I don't hand over a deck and leave. Every engagement ends with something running in production and a team that can operate it without me.
01 — SIOP design & facilitation
A cadence that survives contact with the P&L
Demand review, supply review, reconciliation, executive sign-off — designed around how your business actually decides, not a textbook calendar. One owner per call, one place the override is written down and later checked against what happened, and metrics defined tightly enough that the number means the same thing in every room. I chair it myself until the leads can, then hand it over. I've designed, implemented and run this three times now, in three different operating models — the third one goes in faster than the first.
02 — Planning & replenishment
The system produces the number
Reorder and production logic that reflects how the business actually sells — segmented safety stock instead of a hardcoded weeks-of-supply, real lead times, real minimum runs, shelf life where it binds — so the supply signal, whether that's a PO or a co-man run, is generated rather than negotiated. Planners get a workbench showing the exceptions, not a spreadsheet to rebuild every Monday.
03 — Network balancing & freight
Where it sits, not just how much
Half the inventory problem is placement. The same units in the wrong building are a stockout in one region and a markdown in another, and the correction ships at LTL rates. I model deployment across the network: what to transfer instead of buy or make, which node should own which item, how full the truck is when it moves, how much life is left on the pallet when it lands, and what a line actually costs to serve from each site. With freight where it is, the placement decision is often worth more than the buying decision. Where the question is large enough to need a solver rather than a rule, it becomes line 06.
04 — Decision rights & data governance
Fewer cooks, on the record
Outside the monthly cadence, the same discipline applied to the daily calls: who may change a forecast, who may override a buy, what gets logged and what gets audited. Fill rate and service defined once, published from one source, so the standing argument about whose number is right simply ends.
05 — Data & automation
Every input on the same clock
One governed data model over orders, inventory and vendors, joined to the outside signals that actually move your category, whatever the category is. Then self-serve tools and AI agents pointed at the reconciliation, chasing and re-keying that eat the week — the work nobody should still be doing by hand.
06 — Network & optimization modeling
When the answer needs a solver, not a spreadsheet
Some network questions are genuinely too large to reason through by hand: which carrier gets which lane across a freight bid, which supplier or co-manufacturer runs which volume, which sites should exist and serve which customers, what moves between nodes instead of getting bought. These are mixed-integer programs — assignment, facility location and network flow with the constraints that actually bind: capacity, minimum runs, commitments, service, shelf life, truck fill. I formulate and solve them in Gurobi and hand back the model, not a slide. What comes out isn't just a recommendation — it's a proven bound on how much better the answer could be, and shadow prices telling you what each constraint is costing you.
07 — KPI frameworks & reporting platforms
A metric is a contract, not a chart
Grain, scope, source and owner, fixed and written down — then a dimensional model underneath so cross-functional questions are joins instead of projects, then a platform architecture that survives concurrent users and can be rebuilt from a repository rather than from a person. Tableau or Power BI where the licence and the self-serve appetite already exist; Streamlit where what's needed is an application a planner works inside rather than a dashboard they look at. Sequence matters: definitions, model, architecture, tool, owner. Most failed reporting programs started at "tool."
08 — Platform selection & implementation
I've bought the software three times
I've led three out-of-the-box planning implementations end to end — ToolsGroup, Netstock and E2open — across different operating models. So this is a service line, not a competitive position: I'll run the selection if you should buy one, run the implementation if you've bought one, or come in when a live implementation has stalled and nobody wants to say so out loud. Requirements written from how you actually plan rather than from a vendor's feature matrix, an honest read on which gaps are configuration and which are the product's model of the world, and the data and definitional work that has to be right before go-live regardless of whose logo is on it.
09 — Sponsor-side work
Where the working capital went
For sponsors and boards: what the inventory is really worth, how much of it will never sell at any price, and how many people are employed to compensate for a system that doesn't work. Opportunity sizing that holds up in a board room, and a plan the operating team will actually run.