网络自报的数字——它们会重复计数。",{"type":60,"value":69,"format":10,"indent":11,"version":12,"direction":13,"children":705},[706],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":707},[708,710],{"type":20,"text":709,"format":12,"detail":11,"mode":22,"style":10,"version":12},"有意识地选归因模型。",{"type":20,"text":711,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 效果直投默认末次点击；渠道大规模重叠时切数据驱动。",{"type":60,"value":40,"format":10,"indent":11,"version":12,"direction":13,"children":713},[714],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":715},[716,718],{"type":20,"text":717,"format":12,"detail":11,"mode":22,"style":10,"version":12},"为 iOS 单独规划。",{"type":20,"text":719,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 把 SKAN 数据当作聚合且延迟的，别指望和 Android 做到用户级对齐。",{"type":60,"value":86,"format":10,"indent":11,"version":12,"direction":13,"children":721},[722],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":723},[724,726],{"type":20,"text":725,"format":12,"detail":11,"mode":22,"style":10,"version":12},"放量前验证每条归因链接。",{"type":20,"text":727,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 跑一次 QA 安装，确认来源被正确记功。",{"type":60,"value":435,"format":10,"indent":11,"version":12,"direction":13,"children":729},[730],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":731},[732,734],{"type":20,"text":733,"format":12,"detail":11,"mode":22,"style":10,"version":12},"守住漏斗顶端。",{"type":20,"text":735,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 过滤机器人与无效流量，让归因建立在真人而非噪声之上。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":737},[738],{"type":20,"text":739,"format":11,"detail":11,"mode":22,"style":10,"version":12},"常见问题（FAQ）",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":741},[742],{"type":20,"text":743,"format":11,"detail":11,"mode":22,"style":10,"version":12},"移动归因和移动测量有什么区别？",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":745},[746],{"type":20,"text":747,"format":11,"detail":11,"mode":22,"style":10,"version":12},"归因是把一次安装或事件的功劳记给具体来源；测量是更宽的概念，追踪围绕它的全部指标——安装、留存、ROI、生命周期价值。归因是测量里的一项核心功能，MMP 两者都做。",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":749},[750],{"type":20,"text":751,"format":11,"detail":11,"mode":22,"style":10,"version":12},"AppsFlyer 是归因工具还是 MMP？",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":753},[754,756,761],{"type":20,"text":755,"format":11,"detail":11,"mode":22,"style":10,"version":12},"都是。AppsFlyer 是移动测量合作伙伴，核心功能就是归因，同时也做深度链接、反作弊、人群和分析。想看它与 Adjust、Branch、Kochava 等的对比，请看",{"type":35,"fields":757,"format":10,"indent":11,"version":40,"direction":13,"children":758},{"url":37,"newTab":38,"linkType":39},[759],{"type":20,"text":760,"format":11,"detail":11,"mode":22,"style":10,"version":12},"最佳归因软件",{"type":20,"text":762,"format":11,"detail":11,"mode":22,"style":10,"version":12},"指南。",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":764},[765],{"type":20,"text":766,"format":11,"detail":11,"mode":22,"style":10,"version":12},"iOS 14.5 是不是把移动归因「杀死」了？",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":768},[769],{"type":20,"text":770,"format":11,"detail":11,"mode":22,"style":10,"version":12},"没有，但改变了它。App 跟踪透明度让 IDFA 大面积不可用，iOS 上的确定性归因如今依赖授权，而 SKAdNetwork 提供隐私保护的聚合数据。归因依然能跑——只是在 iOS 上更吵、更聚合。",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":772},[773],{"type":20,"text":774,"format":11,"detail":11,"mode":22,"style":10,"version":12},"什么是移动应用归因链接？",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":776},[777],{"type":20,"text":778,"format":11,"detail":11,"mode":22,"style":10,"version":12},"它是交给广告网络的一条追踪 URL，携带 campaign 参数，被点击时把用户跳转到商店或 App，同时记录互动，好让随之而来的安装记到正确的来源上。",{"type":16,"tag":373,"format":10,"indent":11,"version":12,"direction":13,"children":780},[781],{"type":20,"text":782,"format":11,"detail":11,"mode":22,"style":10,"version":12},"干净流量真能提升归因准确率吗？",{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":784},[785],{"type":20,"text":786,"format":11,"detail":11,"mode":22,"style":10,"version":12},"能。归因的可靠程度只取决于它的输入。在安装之前就剔除机器人、点击注入和无效流量，匹配到的数据才反映真人——这在信号本就稀缺的 iOS 上尤其值钱。",{"type":16,"tag":17,"format":10,"indent":11,"version":12,"direction":13,"children":788},[789],{"type":20,"text":790,"format":11,"detail":11,"mode":22,"style":10,"version":12},"核心要点",{"type":55,"listType":636,"start":12,"tag":637,"format":10,"indent":11,"version":12,"direction":13,"children":792},[793,801,815,823,831,839],{"type":60,"value":12,"format":10,"indent":11,"version":12,"direction":13,"children":794},[795],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":796},[797,799],{"type":20,"text":798,"format":12,"detail":11,"mode":22,"style":10,"version":12},"移动广告归因",{"type":20,"text":800,"format":11,"detail":11,"mode":22,"style":10,"version":12},"把安装与应用内事件，追溯回带来它们的广告、campaign 和素材。",{"type":60,"value":69,"format":10,"indent":11,"version":12,"direction":13,"children":802},[803],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":804},[805,807,809,811,813],{"type":20,"text":806,"format":11,"detail":11,"mode":22,"style":10,"version":12},"匹配要么",{"type":20,"text":808,"format":12,"detail":11,"mode":22,"style":10,"version":12},"确定性",{"type":20,"text":810,"format":11,"detail":11,"mode":22,"style":10,"version":12},"（共享 ID、准确率高），要么",{"type":20,"text":812,"format":12,"detail":11,"mode":22,"style":10,"version":12},"概率性",{"type":20,"text":814,"format":11,"detail":11,"mode":22,"style":10,"version":12},"（统计估算、用作兜底）；现代体系两者混用。",{"type":60,"value":40,"format":10,"indent":11,"version":12,"direction":13,"children":816},[817],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":818},[819,821],{"type":20,"text":820,"format":12,"detail":11,"mode":22,"style":10,"version":12},"MMP",{"type":20,"text":822,"format":11,"detail":11,"mode":22,"style":10,"version":12},"（AppsFlyer、Adjust、Branch、Kochava、Singular）是跨网络去重并统一归因的中立裁判。",{"type":60,"value":86,"format":10,"indent":11,"version":12,"direction":13,"children":824},[825],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":826},[827,829],{"type":20,"text":828,"format":12,"detail":11,"mode":22,"style":10,"version":12},"多触点模型",{"type":20,"text":830,"format":11,"detail":11,"mode":22,"style":10,"version":12},"决定功劳如何分配；末次点击是移动端默认，数据驱动在大体量下更公平。",{"type":60,"value":435,"format":10,"indent":11,"version":12,"direction":13,"children":832},[833],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":834},[835,837],{"type":20,"text":836,"format":12,"detail":11,"mode":22,"style":10,"version":12},"iOS ATT 与 SKAdNetwork",{"type":20,"text":838,"format":11,"detail":11,"mode":22,"style":10,"version":12}," 用隐私保护的聚合数据取代了确定性 IDFA 匹配——抬高了干净信号的价值。",{"type":60,"value":840,"format":10,"indent":11,"version":12,"direction":13,"children":841},6,[842],{"type":24,"format":10,"indent":11,"version":12,"direction":13,"children":843},[844,846,848],{"type":20,"text":845,"format":11,"detail":11,"mode":22,"style":10,"version":12},"DeepClick 是让归因数据更干净的",{"type":20,"text":847,"format":12,"detail":11,"mode":22,"style":10,"version":12},"流量质量与安装基建层",{"type":20,"text":849,"format":11,"detail":11,"mode":22,"style":10,"version":12},"，不取代你的 MMP。",{"id":851,"alt":852,"updatedAt":853,"createdAt":853,"url":854,"thumbnailURL":116,"filename":855,"mimeType":856,"filesize":857,"width":858,"height":859},309,"MMP 移动归因概念图：多广告渠道数据流汇聚到统一归因看板","2026-06-10T06:36:56.603Z","https://cms-r2.deepclick.com/hero-c-20a81fdddb02.png","hero-c-20a81fdddb02.png","image/png",1834330,1536,1024,{"title":861,"description":862,"image":863},"移动广告归因指南：原理、MMP 与 AppsFlyer 实操 2026","一篇讲透移动广告归因：确定性归因 vs 概率归因、MMP 平台、多触点归因模型、iOS SKAdNetwork 隐私冲击，并手把手教你配置 AppsFlyer 归因链接。",{"id":851,"alt":852,"updatedAt":853,"createdAt":853,"url":854,"thumbnailURL":116,"filename":855,"mimeType":856,"filesize":857,"width":858,"height":859},"published","mobile-advertising-attribution-guide-zh",{"id":69,"name":867,"avatar":868,"updatedAt":874,"createdAt":875},"DeepClick",{"id":869,"alt":867,"updatedAt":870,"createdAt":870,"url":871,"thumbnailURL":116,"filename":872,"mimeType":856,"filesize":873,"width":859,"height":859},25,"2026-04-22T08:09:22.606Z","https://cms-r2.deepclick.com/头像-白.png","头像-白.png",26626,"2026-04-22T08:09:35.299Z","2026-04-22T06:42:49.116Z",{"id":40,"titleZh":877,"titleEn":878,"slug":879,"order":69,"updatedAt":880,"createdAt":880},"行业信息","Industry Info","industry-info","2026-04-22T03:32:29.529Z","2026-06-16T15:58:39.587Z","2026-06-16T15:58:50.732Z","2026-06-16T15:58:20.375Z","\u003Cdiv class=\"payload-richtext\">\u003Ch2>什么是移动广告归因\u003C/h2>\u003Cp>\u003Cstrong>移动广告归因（mobile advertising attribution）\u003C/strong>，是把一次 App 安装或应用内事件，追溯回究竟是哪条广告、哪个 campaign、哪个渠道、哪套素材带来的过程。用户点了广告、随后装了你的 App，归因要回答的就是「这次安装该记在谁头上」——算清这一笔，你才能衡量广告投入产出比、优化出价、砍掉不出转化的渠道。\u003C/p>\u003Cp>没有归因，移动买量就是凭感觉花钱；有了归因，每块预算都能追到结果。本文是一篇「原理拆解」型指南：讲清归因是什么、匹配怎么发生、MMP 扮演什么角色、多触点模型如何分配功劳，以及苹果隐私变化如何重塑整套体系。如果你想要的是各家工具的横向对比排名，请看姊妹篇\u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/best-attribution-software-zh/\">最佳归因软件盘点\u003C/a>——本文讲「怎么做」，那篇讲「选哪个」。\u003C/p>\u003Ch2>移动广告归因是如何工作的：基本链路\u003C/h2>\u003Cp>无论用哪家平台，归因系统的骨架都一样：\u003C/p>\u003Col class=\"list-number\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cp>\u003Cstrong>记录广告互动。\u003C/strong> 用户在某个广告网络（Meta、Google、TikTok、DSP）上点击广告，网络记下点击 ID、时间戳、设备信号和 campaign 元数据。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cp>\u003Cstrong>记录安装。\u003C/strong> 用户完成安装并打开 App，你的测量 SDK 触发首次打开（first-open）事件。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cp>\u003Cstrong>尝试匹配。\u003C/strong> 归因服务把这次安装信号与近期广告互动池逐一比对，判断该把功劳记给哪一次。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cp>\u003Cstrong>分配功劳并回传。\u003C/strong> 匹配到的来源进入数据看板，并通过 postback 回传给网络的优化引擎。\u003C/p>\u003C/li>\u003C/ol>\u003Cp>最关键、也最有讲究的，是第 3 步的「匹配」。匹配本质上有两条完全不同的路。\u003C/p>\u003Ch2>确定性归因 vs 概率归因\u003C/h2>\u003Cp>\u003Cstrong>确定性归因（deterministic）\u003C/strong>，靠的是广告点击和安装两端共享的同一个唯一、持久标识——历史上是设备 ID（iOS 的 IDFA、Android 的 GAID），或经深度链接透传的点击 ID。两端出现同一 ID，匹配几乎板上钉钉，精确且可审计。\u003C/p>\u003Cp>\u003Cstrong>概率归因（probabilistic，又叫指纹匹配）\u003C/strong>，在没有共享确定性 ID 时使用。它用一组非唯一信号——IP、设备型号、系统版本、时间临近度——加上统计建模「估」出一个匹配。它是有依据的猜测而非确定结论；当隐私限制把信号一层层剥掉，准确率随之下降。\u003C/p>\u003Cdiv class=\"lexical-table-container\">\n        \u003Ctable class=\"lexical-table\" style=\"border-collapse: collapse;\">\n          \u003Ctbody>\u003Ctr class=\"lexical-table-row\">\n        \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>维度\u003C/p>\n      \u003C/th>\n    \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>确定性归因\u003C/p>\n      \u003C/th>\n    \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>概率归因\u003C/p>\n      \u003C/th>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>匹配依据\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>共享唯一 ID（设备 ID / 点击 ID）\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>统计信号（IP、设备、时间）\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>准确率\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>非常高\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>中等，取决于模型\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>隐私敏感度\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>较高（用到标识符）\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>较低（无持久 ID）\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>iOS（ATT 后）可用性\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>无授权则受限\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>常受苹果政策限制\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>适用场景\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>自有渠道、已授权用户\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>ID 不可用时的兜底\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003C/tbody>\n        \u003C/table>\n      \u003C/div>\u003Cp>实际落地中，现代归因两者混合：授权与标识符允许时走确定性，缺失时用概率（或隐私保护的聚合方法）兜底。\u003C/p>\u003Ch2>移动测量合作伙伴（MMP）：中立的裁判\u003C/h2>\u003Cp>\u003Cstrong>移动测量合作伙伴（mobile measurement partner，MMP）\u003C/strong>，是一个独立第三方，跨你所有广告网络汇总归因数据，充当唯一、中立的事实来源。与其相信每个网络自报的数字——每家都会把同一次安装算成自己的——不如让 MMP 统一去重、一致分配功劳。MMP 还会把数据归一化成同一套结构、在 iOS 上托管 SKAdNetwork、运行反作弊，并把 postback 回传给各网络让其算法优化，是任何认真做移动买量的团队的中枢。\u003C/p>\u003Cdiv class=\"lexical-table-container\">\n        \u003Ctable class=\"lexical-table\" style=\"border-collapse: collapse;\">\n          \u003Ctbody>\u003Ctr class=\"lexical-table-row\">\n        \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>MMP 平台\u003C/p>\n      \u003C/th>\n    \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>突出优势\u003C/p>\n      \u003C/th>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>\u003Ca href=\"https://www.appsflyer.com/\">AppsFlyer\u003C/a>\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>集成网络目录最大；看板深入；反作弊工具强\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>\u003Ca href=\"https://www.adjust.com/\">Adjust\u003C/a>\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>数据干净、自动化、反作弊；已并入 AppLovin\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>\u003Ca href=\"https://www.branch.io/\">Branch\u003C/a>\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>深度链接与 web-to-app 起家；跨平台旅程强\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>\u003Ca href=\"https://www.kochava.com/\">Kochava\u003C/a>\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>灵活、隐私优先、数据管控可配置\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>\u003Ca href=\"https://www.singular.net/\">Singular\u003C/a>\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>营销分析 + 归因；成本/ROI 聚合见长\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003C/tbody>\n        \u003C/table>\n      \u003C/div>\u003Cp>DeepClick \u003Cstrong>不是\u003C/strong> MMP，也不与上述平台竞争。它处在更靠前的一层——流量与安装基建侧——让流入你 MMP 的数据更干净。下文展开。\u003C/p>\u003Ch3>如何配置 AppsFlyer 归因链接（实操）\u003C/h3>\u003Cp>\u003Cstrong>AppsFlyer 归因链接（attribution link，又称 OneLink）\u003C/strong>，就是你交给广告网络、用来给安装记功的追踪 URL。一套可落地的配置：\u003C/p>\u003Col class=\"list-number\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cp>\u003Cstrong>接入 SDK。\u003C/strong> 把 AppsFlyer SDK 加进 iOS 和 Android 包并配好 dev key，让 App 能触发首次打开与应用内事件。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cp>\u003Cstrong>注册广告网络为 partner。\u003C/strong> 在后台启用对应的集成 partner，让 postback 自动配好。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cp>\u003Cstrong>生成归因链接。\u003C/strong> 挂上 \u003Ccode>pid\u003C/code>（媒体来源）、\u003Ccode>c\u003C/code>（campaign）、\u003Ccode>af_adset\u003C/code>、\u003Ccode>af_ad\u003C/code> 等参数，让报表能按来源、campaign、广告组、素材拆分。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cp>\u003Cstrong>加深度链接跳转。\u003C/strong> 用 OneLink 让同一条 URL 跳转到 App Store、Google Play，或在已安装时直接进入某页面，减少流失。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"5\"\n        >\u003Cp>\u003Cstrong>放量前 QA。\u003C/strong> 跑一次测试安装，在原始数据里确认它被正确归到对应的 \u003Ccode>pid\u003C/code>/\u003Ccode>c\u003C/code>。\u003C/p>\u003C/li>\u003C/ol>\u003Cp>而以上每一次匹配的质量，都取决于链接本身控制不了的一件事：打到它上面的流量是不是真的。\u003C/p>\u003Ch2>多触点归因：功劳到底算谁的\u003C/h2>\u003Cp>现实中的转化路径往往不止一次触达——用户可能先看了视频广告，过几天点了搜索广告，然后才安装。\u003Cstrong>多触点归因（multi-touch attribution）\u003C/strong>决定这些触点之间如何分配功劳，一个\u003Cstrong>多触点归因平台（multi touch attribution platform）\u003C/strong>会套用模型，而不是一股脑把功劳全记给最后一次点击。\u003C/p>\u003Cdiv class=\"lexical-table-container\">\n        \u003Ctable class=\"lexical-table\" style=\"border-collapse: collapse;\">\n          \u003Ctbody>\u003Ctr class=\"lexical-table-row\">\n        \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>模型\u003C/p>\n      \u003C/th>\n    \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>功劳如何分配\u003C/p>\n      \u003C/th>\n    \u003Cth\n        class=\"lexical-table-cell lexical-table-cell-header-1\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>适用场景\u003C/p>\n      \u003C/th>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>末次点击\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>100% 给安装前最后一次触点\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>简单、效果直投\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>首次点击\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>100% 给第一次触点\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>衡量漏斗顶端的发现\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>线性\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>所有触点均分\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>每次交互同等重要\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>时间衰减\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>越靠近转化的触点分得越多\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>短、强意图的周期\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003Ctr class=\"lexical-table-row\">\n        \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>数据驱动\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>算法按实际贡献分配\u003C/p>\n      \u003C/td>\n    \u003Ctd\n        class=\"lexical-table-cell lexical-table-cell-header-0\"\n        \n        \n        style=\"border: 1px solid #ccc; padding: 8px;\"\n      >\n        \u003Cp>大体量、多渠道项目\u003C/p>\n      \u003C/td>\n    \n      \u003C/tr>\u003C/tbody>\n        \u003C/table>\n      \u003C/div>\u003Cp>末次点击仍是移动端默认，因为 App 安装窗口短；但当预算大、渠道相互重叠时，数据驱动模型能给出更公平的图景。\u003C/p>\u003Ch2>iOS 隐私冲击：ATT 与 SKAdNetwork\u003C/h2>\u003Cp>苹果的 \u003Cstrong>App 跟踪透明度（ATT）\u003C/strong> 框架，要求 App 在访问 IDFA 前必须征得用户同意。大多数用户会拒绝，这就抽走了支撑 iOS 上经典确定性归因的那个确定性标识符。\u003C/p>\u003Cp>为填补空缺，苹果提供 \u003Cstrong>SKAdNetwork（SKAN）\u003C/strong>——一个隐私保护框架：由苹果自己校验安装，并通过「转化值（conversion value）」回传聚合、延迟的转化数据。代价是实打实的：数据是聚合的（非用户级）、被随机计时器延迟、粒度有限。较新版本（SKAN 4）引入多次 postback 和粗粒度数值，苹果的 AdAttributionKit 又进一步扩展了该模型。\u003C/p>\u003Cp>实际后果是：在 iOS 上你要把 SKAN、已授权的确定性数据和建模估算混着用，并接受更多噪声。这让\u003Cstrong>流量质量与干净的安装链路比以往更重要\u003C/strong>——当信号本就聚合且稀缺，每一份作弊或浪费的曝光，都会污染那本就不多的数据中更大的比例。\u003C/p>\u003Ch2>干净流量与安装基建，落在哪一环\u003C/h2>\u003Cp>归因再准，也只能和喂给它的数据一样准。如果你相当一部分点击是机器人、点击注入或低意图垃圾，MMP 会忠实地把这些垃圾也归因进去——虚高某些渠道、饿死另一些，并教会算法去追错误的人群。DeepClick 正作用在这一层：它优化归因的\u003Cstrong>输入\u003C/strong>，而不取代 MMP。\u003C/p>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cp>\u003Cstrong>干净流量分离。\u003C/strong> \u003Ca href=\"https://deepclick.com/product/shield\">DeepClick 绿盾（Shield）\u003C/a>在流量进入安装链路之前，就把真实合格的访客与机器人、无效流量分开。更少的垃圾会话污染数据池，意味着归因信号更干净、MMP 的匹配建立在真人之上。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cp>\u003Cstrong>降低安装流失。\u003C/strong> 一套快速无摩擦的 \u003Ca href=\"https://deepclick.com/product/pwa-install\">PWA 安装链路\u003C/a>，让用户不在「点广告」和「首次打开」之间放弃——那道缝隙正是归因丢失、安装无法记功的地方。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cp>\u003Cstrong>再触达找回。\u003C/strong> \u003Ca href=\"https://deepclick.com/product/re-engagement\">再触达（Re-engagement）\u003C/a>把流失用户拉回来，找回那些本会从漏斗漏掉、永不被计量的安装与转化。\u003C/p>\u003C/li>\u003C/ul>\u003Cp>这一切都不会让 DeepClick 变成归因厂商——做归因的依然是你的 MMP。DeepClick 只是确保它所测量的流量与安装是真实且完整的。\u003C/p>\u003Ch2>移动广告归因最佳实践\u003C/h2>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cp>\u003Cstrong>用 MMP 作为中立的事实来源。\u003C/strong> 别信各网络自报的数字——它们会重复计数。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cp>\u003Cstrong>有意识地选归因模型。\u003C/strong> 效果直投默认末次点击；渠道大规模重叠时切数据驱动。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cp>\u003Cstrong>为 iOS 单独规划。\u003C/strong> 把 SKAN 数据当作聚合且延迟的，别指望和 Android 做到用户级对齐。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cp>\u003Cstrong>放量前验证每条归因链接。\u003C/strong> 跑一次 QA 安装，确认来源被正确记功。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"5\"\n        >\u003Cp>\u003Cstrong>守住漏斗顶端。\u003C/strong> 过滤机器人与无效流量，让归因建立在真人而非噪声之上。\u003C/p>\u003C/li>\u003C/ul>\u003Ch2>常见问题（FAQ）\u003C/h2>\u003Ch3>移动归因和移动测量有什么区别？\u003C/h3>\u003Cp>归因是把一次安装或事件的功劳记给具体来源；测量是更宽的概念，追踪围绕它的全部指标——安装、留存、ROI、生命周期价值。归因是测量里的一项核心功能，MMP 两者都做。\u003C/p>\u003Ch3>AppsFlyer 是归因工具还是 MMP？\u003C/h3>\u003Cp>都是。AppsFlyer 是移动测量合作伙伴，核心功能就是归因，同时也做深度链接、反作弊、人群和分析。想看它与 Adjust、Branch、Kochava 等的对比，请看\u003Ca href=\"https://deepclick.com/zh-CN/resources/blog/best-attribution-software-zh/\">最佳归因软件\u003C/a>指南。\u003C/p>\u003Ch3>iOS 14.5 是不是把移动归因「杀死」了？\u003C/h3>\u003Cp>没有，但改变了它。App 跟踪透明度让 IDFA 大面积不可用，iOS 上的确定性归因如今依赖授权，而 SKAdNetwork 提供隐私保护的聚合数据。归因依然能跑——只是在 iOS 上更吵、更聚合。\u003C/p>\u003Ch3>什么是移动应用归因链接？\u003C/h3>\u003Cp>它是交给广告网络的一条追踪 URL，携带 campaign 参数，被点击时把用户跳转到商店或 App，同时记录互动，好让随之而来的安装记到正确的来源上。\u003C/p>\u003Ch3>干净流量真能提升归因准确率吗？\u003C/h3>\u003Cp>能。归因的可靠程度只取决于它的输入。在安装之前就剔除机器人、点击注入和无效流量，匹配到的数据才反映真人——这在信号本就稀缺的 iOS 上尤其值钱。\u003C/p>\u003Ch2>核心要点\u003C/h2>\u003Cul class=\"list-bullet\">\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"1\"\n        >\u003Cp>\u003Cstrong>移动广告归因\u003C/strong>把安装与应用内事件，追溯回带来它们的广告、campaign 和素材。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"2\"\n        >\u003Cp>匹配要么\u003Cstrong>确定性\u003C/strong>（共享 ID、准确率高），要么\u003Cstrong>概率性\u003C/strong>（统计估算、用作兜底）；现代体系两者混用。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"3\"\n        >\u003Cp>\u003Cstrong>MMP\u003C/strong>（AppsFlyer、Adjust、Branch、Kochava、Singular）是跨网络去重并统一归因的中立裁判。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"4\"\n        >\u003Cp>\u003Cstrong>多触点模型\u003C/strong>决定功劳如何分配；末次点击是移动端默认，数据驱动在大体量下更公平。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"5\"\n        >\u003Cp>\u003Cstrong>iOS ATT 与 SKAdNetwork\u003C/strong> 用隐私保护的聚合数据取代了确定性 IDFA 匹配——抬高了干净信号的价值。\u003C/p>\u003C/li>\u003Cli\n          class=\"\"\n          style=\"\"\n          value=\"6\"\n        >\u003Cp>DeepClick 是让归因数据更干净的\u003Cstrong>流量质量与安装基建层\u003C/strong>，不取代你的 MMP。\u003C/p>\u003C/li>\u003C/ul>\u003C/div>","https://deepclick.com/zh-CN/resources/blog/mobile-advertising-attribution-guide-zh",{"zh-CN":865},1781753535329]