加入新推出的
Discord 社区,展开实时讨论,获得同行支持,并直接与 Meridian 团队互动!
付费搜索建模
使用集合让一切井井有条
根据您的偏好保存内容并对其进行分类。
了解作为搜索广告混杂变量的搜索查询量
将因果推理应用于营销时,最大的挑战可能是,当产品需求增强时,广告客户往往会在营销上投入更多资金。我们在分析营销支出的因果效应时,最关心的是如何区分 KPI 的增加是由于营销支出的增加,还是由于内在需求的增加。
对于搜索广告,内在需求与营销支出之间的密切关系尤为明显。这是因为只有当搜索查询与一组广告客户所定位的特定关键字相匹配时,页面上才会显示搜索广告。如果内在需求较高,自然搜索查询量也会很高,进而使搜索广告的总支出金额变高。因此,自然搜索查询量是搜索广告的一个重要混杂变量。如果不考虑这一因素,就难以进行有效的搜索广告因果推理。
对于搜索广告预算较高的广告客户,这一问题尤为突出,因为他们的付费搜索广告量与自然搜索查询量的关系往往更为密切。不过,如果预算较低的广告客户在需求旺盛期增加预算,或者仅在这些时期投放搜索广告系列,则也会受到该问题的影响。
Meridian 提供了相应选项,可在模型中加入 Google 自然搜索查询量 (GQV),作为 Google 搜索广告的混杂变量。来自非 Google 搜索引擎的自然搜索查询量往往无法获得。如果您想要将非 Google 付费搜索广告纳入模型,但没有来自相应搜索引擎的自然搜索查询量,以下替代方案可能适合您:
如果 GQV 能准确反映非 Google 搜索查询量,则可以减少偏差。建议您对此假设进行评估。您可以通过绘制图表来评估该假设,例如:

在上图中,y 轴表示媒体展示次数与品牌 GQV 之间的相关性,x 轴表示媒体展示次数与宽泛查询量之间的相关性。
如果您不想假设 GQV 能准确反映非 Google 搜索查询量,则可能需要在模型中忽略非 Google 搜索引擎。
如需详细了解广告定位导致的选择性偏差问题,请参阅在媒体组合建模中对付费搜索广告进行偏差校正。
如未另行说明,那么本页面中的内容已根据知识共享署名 4.0 许可获得了许可,并且代码示例已根据 Apache 2.0 许可获得了许可。有关详情,请参阅 Google 开发者网站政策。Java 是 Oracle 和/或其关联公司的注册商标。
最后更新时间 (UTC):2025-08-04。
[null,null,["最后更新时间 (UTC):2025-08-04。"],[[["\u003cp\u003eUnderstanding the impact of marketing spend on key performance indicators (KPIs) for search ads can be difficult because inherent demand and marketing spend are often correlated.\u003c/p\u003e\n"],["\u003cp\u003eOrganic query volume is a significant confounding factor for search ads as it influences both inherent demand and ad spending.\u003c/p\u003e\n"],["\u003cp\u003eMeridian offers the option to incorporate Google organic query volume (GQV) into the model to address this confounding factor.\u003c/p\u003e\n"],["\u003cp\u003eIf non-Google organic query volume is unavailable, using GQV as a proxy or excluding non-Google search engines from the model can be considered.\u003c/p\u003e\n"]]],["Advertisers face the challenge of distinguishing between increased marketing spend and inherent demand when analyzing marketing's causal effects. Search ad spending correlates with organic query volume, making it a key confounder. High query volume leads to higher ad spending, impacting both large and small advertisers. Meridian allows inclusion of Google organic query volume (GQV) to mitigate this. If non-Google query volume is unavailable, GQV can be used as a proxy or the non-Google search engine omitted.\n"],null,["# Paid search modeling\n\nUnderstanding query volume as a confounder for search ads\n---------------------------------------------------------\n\nPerhaps the biggest challenge in causal inference when applied to marketing is\nthat advertisers often spend more on marketing when there is stronger demand for\ntheir product. Disentangling whether an increase in the KPI is due to an\nincrease in marketing spend or due to an increase in inherent demand is a\nprimary concern when one is analyzing causal effects of marketing spend.\n\nThe strong relationship between inherent demand and marketing spend is\nparticularly salient when it comes to search ads. This is because a search ad is\nonly shown on the page if a search query matches certain keywords targeted by a\nset of advertisers. When inherent demand is high, organic query volume will also\nbe high, and so the total spending on search ads will be high. Therefore,\norganic query volume is an important confounder for search ads. It is hard to\nget good inference on search ads without it.\n\nThis is particularly an issue for advertisers with high search budgets because\ntheir paid search ad volume tends to track more closely with organic query\nvolume. However, this also affects lower budget advertisers who increase their\nbudgets during periods of high demand, or who only run search campaigns during\nthese periods.\n\nMeridian provides the option to include [Google organic query volume\n(GQV)](/meridian/docs/basics/using-mmm-data-platform) in the model as a\nconfounder for Google Search ads. Organic query volume from non-Google search\nengines is often unavailable. If you want to model non-Google paid search ads,\nand organic query volume from the corresponding search engine is not available,\nthe following alternatives might work for you:\n\n- Bias can be mitigated if GQV is a good proxy for the non-Google query\n volume. We recommend assessing this assumption. One way to help assess the\n assumption is by creating a plot, for example:\n\n The previous plot shows the correlation between media impressions and brand\n GQV on the y-axis, and the correlation between media impressions and generic\n query volume on the x-axis.\n- If you don't want to assume GQV is a good proxy for the non-Google query\n volume, you might need to omit the non-Google search engine from the model.\n\nFor more information about the challenges of selection bias due to ad targeting,\nsee [Bias Correction For Paid Search In Media Mix\nModeling](https://research.google/pubs/bias-correction-for-paid-search-in-media-mix-modeling/)."]]