Overview

Founded in 2000, Scanbuy provides mobile data and marketing solutions to over 300 CPG brands, agencies, and retailers worldwide. One of Scanbuy’s core offerings is its CPG transaction and purchase dataset, a rich collection of deterministic in-store data collected from over 20 million opt-in grocery shoppers over the last three years. A second, newer data product, ExtendedAudiences™, uses complex algorithms to create look-alike audiences of “most-likely-to-buy” consumers for thousands of brand- and product-level audience segments. With Scanbuy, CPG brands reach the right grocery-shoppers with the right offers at the exact moments they are most likely to buy.

Problem 

In digital advertising, only about 10% of audience targeting is based on deterministic datasets, while 90% relies on probabilistic models, often generated through opaque “black-box” AI systems. These models introduce bias, inaccuracies, and a lack of explainability, leading to inefficiencies and missed opportunities in ad targeting. Advertisers needed a transparent, reliable, and data-driven approach to improve targeting accuracy and campaign performance.

Solution 

Howso’s Causal AI-powered solution redefined transparency in programmatic advertising. Unlike traditional black-box AI models, Howso’s deterministic approach enabled advertisers to construct customized audience segments based on real-time consumer purchase signals. With counterfactual analysis capabilities, advertisers could explore “what if” scenarios, refining their strategies dynamically.

Scanbuy took Howso and turned a dataset containing 100,000 consumers and expanded it to 1 million using look-alike ML modeling. In a hypothetical example, Scanbuy could predict that people who buy Huggies diapers also often buy Rao’s pasta sauce. Howso’s explainable AI framework allows Scanbuy to make the decisioning logic of all ML models open for scrutiny. And since Scanbuy’s data is anonymized and opt-in, if a customer opts out of the core audience, that person is also removed from the extended audience – ensuring unbroken and fully compliant data privacy. 

Results 

Using ExtendedAudiences, Scanbuy customers routinely connect with far larger number of consumers for no additional modeling costs — dramatically boosting campaign ROI. Typically, when an audience goes from deterministic to probabilistic, it experiences a decay in conversion accuracy of up to 50%. But when Scanbuy uses Howso to build probabilistic models, there is virtually zero reduction in audience quality.  

In one example, a top global beverage brand sent offers to an original deterministic audience and got a 16% engagement rate. Sending the same offer to a Scanbuy extended audience resulted in a 15% engagement rate – virtually the same campaign success rate to an audience 10X larger, with no additional incremental spend. This is a 10X breakthrough in campaign performance.  

Click here to see the impact of Howso on Scanbuy for yourself.