Data-Readiness Assessment: Dynamic Pricing Optimisation

Technical Note · Data & Analytics · 24 February 2026

Initiative: Dynamic Pricing Optimisation Author: P. Sharma

To: David Chen (CFO), Marcus Kim (CIO)
From: Dr. Priya Sharma, Head of Data & Analytics
Re: What our data can and cannot support before we contract a pricing engine

I support the initiative. I want to be equally clear about where our data stands today, because pricing is precise and high-stakes: an automated wrong price is public, instant, and can read as either gouging or a margin-destroying error. We should not contract a vendor on the assumption that our data is ready, because it is not.

1. We lack a clean competitor price feed

Effective dynamic pricing on a fashion and homewares assortment depends heavily on knowing where competitors sit. We do not have this. We have ad-hoc manual price checks done by category buyers: patchy, inconsistent, and weeks stale. There is no automated, structured, matched feed of competitor prices against our SKUs.

Critical gap: any vendor that claims strong results "with a competitor feed" is describing a feed we do not own. Acquiring and maintaining a matched competitor price feed is a separate cost and project, likely $80–120K/yr in data subscriptions plus matching effort. This is not in the $850K and must be planned explicitly.

2. Demand signals are split across online and store

Our online demand data (Shopify Plus) is clean and granular. Our in-store demand data sits in a 10–15 year old POS and inventory stack and is coarse, batch-loaded overnight, and not reliably joined to the same SKU and customer keys.

Signal	Online	Store (legacy POS)
Granularity	Per-event, real-time	Daily batch, aggregated
SKU keying	Consistent	~14% mismatch / unmapped
Price elasticity history	Usable	Sparse, unreliable

A model trained mostly on online behaviour will mis-price in stores, which are still ~65% of revenue. Reconciling these two worlds is foundational work, not a configuration step.

3. Backtest performance is not live performance

Read this before you read any vendor backtest: a backtest replays a pricing policy against history that already happened. It cannot observe how customers and competitors would have reacted to prices we never actually set. Backtests almost always overstate gains. I would discount any backtested margin figure heavily until we have run a live, guard-railed pilot on a narrow category.

Recommendation

Sequence a competitor-feed acquisition project before, not after, engine selection.
Fund the online/store data reconciliation as a named workstream.
Treat any vendor backtest as marketing until validated on a live, capped pilot.
Plan for 12+ months to a credible, trusted rollout. The Board's "this year" margin expectation is not realistic at full scope.

Dr. Priya Sharma, Head of Data & Analytics