Data Driven Decision Making

Part VI · Chapter 17

Operating the D3M System

Turning a stack of analyses into an operating system that compounds instead of decays.

This closing chapter focuses on turning the artefacts analytics leaves behind — cards, memos, dashboards, and case packs — into infrastructure that compounds instead of rotting when the analyst rotates off. It treats each artefact as a data product with a name, owner role, version, contract, and refresh cadence, then funnels them into the one-page decision memo a sponsor can actually sign. From there it scales monitoring to the whole portfolio, names the learning failures that let an analytics system censor its own training data, and ends with a full Bean & Basket expansion case that walks the entire decision ladder against one question: which fifty new cities to enter, in what sequence, on what conditions.

Start reading

Topics covered

the data product view (name, owner, version, contract, cadence)the artefact catalog and case-pack architecturethe eleven-section decision memocounterfactual and reversal thresholdsportfolio monitoring and status roll-upsdrift, decay, and re-investment cadencesclosed-loop targeting and exploration budgetsdecision retrospectives and kill switches

In this chapter

  1. 17.1The Data Product ViewReframes every analytics artefact as a product with a name, owner role, version, contract, and refresh cadence, indexed in a shared catalog.
  2. 17.2Decision MemosBuilds the one-page, eleven-section decision memo that fuses all evidence into a signable recommendation, with a worked Bean & Basket retention example.
  3. 17.3Monitoring, Feedback, and Learning LoopsScales monitoring to the portfolio and exposes learning failures like closed-loop targeting, fixed by exploration budgets, retrospectives, and kill switches.
  4. 17.4Final Integrative Case: The Bean & Basket ExpansionAn integrative Bean & Basket expansion case walking the full ladder to recommend which fifty markets to enter, in what phased sequence.