Business intelligence

Figures that carry a decision

BI is bought as reporting and expected to be steering. In between lies the difference between a screen and an action.

The dashboard

Four figures that carry a decision

Four numbers, two charts — and for every number, what it is for and how it is calculated.

Empty running 18.4% +3.1 points against last year
Orders per trip 2.7 +0.2 — the lever behind it
To the credit note 11 days −5 days since January
On-time rate 94.2% unchanged for three months
Where the empty running actually sitsThe company figure of 18.4 % comes from two lanes, not six. Only those two carry a measure.
limit 18% Vienna – Hamburg Vienna – Hamburg: 24 % 24 % Linz – Rotterdam Linz – Rotterdam: 21 % 21 % Graz – Munich Graz – Munich: 17 % 17 % Salzburg – Milan Salzburg – Milan: 15 % 15 % Wels – Prague Wels – Prague: 12 % 12 % Innsbruck – Zurich Innsbruck – Zurich: 9 % 9 %
below the limitabove the limit — this gets looked atlimit 18%

Example data. The limit is set, not measured — and drawn in for that reason: a traffic light whose limit nobody knows is a colour without information.

From consignment note to credit noteThe released document moves the money, in both directions. Every day before that is money standing still.
0 5 10 15 20 agreed: 12 days Jan: 16 days Jan Feb: 15 days Feb Mar: 15 days Mar Apr: 14 days Apr May: 14 days May Jun: 13 days Jun Jul: 13 days Jul Aug: 12 days Aug Sep: 12 days Sep Oct: 11 days Oct Nov: 11 days Nov Dec: 11 days Dec 11 days

Example data. Only the last point is labelled — a number on every one turns the trend into a table with a line.

MetricWhat it is forHow it is calculated
Empty runningWhere capacity is paid for and not sold. Planning can act on it — per lane, never on the company figure.Empty kilometres ÷ total kilometres, per lane and month. Kilometres from telematics, not from route calculation.
Orders per tripThe lever underneath: two shipments on one trip double the revenue, not the effort.Completed orders ÷ trips driven. Cancellations do not count — otherwise the figure rises when things go worse.
Days to the credit noteThe consignment note moves the money. Whatever sits here sits on both sides: the sub-carrier waits, the customer pays against the document.Calendar days from upload to release, median rather than mean. One dispute running six weeks moves the mean.
On-time rateWhat the customer notices. No daily decision — the question of whether a price holds.Shipments inside the agreed window ÷ all. The window is in the contract, not the report.

A report says that. What follows from it is on KPIs and AI — drag it through twelve months.

Four numbers, four recipients, four decisions. A fifth needs a fifth recipient — otherwise it belongs to nobody.

Explainer · about 35 secondsHow forty metrics become four

    It starts with the collection: forty figures, every one of them correct.

    Starts on its own · pause any time
    State of the art

    The question before the dashboard

    Which decision gets made differently? If that is hard to answer, what is missing is not the report — it is the recipient.

    When the answer is hard

    A dashboard for “management” re-plans no trip.

    When it is easy

    The metric reaches someone who can act the same day. That is the whole difference.

    A metric with no decision behind it is decoration — and it costs as soon as somebody maintains it.

    Measured ≠ calculated

    Recorded and estimated look identical in the report. Two fields instead of one save the first dispute.

    Data quality is not a pre-project

    Master data decays continuously. What helps is a point where an error surfaces — not a spring clean.

    No model repairs it

    A broken master record does not become correct through AI. Only more articulate.

    A metric only tells you that something is the case: 18 per cent empty running. Why it is rising, and who has to act now, is nowhere to be seen. On KPIs and AI it stands right beside it — the explanation next to the figure, the next step below it. And how an AI learns this before it is allowed a say is explained under shadow mode.

    Explainer · about 35 secondsFrom a number to a decision

      A metric comes from closed cases — only there is the outcome actually known.

      Starts on its own · pause any time
      Worked example

      Worked example: the empty-run rate

      A metric everyone has and almost nobody calculates the same way — three companies, three numbers.

      TMS ERP Telematics one definition one number
      Three companies, three numbers: all empty kilometres, return legs only, or any trip below 30 per cent utilisation. All defensible, no two comparable.

      What is missing

      Name, definition, owner — and the provenance: telematics or route calculation?

      What actually counts

      Not the level, the movement. 18 or 23 per cent says little; rising for three months says a lot.

      A metric everyone calculates differently is not a metric. It is an agenda item.

      Tools, briefly placed: Power BI where someone wants to look for themselves. For forms and invoices, classic reporting tools. A data warehouse from the third source that names the same figure differently.

      Explainer · about 30 secondsThe same figure, three numbers — and what follows from it

        Three companies measure the same thing: the share of empty kilometres.

        Starts on its own · pause any time
        Best practice · bench learning

        Where the best solution sits today

        What is named is the pattern, not the vendor — a product name ages, an idea does not.

        One term, one definition

        Every metric with a name, a definition and an owner. It prevents the meeting where two departments present the same figure differently.

        Retire reports

        Without pruning you accumulate screens nobody opens and everyone maintains. Pruning is part of operations.

        Provenance at the value

        Measured or calculated, from which system — that belongs with the number, not in a footnote. Only then does it hold up in a dispute.