Lease clauses, POS and foot-traffic fragmented — Snowflake unifies data for store decisions

Overview

Retail footprint optimization at a fashion brand stalled because lease terms, point-of-sale performance, and foot traffic signals lived in separate systems and formats. Store teams debated screenshots, Legal chased clauses in PDFs, and executives lacked a single view to weigh renewals, relocations, or closures. Intelligex integrated NCR point-of-sale (POS) data, Placer.ai visitation feeds, and lease documents extracted into Snowflake, then surfaced options in a decision app with Legal reviews on flagged clauses. Decisions moved with better confidence, fewer meetings, and less rework because every option referenced the same governed snapshot and approvals.

Client Profile

• Industry: Fashion retail and lifestyle brand
• Company size (range): Enterprise retailer with a multi-region store network
• Stage: Established brand balancing omnichannel growth and lease renewals
• Department owner: Strategy, Analytics & Executive Leadership (Corporate Strategy / Real Estate)
• Other stakeholders: Legal, Store Operations, Finance/FP&A, Merchandising, eCommerce, IT/Data Engineering, Procurement

The Challenge

Footprint choices depended on reconciling store performance, trade area dynamics, and contract constraints. In practice, NCR POS held sales and basket data, Placer.ai provided visitation and co-tenancy signals, and lease terms lived as scanned PDFs in shared drives. Teams copied values into spreadsheets with varying cut-offs and definitions. Legal reviewed clauses late, only after recommendations were drafted, and clauses like co-tenancy, exclusivity, radius restrictions, or termination windows were easy to miss.

Leaders wanted to keep core systems—POS, analytics, and document repositories—while reducing the manual reconciliation that slowed decisions. They asked for a single place to compare options by store, with sourced sales and traffic context, extracted lease constraints, and a lightweight approval path that brought Legal and Finance into the flow before recommendations reached executive forums.

Why It Was Happening

Identity, timing, and evidence were fragmented. Store identifiers differed across POS, traffic, and lease files; visitation and POS figures refreshed on different cadences; and lease terms were hidden in attachments without a structured schema. Analysts reconciled by hand, producing defensible yet inconsistent decks. Decision meetings reopened basics—what the lease allowed, whether a clause applied, which cut-offs were used—rather than weighing options.

Governance arrived at the end. There was no approval gate to ensure a Legal read on sensitive clauses before recommendations, no change log for shifting inputs, and no audit trail binding a decision to its exact sources. The result: repeated meetings, last-minute redlines, and uneven confidence in outcomes.

The Solution

We built a governed pipeline that unified store performance, trade area signals, and lease terms in Snowflake, and exposed them in a decision app designed for portfolio choices. NCR POS and Placer.ai feeds landed on a schedule; lease PDFs were parsed with an AI document service into structured fields. A conformed model harmonized store identity and calendars, applied validations, and flagged clause risks for Legal review. The decision app presented side-by-side options—renew, resize, relocate, or exit—with sourced charts, clause summaries, and an approval-backed change log. Nothing was replatformed: existing tools remained systems of record while orchestration handled integration, rules, and workflow.

• POS integration for sales, basket, and returns from NCR POS
• Trade area and foot traffic feeds from Placer.ai aligned to store geographies
• Lease document extraction for key terms, dates, and clauses using Azure AI Document Intelligence
• Conformed analytics model and snapshots in Snowflake with store identity, calendars, and geospatial tags
• Transformations and validations for identity mapping, clause detection confidence, refresh cut-offs, and outlier screening
• Decision app with scenario toggles and sourced tiles built on Microsoft Power Apps, embedded with maps and clause summaries
• Legal review workflow for flagged clauses and edge cases using Power Automate Approvals, capturing comments and sign-off
• Role-based access enforced via identity groups so Real Estate, Legal, Finance, and executives see appropriate views (Okta Groups)
• Change log linking each recommendation to sources, snapshot ID, clause interpretations, and approvers
• Operational dashboard tracking data freshness, extraction quality, open reviews, and decisions in flight

Implementation

• Discovery: Cataloged NCR POS fields and store identifiers, Placer.ai visitation feeds and coverage, and lease repositories and formats. Mapped common clauses and redline triggers with Legal. Reviewed prior footprint decks to pinpoint repeated reconciliation and late-stage edits.
• Design: Defined the conformed store model with geography and calendar alignment; authored the lease clause schema and confidence thresholds; designed validations for identity, cut-offs, and outliers; scoped scenario views and maps in the decision app; and documented Legal and Finance approval steps with reason codes.
• Build: Implemented scheduled POS and traffic feeds into Snowflake; configured Azure Document Intelligence to extract lease terms; built transformations for identity mapping, clause detection, and validations; developed the Power Apps decision interface with sourced tiles and clause highlights; and wired approval flows with audit logging.
• Testing and QA: Replayed recent renewals to reconcile sales, traffic, and lease constraints; spot-checked clause extraction against original PDFs; tuned confidence thresholds and review queues; verified role-based access and change logs; and ran table-top reviews with Real Estate, Legal, and Finance.
• Rollout: Launched read-only decision views alongside existing spreadsheets; after validation, enabled approvals for selected markets; then expanded to the broader estate with Legal-focused clause reviews. Maintained a manual override for sensitive negotiations with post-review documentation.
• Training and hand-off: Delivered quick guides for Real Estate on scenario toggles and sourcing, for Legal on clause review flows and notes, and for Finance on reading profitability and traffic context. Assigned stewardship for store mappings, clause dictionaries, and refresh calendars with a regular review cadence.

Results

Store decisions moved from slide assembly to option selection in a single application. Each recommendation included sourced POS and traffic context, extracted lease constraints, and recorded approvals. Legal weighed in earlier on co-tenancy, exclusivity, and termination windows, so negotiation paths were set before executive forums. Meetings focused on trade-offs rather than reconciling inputs.

Rework declined because identity and cut-offs were harmonized and clause interpretations were captured once, then reused. Snapshots gave executives confidence that charts and clauses matched a controlled dataset. With clear ownership and a visible audit trail, the team made portfolio calls with shared facts and fewer follow-up cycles.

What Changed for the Team

• Before: Spreadsheets stitched POS, traffic, and lease screenshots. After: A decision app displayed sourced tiles with extracted clauses and approvals.
• Before: Legal reviewed clauses after recommendations were drafted. After: Flagged clauses triggered early review and notes embedded in the option set.
• Before: Store IDs and cut-offs varied by source. After: A conformed model aligned identity and calendars under snapshots.
• Before: Edits and clarifications lived in email threads. After: A change log and approval flow captured rationale and ownership.
• Before: Meetings reopened basic facts. After: Discussions weighed renew, resize, relocate, or exit with shared context.

Key Takeaways

• Unify POS, foot traffic, and lease terms under a governed model so footprint choices start from shared facts.
• Extract and structure key clauses; a searchable clause schema beats hunting through PDFs during negotiations.
• Design decision views, not just reports; scenario toggles and sourced tiles keep meetings focused on options.
• Pull Legal and Finance into the flow with light approvals before recommendations reach executives.
• Keep existing systems; layer orchestration, validations, and a decision app to reduce rework without replatforming.

FAQ

What tools did this integrate with?
We connected NCR POS for sales and basket context, aligned Placer.ai visitation feeds (Placer.ai), and extracted lease terms with Azure AI Document Intelligence. Data landed in Snowflake, the decision app ran on Power Apps, approvals used Power Automate, and access was governed by identity groups such as Okta Groups.

How did you handle quality control and governance?
A conformed store model aligned identity and calendars, validations checked source freshness and outliers, and lease extraction carried confidence scores with a human-in-the-loop review. Flagged clauses routed to Legal for interpretation with notes. Snapshots bound each decision cycle to a controlled dataset, and a change log recorded sources, clause decisions, and approvals.

How did you roll this out without disruption?
We ran the decision app in parallel with existing spreadsheets while teams validated mappings, clause extraction, and scenario logic. After trust was established, we enabled approvals for select markets and broadened coverage. Core systems remained as they were; the workflow layered integration, reviews, and auditability on top.

How were lease clauses extracted and reviewed?
Lease PDFs were processed by the document service to capture term, rent provisions, co-tenancy, exclusivity, radius restrictions, options, and termination rights. Each field included a confidence and a link to the source passage. Items below threshold or marked sensitive entered a review queue where Legal confirmed interpretations and added guidance used in future scenarios.

How did you address privacy and data rights for foot traffic analytics?
Placer.ai feeds were used at aggregate levels consistent with licensing. The model stored trade area metrics and trends rather than individual device data. Access to traffic detail was role-based, and the decision app displayed only the necessary aggregates for footprint choices, with licensing terms referenced in the governance notes.