Electronics Retail.
Consumer electronics chain. ~90 stores, 14 Italian regions. SAP ECC, Toshiba POS, Adobe Commerce and an in-house promotional tool unified beneath a single commercial event model. No system replaced.
A Rome-headquartered consumer electronics and household goods chain operating ~90 stores across 14 Italian regions was watching its gross margin contract while revenue held steady. Promotional events were cannibalising full-price categories, late-cycle markdowns were consuming margin that should have survived the promotional window, and supplier rebate reconciliation was running months behind actual posting. A prior consultancy had produced dashboards, not decisions, and had concluded that SKU-level margin visibility across channels could not be modelled without a unified data layer that did not yet exist. A unified commercial event model replaced four disconnected data flows — SAP ECC, Toshiba POS, Magento, and an in-house promotional tool — with a single source of record for what constituted a promotion, a margin event, and a markdown reason. The decision cycle on margin-impacting price moves compressed from 9–14 days to under 60 seconds. Net margin contribution per promotional event improved by 15–22%. Late-cycle markdown volume on promotional SKUs fell by 25–35% in the first full year of operation.
The operation.
The network runs SAP ECC 6.0 as its core ERP — an older on-premise release common among Italian mid-market retailers that upgraded in the 2010s and have not yet migrated to S/4HANA. Alongside SAP, Zucchetti payroll and Italian regulatory compliance modules handle HR, fiscal reporting, and the country-specific electronic invoicing obligations introduced under the fatturazione elettronica regime.
Store operations run on a Toshiba TCx Sky POS deployment rolled out during the chain's 2019–2021 modernisation cycle. E-commerce operates on Adobe Commerce (Magento 2), with a separate, custom-built promotional planning web tool written internally in 2017 and extended through 2022. The warehouse and distribution network uses Replica Sistemi, a regional Italian WMS common in mid-sized distribution operations.
Each of these systems was functional in isolation. The entity SKU existed in all of them — but it existed as a different object in each: SAP's article master carried procurement cost and rebate entitlements; the POS carried transaction-level margin at register; Magento carried online price history; the in-house promotional tool carried the intent of the promotional event but not its measured result. Margin was calculable in each silo and not calculable across them.
Mid-market retailer, ~90 stores across 14 Italian regions. 14-person commercial / pricing / supply-chain / IT team engaged.
Four temporal frames. No shared spine.
The visible problem was margin erosion. Gross margin on consumer electronics categories had been drifting down across 2023 and into 2024, while revenue held roughly flat. Promotional volume had increased — partly in reaction to competitive pressure from the national chains that dominate Italian consumer electronics retail — but the net margin contribution per promotional event was declining. Late-cycle markdowns on promotional SKUs had grown, and supplier rebate reconciliation was running four to six weeks behind actual posting, creating working capital drag.
Underneath the visible symptoms, the environment had no object representing a commercial event in a way that preserved its causal structure. A promotion on TV category X and a margin collapse in soundbars two weeks later were not linkable in any production system. The decision to mark down a slow-moving SKU and the decision to launch a competing promotion on an adjacent category were made by different teams on different timelines against different data extracts.
The previous work had correctly identified that the data was incoherent. It had not been structured to make the data coherent. Dashboards describe. They do not define. Additional analytics investment against the same fragmentation compounds it.
Stathon engagement assessment
The commercial director needed to know, before authorising a promotional event, what that event would do to margin in adjacent categories — not after the fact, not in next month's report, not at a category aggregate, but at the SKU and sub-category level, in time to change the promotion's structure. That question could not be answered by any system in production at the start of the engagement. It was the one question whose answer would justify everything else.
No object representing a commercial event with preserved causal structure. A promotion on TV category X and a margin collapse in soundbars two weeks later were not linkable in any production system. Each silo recorded fragments; none recorded the connective tissue.
The entity SKU existed in all systems — but as a different object in each. SAP’s article master carried procurement cost and rebate entitlements; the POS carried transaction-level margin at register; Magento carried online price history; the promo tool carried event intent but not measured result. Margin was calculable in each silo and not calculable across them.
Margin reported monthly. Promotions planned weekly. Pricing adjusted daily. The POS processing transactions every second. Four temporal frames operating against four different data extracts, with no shared spine to relate them.
The decision to mark down a slow-moving SKU and the decision to launch a competing promotion on an adjacent category were made by different teams on different timelines. Cross-category cannibalisation was invisible until the next monthly margin report — weeks after the decisions could still be reversed.
Three phases.
Forge-tier deployment. Four modules — Arché, Core, Athena, Aegis — deployed in phased sequence. SAP ECC remained in place, unchanged. The Toshiba POS was not modified. Adobe Commerce was not migrated. The in-house promotional planning tool was not rebuilt.
Definition & Integration Spine
Weeks 1–6 · Aug–Sep 2024Nothing was live-facing in Phase 1. No alert went to a user. No report changed. The commercial team continued operating against their existing tools. What was built was the spine that would carry the Phase 2 capability.
Stathon deployment record
Margin Integrity Monitor
Oct 2024 – Jan 2025Athena's rule-based escalation logic layered beneath a probabilistic scoring engine tracked active and planned promotional events against the unified SKU-level margin object. The Monitor operated in advisory mode for the first 90 days. No automatic price action, no blocking of promotional approvals — every flag went to the commercial team, who reviewed it and either accepted or overrode. Every override was logged and fed back into the tuning cycle.
Initial false positive rate ran approximately double the target threshold. Roughly a third of flagged cannibalisation events were seasonal substitution effects — consumers shifting purchase intent between air conditioning categories in late spring, for example — that the model had not yet learned to exclude. Two tuning cycles over a 45-day window brought the false positive rate into operational range.
During the second tuning cycle, a concrete flag grounded the capability: in pre-planning for the 2024 Black Friday cycle, the Monitor flagged that a planned 20% promotion on a mid-tier LED TV line would cannibalise an estimated €1.6–2.0 million of full-price sales in the soundbar sub-category over the following three weeks. The commercial team restructured the promotion into a bundled offer that preserved the bulk of the expected margin exposure.
Three regional category managers — particularly in white goods categories where promotional planning was historically driven by supplier-co-funded campaigns — resisted through October and November. Calibration sessions in December and January, walking through specific flagged events against prior-year actual margin outcomes, closed the adoption gap. By end of January 2025, the Monitor's flags were being reviewed on every major promotional event across the network.
Expansion & Roadmap
Feb 2025 onwardA markdown timing advisor moved into advisory-mode production, recommending optimal markdown initiation points for slow-moving and end-of-promotion SKUs. Network rollout completed in September 2025 after a six-week calibration extension for the southern regional cluster. A supplier rebate reconciliation engine entered pilot in March 2025 and remains in pilot at the client's finance-governance preference.
Workflow state model linking inventory position, promotional predecessor events, and category-level seasonal demand curves. Operates as an advisory layer to the inventory managers, not an automated price-change authority. Live across all ~90 stores and the e-commerce channel.
Reconciles contractual rebate entitlements against SAP ECC postings and flags discrepancies for review. In March 2025 surfaced a ~€340,000 discrepancy that on review was a mis-filed contract amendment never posted into SAP's vendor master. Remains in pilot — finance retains a quarterly review gate before authorising production-mode use.
Tuning continues against southern-region markdown behaviour patterns that deviate meaningfully from the national baseline. The on-site calibration extended the network deployment by approximately six weeks beyond the Lazio cluster timeline.
What changed.
Measurement window: first promotional cycle post-deployment (Q4 2024) through first full annual cycle completed (Q1 2026). Comparison baseline: 2023 full year and H1 2024 full-price-versus-promotional margin performance. SKU-matched, with seasonality adjustment applied.
The client received structural capacity, not software. What the previous engagement had correctly diagnosed — that the data was incoherent — was not resolvable at the dashboard layer. It was resolvable only once an entity existed within the infrastructure that could represent a commercial event across the systems that had never before shared a definition of one.
The 1,100–1,250 hours per month freed from monthly margin reconciliation, post-event promotional forensics, and manual rebate variance investigation have not been eliminated. They have been redirected from backward-looking reconciliation to forward-looking promotional design.
What this infrastructure is.
Not a tool. Not a platform.
The Stathon infrastructure is the operational logic layer beneath a retail network that, before the engagement, did not possess a shared definition of what constituted a commercial event across its own systems. SAP ECC was not replaced. The Toshiba POS was not modified. Adobe Commerce was not migrated. The in-house promotional planning tool was not rebuilt. None of what the organisation had built was displaced.
Definitional layer beneath every system.
Every one of those systems now operates beneath a definitional layer that can represent a commercial event as a single object with causal structure, a continuity layer that guarantees the event’s record does not drift across extraction cycles, an inference layer that can model the event’s projected consequences before it occurs, and a sovereignty layer that governs who — human or model — can query any part of it.
Dashboards describe. Definitions decide.
A 2023 engagement with an external consultancy had produced a Power BI dashboard layer on top of the same nightly exports and concluded that predictive margin modelling was not feasible without a unified SKU-level margin object. Additional analytics investment against the same fragmentation compounds it. The fragmentation is resolvable only at the definitional layer.
The integration is not visible from the surface. Its absence would be.
Stathon deployment conclusion
Forward roadmap.
Promotional planning simulator
Forward modelling of promotional structures against projected margin impact before commercial authorisation. Decision problem: restructure promotions before launch rather than flag them after scheduling.
Channel-integrated pricing governance
Unified pricing decision authority across store, e-commerce, and marketplace channels. Decision problem: eliminate cross-channel price inconsistency that currently drives unplanned markdown.
Inventory receipt anticipation model
Pre-arrival markdown risk assessment against incoming supplier shipments. Decision problem: surface over-ordering before goods arrive rather than after they enter the warehouse.
Deployment record.
Stathon · Definitional Infrastructure Company. Client identity withheld by agreement. Deployment metrics reflect production conditions as of April 2026.