Mobile-phone advertising enables marketers to reach customers at a personal level and it enables the measure of costumers reaction by novel approaches, in real time, and at scale. By keeping a device anonymous, we can deliver custom adverts and we can check when the device owner will visit a specific mortar-and-brick location. This is the first step in a sale. By measuring visits and sales, the original marketers can determine their return on advertising and they can prove the efficacy of the marketing investments. We turn our attention to the measure of lift: we define it as the visit acceleration during the campaign flight with respect to a controlled baseline. We present a theoretical description; we describe a general and a simplified approach in composing the exposed and the control baseline; we develop two different vertical approaches with different comparable solutions; finally, we present how to carry the experiments and the measures for a few dozens campaigns; these campaigns range from hundred thousands devices and counting a few hundred visits to a handful locations, to sixty million devices and counting million visits to thousands locations. We care about experiments at scale.

翻译：移动电话广告使营销人员能够在个人层面触达客户，并通过新颖方法实时、大规模地衡量消费者反应。通过保持设备匿名性，我们可投放定制化广告，并监测设备持有者何时到访特定实体门店——这是销售流程的第一步。通过衡量到访与销售数据，原始营销人员可计算广告投资回报率，并验证营销投入的有效性。我们聚焦于"提升"这一度量标准：将其定义为广告活动期间相对于控制基线的到访加速现象。我们首先阐述理论框架；在构建实验组与控制基线时，提出通用方法与简化方案；继而开发两种采用不同可对比解决方案的纵向方法；最后展示数十个广告活动的实验设计与数据采集方法——这些活动覆盖规模从追踪十万设备级（对应数百次门店到访、数个营业网点）到六千万设备级（对应百万次到访、数千个网点）。我们尤为关注大规模实验的可操作性。