Empowering Growth without guesswork through Data Science at Attest by Alastair James, Magda Zubala & Reyhaneh Esmaielbeiki
Attest is a consumer research platform that enables brands to create and run targeted surveys to generate actionable insights. Therefore, it is important for Attest to understand the behaviour of respondents of a survey and to remove low-quality answers to provide fair decision-making. In addition, presenting survey findings in an intuitive and robust manner is one of our goals. Here we present how the use of statistics, language models, and scientific experimentation has enabled us to better understand our users. We will explore the opportunity around open text questions: we explain how we leverage them at Attest by offering automated smart insights using customised NLP models. Next, we outline the various challenges survey data poses and present some of the pipelines we have built at Attest. We also introduce how we use NLP models alongside other statistical models to understand the respondent’s behaviour and remove potentially biased responses. Finally, we demonstrate how we leverage our data infrastructure to provide real-time model predictions, by providing an example of a model pipeline that we have deployed in production.
Supported by
