Case Study 05 — Supply Chain Control Tower

When the brief was a list of
KPI names, design
wrote the spec.

A control tower for a Fortune-100 semiconductor manufacturer — one landing page, three strategic lenses, sixteen executive-grade KPIs — authored as four paradigmatic templates the rest of the engagement scaled.

Client
Fortune-100 Semiconductor Manufacturer
Industry
Supply Chain — Global Manufacturing
My Role
SME / Lead Designer
Timeline
2024
Stack
Figma · Miro · PowerBI
Engagement
Deloitte Consulting
16+
KPIs Across the Control Tower
3
Strategic Lenses
4
Paradigmatic KPI Templates
Weeks
From Name-Only Brief to Exec Sign-off
01 — The Problem

Complex business. No spec.
Just a list of KPI names.

The client knew they needed an exec control tower. They did not know — yet — what the dashboards should show, what decisions the KPIs informed, or how each team would consume them. The brief was a spreadsheet of metric labels.

01
A list of KPI names — nothing else
The client handed us a brief that was a spreadsheet of KPI names. No definitions, no data model, no consumer personas. Executive decision-making couldn't start until someone invented the spec.
02
Complex business, divergent interpretations
Forecast Accuracy meant one thing to Planning, another to Finance, a third to Ops. Without a shared framing, every team was already building its own private dashboard in parallel.
03
Data engineers couldn't source what wasn't defined
DS was blocked on us: they couldn't pipeline data for a KPI whose framing the stakeholders hadn't agreed on. Months of ETL risk sat downstream of an unresolved design question.
04
Exec bandwidth, not workshop bandwidth
VPs wouldn't sit through three workshops to co-define a metric. They'd react to a prototype in ten minutes. The method had to match the audience — deliverables, not discovery sessions.
02 — Strategic Frame

Design as requirements-elicitation.

When the brief is ambiguous, the instinct is to push for more discovery — more workshops, more interviews, more decks. On this engagement, exec bandwidth didn't allow it, and honestly, another round of workshops wouldn't have produced a spec either.

We inverted the order. Rather than waiting to know what to build, we built paradigmatic prototypes with dummy data for four representative KPIs, then put the prototype in front of the executive sponsor. Within a single review, the pixels surfaced every hidden assumption — what "accuracy" should compare against, whether "met demand" meant units or dollars, which time-windows mattered. The prototype was the requirements document.

With exec sign-off on the four templates, data scientists had an exact spec to source against — no wasted ETL, no rejected framings. The team filled in the remaining roster by applying the same four patterns. Design didn't follow strategy; design unlocked it.

03 — Information Architecture

One landing. Three lenses. Every KPI drills the same way.

The control tower is a three-level hierarchy. The exec sees the full scoreboard on one landing page. A click on any KPI opens a dedicated KPI Overview — headline tiles, trend, breakdown, outliers. Two sibling tabs carry the Data Model and Documentation, so anyone downstream can trace a number to its source. Consistency across every KPI means the exec learns the grammar once.

Tier 1 · Executive
Insights Overview
Landing page — every KPI on one screen, grouped by strategic lens
Revenue GrowthOperating MarginAsset Efficiency
Tier 2 · KPI Owner
KPI Overview
Headline tiles · trend · breakdown · outliers — the exec can read this and decide
KPI OverviewKPI Data ModelKPI Documentation
Tier 3 · Team Detail
Go to Detailed Dashboard
Deep-dive into node-level data, planner working views, and raw source — owned by the team, not the exec
04 — The Landing Page

Sixteen KPIs. Three lenses.
One scoreboard the exec can read in a minute.

Revenue Growth
Forecast Accuracy · Met Demand % · Order Fill Rate · On-Time Delivery · Total Revenue · Expedite %
Operating Margin
Operating Margin · Supply Adherence % · Gross Margin % · Lead Time Variability · Strategic Objectives · Time-to-Market vs Commitment
Asset Efficiency
Inventory Level ($) · Capacity Utilization % · Supplier OTP · Cycle Time
Insights Overview landing page — sixteen KPIs across Revenue Growth, Operating Margin, and Asset Efficiency
Insights Overview — click any KPI tile to drill into its dedicated overview.
05 — The Four KPI Templates

Four KPIs, authored in full.
Every design decision, explicit.

I designed four paradigmatic KPIs end-to-end — headline framing, drill-down logic, and internal data structure. Each one carries a stated executive question and a stated design decision, so the team that scaled the pattern onto the rest of the roster had the reasoning, not just the layout.

KPI 01

Forecast Accuracy

Revenue Growth
Exec question
“Can we trust our demand signal going into next quarter?”
Design decision

Split accuracy from bias from volatility on the same tile row. A single accuracy % masks whether the error is systematic (bias) or stochastic (volatility) — different root causes, different remediation. The drill-down breaks accuracy by market-sub-segment and product type so the planner can see where the forecast is breaking, not just that it is.

Forecast Accuracy — KPI Overview
KPI 02

Met Demand %

Revenue Growth
Exec question
“Are we fulfilling what customers ordered — and is the trend worsening?”
Design decision

Four tiles side by side — CQ, CQ-1, CY, CY-1 — because a single Met Demand % tells an exec nothing without context. The four-way framing puts the quarter-over-quarter swing (am I trending down right now?) and the year-over-year comparison (versus the same period last year) in the same eye-line. The drill-down exposes the highest un-met demand markets and the lowest-performing nodes so operations knows which products to push, not just that we missed.

Met Demand % — KPI Overview
KPI 03

Capital Utilization

Asset Efficiency
Exec question
“Are we using the capacity we're paying for — and is it degrading?”
Design decision

Separate state from dynamics on the same tile row: CY and CY-1 capacity utilization on the left, CY and CY-1 cycle-over-cycle change on the right. Execs and operators ask different questions of the same number — exec wants "where are we now," operator wants "is it changing fast." Same source, two lenses, zero dashboard duplication. The drill-down is a node-level data table with color-coded deltas so the planner can pinpoint the specific fabs trending outside the capacity corridor.

Capital Utilization — Capacity Utilization %
KPI 04

Inventory Level

Asset Efficiency
Exec question
“How much working capital is locked in inventory — and at what stage is it stuck?”
Design decision

Dual lens by design — Dollars for finance and strategy, Units for ops and planning — on one dashboard as tabs, not two separate reports. Finance and ops literally look at the same source when they argue about it. The stage-of-manufacturing breakdown (Wafer Bank → Die Bank → TKDI → SFG → FG) is the real insight: it's not "how much inventory," it's "at what stage of the flow is it getting stuck." Location snapshots + highest/lowest cycle-change callouts turn the insight into a next action.

Inventory Level — Inventory in $
06 — Pattern, Scaled

Four templates. A dozen more dashboards,
built by the team without further design input.

The four KPI templates weren't four deliverables — they were a grammar. Headline tile row, by-period trend, by-segment breakdown, outlier callouts, and a consistent Overview / Data Model / Documentation tab structure. Once the grammar was in place and exec-approved, every additional KPI in the roster was a fill-in-the-blanks exercise.

The rest of the engagement scaled on that pattern: twelve-plus KPI dashboards authored by the team against my templates, plus Data Model and Documentation pages per KPI. The SME deliverable was never 16 dashboards. It was a method the team could execute against.

07 — Outcomes

What the method actually bought us.

Design as requirements spec
The 4 dummy-data prototypes became the requirements document. Exec review on the pixels was faster and more decisive than three rounds of workshops would have been.
Unblocked the data pipeline
DS stopped waiting on ambiguous briefs. With exec-signed prototypes in hand, the team knew exactly which fields to source — zero ETL spent on rejected framings.
Pattern over pixels
4 paradigmatic KPI templates set the grammar for the rest of the roster. The team scaled the pattern across the remaining 12+ dashboards without further design input.
One ontology across three lenses
Revenue Growth, Operating Margin, and Asset Efficiency became a shared vocabulary inside the client. Cross-functional meetings stopped relitigating what each KPI meant.
08 — Reflection

Design isn't always the output of requirements.
Sometimes it's the instrument that produces them.

Strategic frame: When the brief is ambiguous, design isn't the output — it's the instrument that produces the brief. Dummy-data prototypes aren't low-fidelity placeholders; they're the fastest path to stakeholder alignment on what to build.
SME leverage, not SME labor: My deliverable wasn't 16 dashboards. It was 4 templates with enough logical scaffolding that the team could self-serve the rest. A senior designer's output is a method the juniors can execute against.
Separate the question from the metric: Every KPI on the landing page was paired with an explicit exec question it answered. That discipline killed the vanity metrics that usually sneak into control towers and made every tile defensible.