The definition of cost has to be settled first
When different people give different answers for what one process step costs, every optimisation after that is built on the wrong base.
Industrial FDE · through cost
Industrial FDE means running forward-deployed delivery inside a manufacturing plant. There are two ways in: through data and models, or through money and process steps. Yidao goes in through the second one — first work out what every machine and every process step actually costs, then decide where a tool is worth building.
Definition
An FDE is someone who understands the business, walks onto the client site, pulls out the work worth fixing, turns it into process, matches it with resources, codifies it into a tool, and stays to keep it working. Do that between the ledger and the shop floor and you have industrial FDE.
In the same plant, the two ways in produce completely different first months. They are not mutually exclusive, but they have an order: until the cost structure is clear, a model does not know which step to optimise.
| Way in | Where it starts | First-stage output |
|---|---|---|
| Through models | Start with data and models: govern the data, train the model, then look for a scenario it fits | A model or system that runs |
| Through cost | Start with money and process steps: push costs down to cost objects, then find the most expensive step | A cost structure that reconciles to machines and process steps |
Why cost
Going in through cost does not call for modelling skill. It calls for the ability to push money down to the process step, and to make the result stand up in the ledger and against the vouchers. Those are our own trade.
| Capability | Where it comes from | Where it is used on site |
|---|---|---|
| Cost reconstruction | Manufacturing cost accounting and material-labour-overhead reconstruction, the founding team's own trade | Build cost objects, tie energy use and output to machines and process steps |
| Evidence and definitions | Ten-plus years in tax controversy and internal audit | Every cost figure must point back to a line on a source document |
| Codifying it into tools | Six in-house tax and finance tools, eight enterprise AI modules | Automate what can be automated, write the rest up as work procedures that stay |
When different people give different answers for what one process step costs, every optimisation after that is built on the wrong base.
A number without an index does not go into the conclusion. That is the internal audit rule, and it applies to cost: to be accepted, it must first be checkable.
Reverse the order and you get a tool nobody uses. Prove which step is most expensive, then decide whether that step is worth a module.
Three hard problems
These three are not ours. They come up again and again whenever industrial FDE is discussed. What we do about them is in the third column.
| The problem | Why it is hard | What we do |
|---|---|---|
| A senior FDE costs too much for a mid-sized plant | Keeping a hybrid specialist on site is an annual cost that does not match the return on a single project | Priced by person-day and by role, delivered module by module. Start with one cost review; no need to carry a person first |
| The knowledge sits with the old hands and on paper, not in structured data | Data governance first means a long run with nothing to show along the way | The 109 standard work units are that kind of experience already codified: there are executable work procedures from day one |
| When the project ends and the people leave, the capability leaves too | The deliverable is a bespoke system or a report, with no reusable unit | The unit of delivery is sixteen product modules and standard work units. What stays is procedure and working papers, not a one-off result |
We do not promise a percentage of cost or tax saved. The third column is about how we deliver, not a guarantee of results; results are accepted against the plant's own business metrics, with scope, input and acceptance criteria set before work starts.
What going in through cost looks like
Five stages, each with something you can look at. If one stage does not hold up, the next one does not start.
| Stage | On site | What comes out |
|---|---|---|
| Review | Take one full period of cost, output and process data, build cost objects | A cost structure: fixed and variable split apart, unit cost comparable |
| Reconstruct | Rebuild material, labour and overhead; tie energy use and output to specific machines and process steps | What every process step and every machine actually costs |
| Locate | Find the most expensive stretches; judge which yield to process change and which need a tool | An improvement list ranked by value and feasibility |
| Codify | Turn what can be automated into modules, write the rest up as procedures | Modules plus working papers the plant can run itself next period |
| Accept | Compare before and after against the metrics agreed up front | An acceptance conclusion in which every figure traces back to a voucher |
For where FDEs at other companies entered the shop floor, the FDE casebook (in Chinese) collects twenty-four external cases, each with a discussion thread. They are not Yidao projects and are there only as a reference.
Already proven
All three figures come from completed engagements, described by industry and province.
50.9%
Electroplating and surface treatment · Jiangsu. Twelve months of daily data: splitting all cost into fixed and variable showed more than half does not move with output — which reverses where the savings lever points.
51%
Same plant: output up 4.6 times, electricity cost per unit down 51%. The conclusion was sitting in the client's own reports; nobody had ever worked it out for the owner.
97%
New materials · Jiangsu. The real structure measured during an internal audit, quantified and priced with the full set of coded working papers, with a remediation path.
The method holds. What remains is scale.
What this is not
These five are the usual comparisons. The line between them is not which is better, but which end each one starts from.
| Often confused with | What it solves | Where the line is |
|---|---|---|
| Shop-floor systems / MES | Makes production visible and schedulable | It collects process data; going in through cost turns that data into money that reconciles to the vouchers |
| Lean consulting | Removes waste through improvement on the floor | It changes the motions; cost first proves which stretch is most expensive, and the change still has to be accepted in the ledger |
| Audit firms | Issues reports and opinions under the standards | They are answerable outward for a compliance conclusion; cost is answerable inward for the definitions management decides on |
| ERP cost modules | Collects and allocates under rules already set | Whether the rules themselves are right is not its call; rebuilding cost objects and cost drivers is exactly the work |
| Industrial AI, the model route | Optimises process and scheduling with models | It needs a clear objective; going in through cost works out what that objective is worth |
None of the five is a competitor. On most engagements the cost route makes the structure of the money clear first, and then the others know where to push.
Questions
Industrial FDE means running forward-deployed delivery inside a manufacturing plant: someone who understands the business goes in, pulls out the work worth fixing, turns it into process, matches it with resources, codifies it into a tool, and stays to keep it working.
There are two ways in — through data and models, or through money and process steps. Yidao goes in through cost.
An implementation consultant delivers features against a spec; on-site development delivers code against hours.
An FDE is answerable for the business result: the unit of delivery is product modules and standard work units, and acceptance is measured on the plant's own metrics, not on a checklist of features.
The model route governs data and builds a model first; its first-stage output is a model that runs. The cost route pushes cost down to cost objects first; its first-stage output is a cost structure that reconciles to machines and process steps.
They are not mutually exclusive, but they have an order: until the cost structure is clear, a model does not know which step to optimise.
Yes, and that is the most common starting point for the cost route.
Getting started needs one full period of cost, output and process data. Most plants have all three; what is missing is the link between them.
We do not station someone for a year. Work is priced by person-day and by role, and delivered module by module.
Start with one cost review, see the structure, then decide how much to do next.
The first stage is a cost review; how long it takes depends on how complete the data is.
We do not promise a percentage. Scope, input and acceptance criteria are set before work starts, and at the end we compare before and after on the plant's own metrics.
What stays in the plant is modules and working papers, not a one-off result.
Procedures are written as standard work units your own people can run next period. That is why the unit of delivery is a module and not a project.
One conversation, forty minutes, about the money in your plant you currently cannot account for.
If it is worth doing, we start with one cost review.
More questions? The FDE Q&A collects thirty frequently asked questions, each answered in its first paragraph.
In manufacturing, the most expensive money is often not on the quotation. It is between the ledger and the shop floor. Working it out is the first step of industrial FDE.
Yidao Financial Consulting · Industrial FDE
Tell us the facts and we will tell you what kind of problem it is, how it is normally handled, and how far it can realistically be pushed on the evidence you have.