Economics of the Data Office

Pavel Myagkikh25 May 2024

Problem

A chief data officer (CDO) in a company where every unit is judged by P&L must show what the data office earns. Direct effects such as selling data or bank scoring are easy to count: man-hours and infrastructure against money received. Most of the value is indirect: the data office changes decisions in business processes, and nobody prices those decisions. It also needs a rule for choosing the first projects when the company has no data culture.

Example from the article: a fashion retailer with 7.5 billion rubles turnover decides a year ahead how many dresses to sew, often by eye, and a 30% forecast error costs about 750 million rubles of margin.

Solution in brief

Treat the company as a stream of decisions and find the large processes. Price a process by what an error in its decision costs, then credit the data office with the share of that cost the model removes. Start with optimization of an existing process, where a baseline and KPIs already exist.

Questions this case answers
  • How to estimate the monetary value of an analytics function from the decisions it changes inside business processes, where most of the value is indirect
  • How to choose the first analytics projects in an organization without a data culture so that their economic return can be shown
  • How to estimate the economic value of a data office and its indirect effects
  • How to choose the first data projects when the company has no data culture

Tools and methods

  • Decision-stream view of the business — find processes with many costly decisions, then score them by money at stake.
  • Demand forecasting — the worked example of an optimization project.
  • A/B testing with a proper control group — the guard against inflated effects.
  • Global control group — a small share of users with no communication, kept for checks.
Continue reading — it's free

Google or a one-time email code. No password.