32. Retail Use Case

32.1 Scenario Context #

“Calder & Rowe” is a constructed omnichannel retailer: 410 stores, an e-commerce platform, a marketplace presence, three distribution centres, and 2,100 SKUs across seasonal and continuity ranges.

The problem it cannot solve. Commercial outcomes are determined by interactions the organization measures separately. A store’s sales decline. Merchandising attributes it to range performance. Supply chain sees adequate stock in the DC. Store operations sees normal staffing. E-commerce sees a regional uplift. In fact: a delivery was late, causing a two-day gap on a key line, during a local competitor promotion, in a store where the replenishment task was deprioritized because a colleague was absent, and the customers affected shifted to online where margin is lower.

Each function’s data supports its own explanation. None supports the actual one.

32.2 Correlation Architecture #

flowchart TB
    subgraph SRC["Sources"]
        direction LR
        S1["POS<br/><i>stream · transactions, baskets</i>"]
        S2["E-commerce<br/><i>stream · sessions, carts, orders</i>"]
        S3["Inventory<br/><i>CDC · stock position by location</i>"]
        S4["Supply chain<br/><i>stream · shipments, ASNs, delays</i>"]
        S5["Store ops<br/><i>stream · tasks, labour, compliance</i>"]
        S6["Loyalty / CRM<br/><i>CDC · membership, history</i>"]
        S7["Pricing & promo<br/><i>CDC · price changes, campaigns</i>"]
        S8["Competitor & market<br/><i>batch · external feeds</i>"]
        S9["Weather & local events<br/><i>batch/stream · external</i>"]
        S10["Customer service<br/><i>stream · contacts, returns</i>"]
    end

    subgraph SCOPES["Correlation Scopes"]
        direction TB
        SC1["<b>comm.demand_disruption.v3</b><br/>M1·M2·M3·M5 · 14 d<br/>What combination explains<br/>location-level demand deviation?"]
        SC2["<b>comm.channel_shift.v2</b><br/>M2·M1 · 30 d<br/>Which customers shifted<br/>channel, and following what?"]
        SC3["<b>supply.availability_impact.v3</b><br/>M1·M3 · 7 d<br/>What is the realized sales<br/>impact of each stockout?"]
        SC4["<b>cust.lifetime_risk.v4</b><br/>M2·M4·M5 · 180 d · batch<br/>Which experience sequences<br/>precede value decline?"]
        SC5["<b>loss.shrink_pattern.v2</b><br/>M2·M5 · 30 d<br/>Which conjunctions correlate<br/>with unexplained shrink?"]
    end

    subgraph OUT["Illustrative Insight"]
        direction TB
        O1["<b>I-5512</b> — confidence 0.81<br/>'Store 214 demand decline is<br/>primarily attributable to a<br/>48h availability gap on 3 lines<br/>coinciding with a competitor<br/>promotion, not to range<br/>performance.'<br/><br/><b>Evidence:</b> 6 correlations<br/><b>Estimated displaced demand:</b><br/>62% shifted online (lower margin),<br/>38% lost<br/><b>Modifiable factor:</b> replenishment<br/>task deprioritized during<br/>absence cover<br/><b>Alternative:</b> range decline (0.14)"]
    end

    SRC --> SCOPES --> OUT

    classDef src fill:#1b2f3d,stroke:#5b8fb0,color:#e6f2fa
    classDef sc fill:#0a5570,stroke:#3fd0f0,stroke-width:2px,color:#ffffff
    classDef out fill:#3d3313,stroke:#d4a636,stroke-width:2px,color:#fff8e6
    class SRC,S1,S2,S3,S4,S5,S6,S7,S8,S9,S10 src
    class SCOPES,SC1,SC2,SC3,SC4,SC5 sc
    class OUT,O1 out

Figure 46. Retail unified commerce correlation workflow (illustrative). The insight attributes a demand decline to a specific conjunction and explicitly names the alternative explanation the organization would otherwise have adopted — range performance — with its lower confidence. This attribution correction is the scenario’s central value.

32.3 Channel Shift as a Correlation Problem #

comm.channel_shift.v2 illustrates a pattern that is invisible to conventional analytics.

A retailer’s channel reporting shows store sales down and online sales up. The natural interpretation is structural channel migration — an industry-wide trend requiring strategic response.

The correlation substrate can distinguish this from induced channel shift: individual customers who purchased in store, encountered a specific failure (unavailability, queue, unhelpful interaction), and subsequently purchased the same category online. This is not a trend; it is a set of remediable service failures with a measurable margin cost.

The distinction requires entity-resolved correlation across POS, loyalty, e-commerce, and store operations data with a customer-level view — precisely the cross-domain, entity-anchored correlation the framework provides and that channel-level aggregate reporting structurally cannot.

32.4 Privacy Constraints in Retail #

Retail is where the framework’s privacy risks are most acute, because retail correlation readily produces special-category inferences from entirely non-special-category inputs.

Framework requirements applied here:

• Scopes producing customer-level behavioural correlations MUST declare a specific purpose, not “personalization” as a category.
• Special-category determination MUST be performed by someone with the imagination to anticipate non-obvious inferences, reviewed by the privacy office.
• Correlations capable of revealing special-category information MUST be classified as such on formation, inheriting the restrictions rather than the permissions of their inputs.
• Marketing purposes MUST NOT inherit from service-delivery or fraud-prevention collection bases — the intersection rule prevents this, and attempts to override it should be treated as a governance escalation.

32.5 Illustrative Outcomes #

Table 67. Illustrative outcome directions for retail. Three rows are marked as new capability rather than improvement, because conventional channel-level reporting cannot produce them at all.

32.6 What Makes This Hard #

Identity resolution across anonymous and known. A substantial proportion of store transactions are anonymous. Linking anonymous POS activity to a known customer requires payment token matching, loyalty scanning, or probabilistic inference — the last of which carries both accuracy risk and privacy risk. The framework’s precision-favouring thresholds (Chapter 13) apply with particular force: an over-merged customer identity in retail produces both wrong analysis and a potential data protection breach when one customer’s history is exposed to another.

External data quality. Competitor pricing, local events, and weather feeds vary substantially in reliability and timeliness. Correlations depending heavily on external feeds inherit their quality, and the contract discipline that applies to internal sources must apply equally to purchased data.

Seasonality dominates behavioural baselines. Retail behaviour is profoundly seasonal, and a naïve baseline will flag every seasonal transition as anomalous. Baselines must be seasonally adjusted, and the adjustment must be robust to the fact that each year’s seasonality differs.

Commercial pressure toward higher autonomy. Retail organizations will want automated pricing, promotion, and personalization actions at high autonomy levels because the volume justifies it. Many such actions are less reversible than they appear — a promotional price communicated to a customer cannot be un-communicated, and a personalized recommendation revealing an inferred pregnancy cannot be un-shown. The reversibility test in Chapter 20 should be applied strictly against commercial pressure.

Next: 33. Smart Cities Use Case →