Supply Chain & Analytics
Analytics Operating Model
Organisations that have invested in analytics for a few years tend to arrive at a recognisable position: a capable central team, a respectable platform, a substantial backlog, a portfolio of dashboards with thin usage, and an executive who can no longer articulate what the investment has changed.
This is not usually a talent problem or a technology problem. It is a structural one, how work gets prioritised, where analysts sit relative to the decisions they serve, who owns a model once it is in production, and whether anything is ever retired. Those are operating model questions and they are answerable.
How the work runs
Four phases, and what comes out of each.
2–3 weeks
Inventory the decisions, not the reports
We catalogue the decisions that genuinely move cost, revenue or risk (who makes them, at what cadence, on what information) and estimate the value at stake in each. This becomes the map against which the entire analytics portfolio is assessed, and it usually shows a substantial share of output serving no decision on the list.
Outputs
- Decision inventory with owners, cadence and value at stake
- Existing analytics assets mapped to decisions, or shown to map to none
- Usage analysis across the existing dashboard estate
2 weeks
Assess the capability honestly
Skills, tooling, delivery process and the relationship between the analytics function and the business are assessed against what the decision inventory actually demands. The common finding is not a skills gap but a proximity gap, capable analysts too far from the decisions to know which outputs would matter.
Outputs
- Capability assessment against demonstrated need
- Delivery process review from request through to adoption
- Proximity analysis between analysts and the decisions they serve
3–4 weeks
Design the model and the governance
We design the structure (centralised, federated or hub and spoke) against your decision geography rather than against a reference model, and define what the centre owns versus what sits in the business. Then the mechanics that determine whether it works: prioritisation on value at stake, model ownership in production, and a retirement process, which almost nobody has and everybody needs.
Outputs
- Operating model with the centre-versus-business split defined
- Prioritisation framework weighted by value at stake
- Production model ownership, monitoring and retirement process
4–6 weeks
Prove it on real work
The design is run against two or three live pieces of work rather than announced as a reorganisation. This surfaces where it does not fit while changing it is still cheap, and it gives the organisation a demonstration rather than a memo.
Outputs
- Two or three initiatives delivered under the new model
- Adjustments made on the evidence from running it
- Measurement framework so the investment can be assessed next year
What good looks like
The ranges we design toward.
30–60%
Assets retired
Share of the existing dashboard estate typically found to serve no decision on the inventory.
2–4×
Adoption rate
Improvement in analytics output actually used once work is prioritised by value at stake.
0
Vendor commissions
We hold no partnerships and take no referral fees, which is why the assessment is worth reading.
Deliverables
What you are left holding.
- 01Decision inventory with value at stake. The map everything else is judged against
- 02Honest capability and proximity assessment of the current function
- 03Operating model designed against your decision geography, not a reference diagram
- 04Prioritisation, production ownership and retirement processes
- 05Two or three initiatives delivered under the new model to prove it works
What an engagement looks like
A representative shape.
Illustrative. This is the shape of work in supply chain & analytics, not an account of a specific client. We do not publish named results, and we would be wary of a firm that publishes them without the client in the room.
12 weeks
- 01Situation
A consumer goods manufacturer was opening a third distribution centre on the strength of a regional sales forecast, with freight cost per unit rising and no cost-to-serve view below channel level.
- 02What we did
Reconstructed eighteen months of shipment data into a true origin–destination model, built cost-to-serve to customer level, and tested the proposed footprint against demand growth, rate shock and node-loss scenarios.
- 03Where it landed
The third site turned out to be robust only under the most optimistic growth case. Redrawing catchments and re-tendering two lanes captured most of the projected saving without the capital, and the site decision was deferred rather than cancelled.
Questions
The ones we get asked.
Should analytics be centralised or embedded in the business?
It depends on your decision geography, and reference models are unhelpful here. Where decisions are similar across units, a centre builds once and deploys widely. Where they differ substantially by market or product, embedded analysts with far better context outperform a central team. Most organisations at scale land on hub and spoke, a centre owning platform, standards and hard methods, with analysts sitting close to the decisions. The failure mode is picking a structure first and rationalising it afterwards.
How do you measure the return on analytics?
Through the decision inventory. For each decision, estimate the value at stake and the improvement attributable to better information, which is imprecise and still far better than counting dashboards or model deployments. The discipline that matters most is agreeing the measure before building, because attribution after the fact is an argument nobody wins.
We have a big backlog. How should we prioritise it?
By value at stake in the decision each item serves, divided by effort, and everything that serves no decision on the inventory goes to the bottom regardless of who requested it. This is straightforward to describe and politically hard to apply, which is usually why it is not being done. Applying it publicly, with the reasoning visible, is most of the battle.
Do we need MLOps tooling?
If you have models in production, you need the practices, versioning, monitoring, retraining triggers, named ownership. Whether that requires dedicated tooling depends on how many models and how often they change. With three models retrained quarterly, disciplined process and existing tooling is enough. With thirty and continuous retraining, it is not.
Insights
Related reading.
The bullwhip effect: why your demand signal lies
Factory schedules swing by forty per cent while end-consumer demand moves by five. Nobody is behaving irrationally. That is precisely the problem.
Multi-echelon inventory: beyond the safety stock formula
The standard safety stock formula is correct and it is answering a question nobody in a multi-tier network is actually asking.
Cost-to-serve: the analysis most companies skip
Most businesses know their gross margin by customer. Rather fewer know which customers are profitable, and the two lists differ more than anyone expects.
S&OP versus IBP: what actually changes
Most descriptions of the difference are vendor taxonomy. The distinction that matters is whether a trade-off gets settled by someone with the authority to settle it.
Why people systematically under-order high-margin products
The mathematics of a single-period stocking decision is a hundred years old and unambiguous. Experienced buyers still get it wrong in the same direction, every time.
Closely related
More on Supply Chain & Analytics →Data Foundations
The unglamorous layer everything else rests on, pipelines, models, tests and lineage, scoped to a capability rather than to the whole enterprise.
Supply Chain & AnalyticsDecision Science
Allocation, scheduling and network problems solved properly, with the constraints that genuinely bind and a model you can interrogate.
Supply Chain & AnalyticsControl Towers
Control towers, planning system selection and analytics that change a decision, assessed before you sign a licence.
Start a conversation
Talk to us about analytics operating model.
Most useful first conversations are thirty minutes and involve no proposal. Describe the problem you are looking at and we will tell you honestly whether it is one we can help with.
Or write directly to hello@shyenova.com
