Approach
Decision first, baseline second, everything else after that.
Most supply chain analyses fail in one of two places: they answer a question nobody was deciding, or they rest on a baseline that does not survive scrutiny. The way we run engagements is built around avoiding both.
Engagement phases
Four phases, in this order, every time.
Durations shift with scope. The sequence does not, and the baseline phase does not get compressed to make room for analysis.
Scope
1 to 2 weeks
We write down the decision the work has to inform, and what evidence would change it.
Scoping produces a list of questions, not a list of workstreams. If we cannot state what decision an analysis would change, that analysis does not go in the plan.
We also do a data readiness check here, because the honest constraint on most engagements is what can be reconstructed from your systems rather than what is analytically possible.
Phase exit criteria
- Decision statement agreed in writing
- Data availability assessed before commitment
- Success criteria defined as a decision, not a deliverable
- Explicit list of what is out of scope
Baseline
2 to 4 weeks
We rebuild current state in the model and reconcile it against what finance reports.
This phase is longer than most proposals allow for, and it is the one that determines whether anyone believes the result. The model, run against last year of actual activity, has to reproduce last year of actual cost by category rather than in total.
Reconciliation reliably surfaces problems worth knowing about independent of the study: lanes billed to the wrong site, stale product dimensions, accessorials nobody allocates, customers served from a facility the master data says does not serve them.
Phase exit criteria
- Baseline reconciles by cost category, not just in total
- Every material variance explained rather than absorbed
- Finance signs off before scenarios begin
- Data quality findings documented and handed back
Analyze
3 to 6 weeks
Optimization, simulation, and scenarios run against the agreed questions.
Findings are reviewed in working sessions as they emerge. There is no final reveal, because a result that surprises the operating team in a steering committee is usually a result that missed a constraint.
Every number carries the assumption it depends on and the range it moves across. Where two assumptions matter more than the rest, we say so explicitly rather than burying it in an appendix.
Phase exit criteria
- Working sessions instead of a staged reveal
- Sensitivity ranges reported alongside every headline figure
- Constraints validated with the people who operate them
- Alternatives kept on the table until the evidence closes them
Hand over
1 to 2 weeks
You get the model, the documentation, and the cadence that keeps it current.
The engagement is not finished when the recommendation is accepted. It is finished when your team can rerun the model against new data and get an answer they trust.
That includes the unglamorous parts: refresh schedules, named owners for parameters that decay, and a written note on where the model is weak and should not be pushed.
Phase exit criteria
- Model, code, and data pipeline transferred to your environment
- Documentation covering assumptions and known limitations
- Training sessions with the team who will run it
- Refresh cadence and parameter ownership agreed
Commitments
What we will and will not do.
These are not values on a wall. They are the things we get asked about in second meetings, so they are written down.
The people who scope it, build it
There is no handoff from a sales team to a delivery team, and no rotating bench. The person who scopes your engagement is the person writing the model.
No phase two attached to the readout
If follow-on work is genuinely warranted we will say so, once. A final presentation that doubles as a proposal compromises the recommendation inside it.
We will tell you not to do it
Some questions do not justify the analysis, and some are already answered well enough by a spreadsheet. Saying so early costs us a project and saves you a quarter.
Ranges over false precision
A single confident number with no sensitivity behind it is a marketing claim. We report what moves the answer and by how much.
Working arrangements
How engagements are staffed and priced.
Most engagements are fixed scope and fixed fee, quoted after scoping rather than before it. We would rather spend two weeks understanding the problem and quote accurately than quote fast and manage the gap with change orders.
Teams are small, usually two to three people, and they are the people doing the modeling. Where an engagement needs deep sector or systems knowledge we do not have, we say so and bring in a specialist rather than learning on your budget.
Some clients keep us on a retained basis after handover, a few days a month to maintain models and pressure test decisions. That is offered where it is genuinely useful and not proposed by default.
Related