Commerce OS guide

Opportunities

Guide status: Operating-system guidance. Last reviewed 14 August 2026. Opportunity controls and readiness labels can vary by installed version; verify them in the current workspace before relying on this guide operationally.

Purpose

Opportunities converts store evidence into a manageable set of commercial questions. Its job is to help a team decide what deserves investigation, not to declare that every pattern is causal or every idea will produce an uplift.

An opportunity should keep the evidence, affected scope, proposed direction, confidence, constraints and next decision connected. That makes it possible to disagree constructively, defer work without losing context, and later understand why an action was taken.

Before you begin

  • Confirm the store and reporting window.
  • Resolve material data or placement problems in Store Health first.
  • Identify the commercial objective: for example, discovery, conversion quality, order composition or retention.
  • Check who can approve any follow-on work.

Read an opportunity

Start with the observed condition. This should describe what the evidence shows without claiming a cause. Then read the hypothesis: the reason the condition may matter and the proposed relationship to test or investigate. Keep those two statements distinct.

Inspect the scope. Product, collection, audience, placement, market and time-window limits can materially change the meaning of the evidence. Review confidence alongside evidence class and volume; a confident inference is still an inference.

Priority is useful only when its inputs are visible. A high-value possibility with low confidence may deserve an experiment, while a lower-value but urgent configuration problem may deserve immediate remediation. Treat the priority as a decision aid, not an autonomous instruction.

From question to controlled work

  1. Qualify: confirm the condition is not explained by missing data, seasonality or a recent store change.
  2. Scope: state what is included and excluded.
  3. Choose a route: dismiss, watch, investigate, turn into a recommendation, or design an experiment.
  4. Set authority: record whether work can be prepared, needs approval, or may execute within an existing policy.
  5. Define success: choose a primary measure, guardrails and an appropriate observation window.

Safe use and fallback

If evidence is insufficient, preserve the opportunity as Learning or Unknown rather than forcing a rank or forecast. If an opportunity conflicts with margin, inventory, brand, consent or customer-experience rules, those constraints take precedence. A dismissed or deferred item should retain its reason so that the same question is not repeatedly rediscovered without context.

You are finished when

  • observation and hypothesis are clearly separated;
  • scope, evidence class and limitations are visible;
  • the next decision and owner are recorded; and
  • any proposed measurement can distinguish useful learning from noise.