Machine Learning for Managers [Book]

Marketing analytics stages: Demystifying and deploying machine learning

“…Quantitative academic marketing (QAM) research is predominantly deductive in nature; it implies deriving hypotheses from academic literature and collecting data to test them (Golder et al, 2023; Ma and Sun, 2020). In contrast, AMA is often inductive; researchers use ML algorithms to detect patterns in previously collected data and then interpret the patterns using their domain knowledge, academic theory, or common sense (Chapman et al, 1999; Geertsema, 2022). In AMA, BI contains deductive elements; the analyses might be based on managers’ expectations or questions.…”

“…QAM tends to highlight the relevance of unexpected, novel findings for theory development. In AMA, counterintuitive outcomes and originality are less relevant; outcomes are evaluated in terms of their capacity to solve problems or identify business opportunities (Chapman et al, 1999; Geertsema, 2022). For example, AMA results might suggest ways to improve communication by customer-facing employees, recommend adjusted investments in different advertising channels, or establish codes for credit scoring loan applicants.…”

“…Logistic regression and other supervised ML algorithms can be applied for many other AMA purposes, such as predicting churn probabilities or credit scoring to quantify probabilities that customers default on loans (Chapman et al, 1999; Geertsema, 2022; Verhoef et al, 2022). Analysts may adopt bivariate techniques or data visualisations to assess supervised ML models too.…”

“…The result of applying the programmed models would be an n × m table, with n customers and m possible behaviours, from which a fixed number of offers gets selected for each of the n customers, according to automated program codes. As always, the deployed models must be monitored (Chapman et al, 1999; Geertsema, 2022).…”

“…We emphasize transactional data as such data provide competitive advantages (Bhat and Darzi, 2016;Zhang et al, 2006), relative to survey or online web data that other organisations can collect as well. Transactional data are especially valuable for banks, airlines, retailers, and public organisations (Geertsema, 2022). Other industries analysts use alternative data sources.…”