The data layer underneath
Where an order stops when the data is not clean enough to act on.
Matching, master data, document intake, and the tail of exceptions nobody owns. This is the least glamorous of the three things we do and the one that most often has to happen first — because a system built on a catalogue nobody can search inherits the problem rather than solving it.
Most people arrive here for one of three reasons: there is an open vacancy for order entry and nobody wants to fill it again, the vendor already in place is not covering what was promised, or the catalogue has become impossible to search.
The remainder is the part that needs a judgement.
Every tool in this category quotes a share it handles automatically. Take the number at face value — the question is the rest. In industrial distribution the rest is the line that describes a part instead of naming one, the customer who writes wie letztes Mal, the unit of measure that does not match how you sell it, and the substitution that needs an engineer rather than a parser.
When waiting stops being acceptable, the line goes out on somebody's judgement — the standard unchecked, and no record of who decided. That is the expensive part, not the wait. It is the thing an EN 10204 3.1 certificate exists to make impossible.
We work on that remainder — the lines that cannot be answered in seconds. The people who answer them keep answering, with the standard already checked, the source attached and the decision recorded.
What sits in this layer
Entity resolution against technical specifications — deciding which item a line describes when the line describes it rather than naming it.
DIN and ISO standards, tolerances, materials, pressure classes — the attributes that decide whether a substitution is correct, and the rules that say who is allowed to make it.
Cross-references between supplier catalogues — including the ones where a competitor's number maps to two of yours and neither is exact.
Master data normalisation — turning an item file that grew for fifteen years into one that can be matched against.
Document intake from email, parsed into structure — attachments, forwarded threads, scans, spreadsheets that were built by a person for a person.
An exception queue with routing and rules — where everything above sends what it could not finish, with the question already formed.
Onboarding new sender formats — because the next customer will send something the last five did not, and that has to be a 2-day job rather than a project.
How the work runs
- Understand — 1 week on your actual stream. How many formats are really in it, where it breaks, and what tail will still be there in a year.
- Design — the attribute model for your catalogue, and the rules that decide when a substitution is allowed. Those rules come from an engineer, not from purchasing.
- Build — intake, decomposition, matching, the queue.
- QA — a fixed set of lines with known correct matches, run before release and again whenever the catalogue or the rules change. It catches the rule that fixed one family and broke another.
- Integrate into your landscape — into your ERP as it is configured today. Your ERP stays as it is configured; we deliver into it. Integration is planned first and priced first.
- Hand it over — the repository, the documentation, the configuration, the prompts, and the corpus of resolved exceptions. Yours, with no lock to us and none to a model provider.
- Support — the tail routed and on a clock rather than swept into a shared mailbox, and the rules corpus still growing.
The whole flow, and where it stops
| Where it stops | What closes it |
|---|---|
| The document does not read | Document intake — a new sender's format is onboarded, not agreed with him first |
| The line describes the part instead of naming it | Specification matching |
| Two candidates, and neither is exact | Cross-references, and the substitution rules that decide between them |
| The answer needs a person | The exception queue |
| Answered once, and not remembered | Rules written out of the exceptions that closed |
Every place the flow stops has an answer under it, named in words. This is the one figure only a hub can carry: the pipeline whole, and the reader's own place on it.
What it costs
| Package | Duration |
|---|---|
| C0 · Five-order teardown | 3 days |
| C2 · Flow diagnostic | 1 week |
| C3 · Pilot on one flow | 3 weeks |
| C4 · Project | 8–12 weeks |
| C5 · Specification matching | 6–10 weeks |
| C6 · Exception Desk | retainer |
Start at the top of that list. Five real orders, 3 days. You get them parsed, and an honest account of what did not parse and why. It is credited against a project within 60 days.
The packages are cut along engineering boundaries, not price tiers. That is why one of them is 3 days on five documents and another is a standing arrangement: they do different things to different parts of the same pipeline.
The first call is free: we run two or three of your documents on screen and tell you what we see. What leaves that call is an opinion, not work — engineering time starts at the first paid step.
Deeper on any of these
Specification matching — how a line becomes eight attributes, four of which were never written down. Our flagship.
The exception queue — five kinds of exception sorted by what closes them, and why two of the five will never go to zero.
What we can show you, and what we cannot
Catalog Brain is our own system for industrial distribution, and specification matching is its proven capability.
Order intake is a demonstrator. We have built it and we will run your documents through it. It has not yet run a live order stream at a customer, and we would rather you read that here than discover it on the third call.
We would rather spend 3 days on your own orders than hand you someone else's reference.
What this layer assumes
This layer exists for everything arriving as email, PDF, spreadsheet and free text. Portal and EDI traffic already carries your part numbers and is already matched.
The arithmetic starts working above a few dozen exceptions a month. At a few dozen, a named owner and a good spreadsheet beat us on cost, and we say so on the call.
Your ERP stays as it is configured. We deliver into it.
Commercial and engineering decisions stay with your people. Price deviations and substitution questions route to them, always.
What is sold is a system that runs, not people who are available. Scope sets the price; retainers are priced by volume handled. If what you need is capacity at desks, that is cheaper elsewhere.
A system built from zero usually contains agents, and stands on a layer like this one that had to be built first.
Forty-five minutes. Bring the last five orders that came back to a person — the messy ones, as they arrived. We will run them on screen and tell you which of them will ever go away.