Process design, Operations, Controlled vocabulary, Margin recovery, Cross-functionalDiseño de proceso, Operaciones, Vocabulario controlado, Recuperación de margen, Cross-functional

Forty thousand euros, quietlyCuarenta mil euros, en silencio

A quarterly dead-stock process I designed and operate at Eleshop. Catches slow-moving inventory cycle by cycle, routes each item through a shared action vocabulary, and feeds insights back upstream into procurement and content. Sibling intervention to the Eleshop UX case: same diagnosis (structural problems hiding products), different lever (catch what slips through, downstream). Over forty thousand euros in cumulative recovered margin since late 2024.

Un proceso trimestral de stock muerto que diseñé y opero en Eleshop. Detecta inventario lento ciclo a ciclo, enruta cada artículo a través de un vocabulario compartido de acciones y devuelve los aprendizajes aguas arriba a compras y contenido. Intervención hermana del caso UX de Eleshop: mismo diagnóstico (problemas estructurales ocultando productos), palanca diferente (capturar lo que se cuela, aguas abajo). Más de cuarenta mil euros de margen recuperado acumulado desde finales de 2024.

My RoleMi rol

Content Manager by day, process designer on this. Solo. Designed and now run the quarterly cycle end to end: list generation, preparation, action assignment, supervisor review, application, debrief. The intelligence is not a model. The intelligence is a controlled action vocabulary, staged checkpoints, and a feedback loop that turns each cycle into upstream improvement. The design problem was: how do you give a non-technical operator a workflow that catches slow stock early, encodes shared judgement, and improves itself every quarter without burying the operator in tooling?

Gestor de Contenido de día, diseñador de proceso aquí. En solitario. Diseñé y ahora opero el ciclo trimestral de principio a fin: generación de lista, preparación, asignación de acciones, revisión por supervisor, aplicación, debrief. La inteligencia no es un modelo. La inteligencia es un vocabulario controlado de acciones, checkpoints por fases y un bucle de retroalimentación que convierte cada ciclo en mejora aguas arriba. El problema de diseño era: cómo darle a un operador no técnico un flujo que detecte stock lento pronto, codifique el criterio compartido y se mejore cada trimestre sin enterrarlo en herramientas.

Problem Statement

Eleshop carries hundreds of live product listings across specialist electronics categories. A meaningful share turns slowly. Before the process, slow movers were caught by ad-hoc review, often months after the stall began.

Late detection meant late clearance. Late clearance meant tied-up cash, lost margin, and shelf time competing against newer arrivals.

Goals

Surface slow stock on a predictable cadence the team can plan around.
Encode shared judgement so two operators reach the same decision.
Keep a supervisor in the loop without making them the bottleneck.
Track recovered margin and item counts cycle over cycle.
Close the loop: feed cycle insights back into procurement and content.

Constraints

Sole designer / operator. No engineering budget.
Run inside existing systems (PIM + e-commerce admin + Google Sheets). No new vendor.
Must be operable by anyone in the role, not just the designer.
Cannot break list pricing, vendor agreements, or live content.
Sensitive data stays inside Eleshop. This case studies the process, not the catalogue.

Part I: The process

A quarterly cycle the team can plan around. List generation pulls the deadstock candidates from the e-commerce admin into a Google Sheet built from a template. Preparation cleans the data, sorts by stock value, and resolves anything inherited from the previous cycle. Action assigns each row a command from a shared vocabulary. Review hands the sheet to a supervisor who validates judgement calls and resolves the explicit REVIEW flags. Application implements the approved actions through the existing admin tools. Debrief writes the cycle up and feeds insights back into procurement and content.

1. List genPull the deadstock report from the e-commerce admin into a templated Google Sheet. Carry forward last cycle's actions for context.

2. PreparationClean, format, sort by stock value, resolve IGNOREs, fix inherited errors. The sheet is now scannable.

3. ActionFor each item, assign one command from the action vocabulary. Add a remark when context is needed (e.g. promotion %, competitor link).

4. ReviewHand off to supervisor. Supervisor validates judgement calls and resolves explicit REVIEW flags. Returns to operator.

5. ApplicationImplement actions through the e-commerce admin and PIM. Catalogue ops workflow unchanged.

One cycle of the deadstock process. Sensitive specifics live inside Eleshop.

Part II: The action vocabulary

The heart of the process is a controlled set of action commands. Each row in the cycle gets exactly one command and an optional remark. A small fixed vocabulary that everyone in the chain understands is faster to write, faster to review, easier to audit, and produces inter-rater agreement that free text never will.

The vocabulary has sixteen commands grouped by intent. Below, four representative ones that show the range, from upstream procurement levers to escalating responses to the human-judgement override.

Upstream leverDEM. Switch the item from "maintain stock level" to "purchase on demand." Catches the leak at procurement, before the next batch arrives.

Staged responsePRMTN-1, PRMTN-2. First cycle: modest promotion. If still stalled the next cycle: deeper discount toward cost. Escalation is a path, not a single choice.

Exit leverRETURN. Send the item back to the supplier. Asymmetric: bundled annually in Q1 rather than every cycle (see Decision 2 below).

Judgement overrideREVIEW. Operator is unsure. Assigns the row to the supervisor with a remark. The vocabulary admits its own limits.

Decision 1: Controlled vocabulary vs free-text actions

Option A

Free-text "what to do" field

Operator writes a sentence per row. Maximum flexibility, zero consistency. Two operators reach two different conclusions on the same item. Auditing a quarter takes hours.

Option B

Three coarse buckets

Sell / hold / return. Easy to learn but too coarse to encode real choices (upstream procurement change vs price drop vs disposal are not the same).

Option C

Sixteen-command vocabulary + remark field

Each command is a distinct lever (upstream, price, content, exit, escalate). The optional remark holds the specifics. Vocabulary is closed so reviewer can scan a sheet in minutes.

Why: The whole point of the cycle is to encode shared judgement so two people reach the same answer. Free text destroys that. Three buckets are too coarse to capture the real moves available. A medium-sized closed vocabulary plus a remark field is the smallest design that still expresses the catalogue's real choice space.

Decision 2: When to action RETURN. Every cycle or once per year?

Option A

Every cycle, individually

Whenever an item is flagged RETURN, contact the supplier. Faster recovery but high per-return overhead (negotiation, packaging, shipping coordination, supplier patience).

Option B

Never, always discount instead

Skip the return path. Simplifies the vocabulary but loses recovery on items where the supplier relationship actually supports returns.

Option C

Bundle annually, in the Q1 cycle

Collect every RETURN action across the year, group by brand, negotiate one annual return per supplier. Items still get marked when first identified. The action queues; the execution batches.

Why: Returns have meaningful per-instance overhead and many suppliers will accept one well-prepared annual return but reject a string of small ones. Batching gives the operator and the supplier predictability. The cost is a few months of carrying cost on items already condemned to return. The trade is usually positive.

Part III: The feedback loop

The cycle is not just an inventory clean-up. It is the most honest data the team has about which products, brands, and procurement choices are quietly costing money. The debrief at the end of each quarter exists to convert that data into upstream change.

Each debrief surfaces three classes of pattern. Sensitive specifics are not in this case study.

Procurement signalBrands or categories that keep ending up on the list cycle after cycle. The signal: change the default purchasing strategy for the next batch, not the discount on the current one.

Content signalListings flagged LIST-ERROR repeatedly. Wrong visibility, missing categories, missing photos. The signal: a category-level content audit, not a per-item fix.

Brand-level signalAn entire brand under-performing. The signal: a brand-level review with sales and procurement. Sometimes the answer is to stop carrying the brand.

The cycle ends by changing the inputs to the next cycle. That is the systems thinking, not the spreadsheet.

The same diagnosis that drove the Eleshop UX case ends up here too: structural problems hiding products on the homepage become "LIST-ERROR" patterns in the deadstock debrief. The redesign is the upstream fix. The deadstock process is the downstream catch. One root, two interventions, both pulling on the same lever from different ends.

Impact HighlightImpacto

€ 40,000+ in cumulative recovered margin since late 2024.
Detection lag dropped from "months, when a person noticed" to a predictable quarterly cadence.
Controlled action vocabulary makes inter-operator agreement measurable for the first time. Audit time dropped from hours to minutes per cycle.
Debrief loop converts inventory clean-up into upstream change. The deadstock process now drives procurement and content decisions that shrink the next cycle's input.
Zero new tooling. Runs inside the existing e-commerce admin + PIM + Google Sheets stack. Operable by anyone in the role, not just the designer.

40.000 €+ de margen recuperado acumulado desde finales de 2024.
El retraso de detección bajó de "meses, cuando alguien se daba cuenta" a una cadencia trimestral predecible.
El vocabulario controlado de acciones hace que el acuerdo entre operadores sea medible por primera vez. El tiempo de auditoría bajó de horas a minutos por ciclo.
El bucle de debrief convierte la limpieza de inventario en cambio aguas arriba. El proceso de stock muerto ahora informa decisiones de compras y contenido que reducen el input del próximo ciclo.
Cero herramientas nuevas. Corre dentro del stack existente de e-commerce admin + PIM + Google Sheets. Operable por cualquiera en el rol, no solo por el diseñador.