SELECTED CASE STUDIES
Complex problems. Practical decisions.
These cases show how I reduce operational complexity: what I chose to model, what I deliberately left out, how I weighed the trade-offs and where the change created business value.
SELECTED CASES
The goal is not to make every problem more complex. It is to identify which complexity actually changes the decision.
From a static trigger list to a live production planning system
I replaced an inventory-threshold planning rule with a live decision model built from BOM, WIP, material and routing data so the planner could see what was actually happening on the floor.
THE SITUATION
The old rule identified low stock. It did not explain what the factory could finish next.
Planning started from parts that had fallen below roughly one month of production coverage, with the review often extending out to three months. But priority decisions still required separate checks for what was already on the floor, where each order sat in the routing, how long it would take to finish, whether raw material was available and which stations were constrained.
WHAT I NOTICED
The constraint was not missing data. It was that the data was not connected around the planning decision.
The business did not have a fully integrated system connecting inventory, BOM, raw material, WIP and routing status in one live view. Instead of waiting for a new system, I used the available operational data to reconstruct the floor state manually and turn it into a repeatable planning model.
THE DECISION
Build the live production picture first, then simplify the sequencing logic where the impact was concentrated.
A product-by-product model across roughly 3,000 products would have taken weeks and added complexity that did not materially improve many setup decisions. The analysis showed that more than 80% of the meaningful setup exposure was concentrated in about 27 families, while setup times across much of the remaining population did not vary enough to justify separate logic.
WHAT CHANGED
Planning moved from a static exception list to a live operating model.
The planner no longer had to treat “below target” as the decision. The model could distinguish what was needed from what was already moving, what would finish soon, what could actually run and where sequencing around families or stations could reduce avoidable setup loss.
EVIDENCE
What improved
The improvement was not a better list inside the old planning process. It was a new decision system built around what was actually happening on the floor.
Using visual hierarchy to prevent a high-cost execution error
A small interface change reduced execution errors by making the intended action obvious at the exact moment a trader had to decide.
THE SITUATION
The interface technically showed the right control, but attention was pulled toward faster-moving information.
At execution, the trader was processing asset, quantity, price and a moving market. Buy or Sell was present, but it competed with information that naturally demanded more attention. The resulting error looked simple on the surface, but solving it without slowing the workflow required understanding where attention was going and what kind of intervention would actually change behaviour.
WHAT I NOTICED
The user did not need another instruction. The system needed to make the intended action easier to recognize.
This was not primarily a training problem. The trader already knew the intended direction. The design question was whether the interface could support that decision at the point of action without adding another control, another click or another repeated confirmation that users might learn to ignore.
THE DECISION
Choose the intervention with the strongest risk reduction and the lowest operating cost.
I compared three responses: make the control larger, insert a confirmation step or strengthen the visual state. The third option was selected because it improved recognition immediately without creating a permanent cost on every transaction. That is what made the solution valuable: the implementation was small, but the avoided error cost could repeat every day.
WHAT CHANGED
The control became a visual state instead of another field to remember.
Functional color, grouping and hierarchy made Buy and Sell recognizable at a glance. The process stayed the same, but the system carried more of the cognitive burden.
EVIDENCE
What improved
Simple solutions can have very high returns when they remove the right source of error without adding cost to the process.
Replacing one annual average with inventory logic tied to business impact
I analysed more than 3,000 products so seasonal demand, unit economics and customer exposure could determine where a different inventory rule was actually worth using.
THE SITUATION
One annual average could create both excess cash tied in stock and customer risk.
When demand was below the annual average, the rule could hold more inventory than the business needed. When demand moved above it, the same rule could leave the business short. Overstock tied up working capital and increased financing, storage and obsolescence exposure; understock could trigger expediting, production re-prioritization and service failures that put customer relationships at risk.
WHAT I NOTICED
A seasonal pattern was only one part of the decision. The financial and customer exposure determined whether the pattern mattered.
The same demand shape can represent very different business risk. A product moving 500 units at $1 carries about $500 of annual unit value, while a product moving 15,000 units at $21 represents about $315,000. Customer concentration changes the decision again: a shortage on a strategically important or highly concentrated product can be far more disruptive than the unit count suggests.
THE DECISION
Segment the portfolio by business impact before changing the inventory rule.
Instead of applying a seasonal method to every SKU, I evaluated the portfolio across demand pattern, annual volume, unit cost, customer concentration and resulting inventory exposure. More sophisticated logic was used only where the expected business value justified the additional complexity.
WHAT CHANGED
Inventory targets began to follow both demand behaviour and economic consequence.
The result was not a more complicated formula for every product. It was a more selective policy: seasonal where the data and business exposure justified it, simple where extra sophistication would not create enough value.
EVIDENCE
What improved
A mathematically correct average can still be a poor business rule when it ignores when demand occurs, how much cash is exposed and what a shortage means to the customer.
ABOUT
The projects are different. The pattern behind them is not.
I use business, operations, data and user experience to make processes, systems and decisions easier to understand and easier to use.
See how I got here