Go/No-Go Checkpoint and Project Risk Portfolio

Description

Build an agreed feasibility checkpoint and stopping rule into risky projects, and balance risky work with safer work across the team's portfolio.

Data projects fail because the data is not what was expected, contains no signal, or the customer did not want the result. Nolis and Robinson manage this risk up front. Get data samples before committing; where that is impossible, place an early go/no-go review in the timeline at which stakeholders reassess feasibility, and scope risky projects with an explicit cut-off, such as stopping if good data is not found within a month. Present the possibility of stopping at the outset so that stopping is not seen as failure. Run several projects at once so a risky model is backed by safer reports, show an MVP to customers early, and do not remove risk entirely, or the team stagnates. The practice transfers directly to ambitious research, AI features and analytics work in UX.

Attributes

Loss type
Siloed
Effort
Hours

Steps

  1. Request samples of the data or materials before the project starts.
  2. If that is impossible, schedule an early go/no-go review with stakeholders in the project timeline.
  3. Define a cut-off condition and date for risky projects (for example, stop if usable data is not found within one month).
  4. Present the possibility of stopping to stakeholders at the outset.
  5. Hold a mix of risky and safe projects so the portfolio still delivers if one is stopped.
  6. Show an MVP early and talk to customers.
  7. If data is lacking, try substitutes with caution or start a separate instrumentation project.

Sources

  • Build a Career in Data Science (Nolis & Robinson, 2020)

Also known as

  • Planned go/no-go checkpoint and project risk portfolio
  • Early stopping point for data projects
  • Feasibility gate
  • Kill criterion

Categories

  • Design
  • Design & Product Practice