Dynamic pricing

Automated price optimisation: high value, high stakes, and the data isn’t ready. ($850K, funded.)

Your group is the delivery lead for this funded initiative.


The mandate

The board has approved $850K to build AI-driven dynamic pricing, automatically adjusting prices across fashion and homewares lines in response to demand, competitor moves, inventory levels and margin targets. The case that won funding: a “significant revenue opportunity” and competitors “already doing it.”

What the board thinks it bought: “smarter prices that lift margin, live this year.”

Your job: deliver it without the predictable failures, and define what “good enough” pricing decisions look like before any price goes live automatically.


Why this one is hard to deliver

  • Pricing is precise and high-stakes, squarely in the danger zone of the trust tool. A wrong automated price is public, instant, and can read as price-gouging or trigger a margin loss across thousands of SKUs.
  • It depends on competitive data RetailFlow doesn’t cleanly have, plus demand signals split across online and store systems.
  • A pricing model that looks great in backtesting (the “demo”) can behave badly in live conditions it never saw. That is exactly the demo-to-production gap, with money attached.

The data reality (ask Priya)

Priya’s honest read: “high potential but high complexity: needs better competitive data, staff capability gaps. Recommend Phase 2, not Phase 1; 12+ months realistic.” Expect her to push back on any plan that promises live automated pricing inside a quarter. Your scope has to confront this.

Who to interview, and the tension they bring

  • Priya Sharma (Data): will tell you the honest timeline; the central tension of your sprint.
  • Marcus Kim (CIO): wants speed and sees the revenue prize; “competitors aren’t waiting.”
  • David Chen (CFO): owns the margin numbers and the go/no-go criteria; wants certainty AI can’t give.

Your starting question for “good enough”

Should pricing be fully automated, or human-approved above a threshold? What guardrails (price floors/ceilings, max change per cycle) make automated pricing safe, and where must a human approve before a price changes?