Stonefield Advisory  presents

Behind the Curtain
Director's Cut

We're transparent. We say everything in plain English. Nothing's off the table. Nothing's hidden — except the dial settings, and we'll tell you that too.
For the geeks. For the nerds. For the review board asking “where did that number actually come from.”
Scene select

Three numbers, traced end to end

Only three here — not because the logic stops, but because nobody's asked about a fourth yet. The reasoning lives in our playbooks; the logic framework does the thinking, AI just draws the picture.

Scene 1 — Location Intelligence

Why does a city's labour-cost score come out the way it does?

Eurostat hourly cost × OECD annual hours = annual employer cost per head. The regional layer stays separate. The user's own weights do the ranking. Every cell reproduces on a calculator.

NOT blended
Eurostat hourly labour cost
(official)
Multiply: hourly × annual hours
OECD annual hours actually worked
(official, country-specific)
Annual employer cost per head
NL: €47.93 × 1,452h = €69,591
NUTS2 regional data
(official)
Regional €/hr — separate labelled layer
Normalise, apply the user's weight
e.g. Labour cost 7% of 100%
Other four priority sliders
(always sum to 100%)
Sum weighted contributions
City ranking score, e.g. 62 ±4
On the tool face: hourly figure is a
full-time / part-time blend — disclosed
Scene 2 — Company Intelligence

Why does a company get a footprint-pressure flag?

Public information only, in. A ±4 band on every score. Overlapping bands are never ranked against each other. A flag is a reason to look, not a conclusion.

no
yes
yes
no
yes
no
Eurofound European Restructuring Monitor
real, dated, sourced events
Public information only?
Excluded before it enters the scoring log
Pressure score from signal
frequency and recency
Attach ±4 confidence band
Bands overlap with a comparator?
Shown as tied — the tool will NOT rank them
Crosses the flag threshold?
FOOTPRINT-PRESSURE FLAG
Watch / No flag
A flag is a reason to look,
not a conclusion
Scene 3 — Executive Intelligence

Why does the rollup say “cannot tell you” instead of guessing?

Each executive question checks for its own required inputs. Missing input → the gap is named, never estimated around. Three honest labels: answered, partial, cannot tell you.

all
some
none
never
Eight standing executive questions
Identify the inputs each question requires
Check SCI / SLI / SPI / SHI:
does each input exist and connect?
Required inputs present?
ANSWERED
PARTIAL
CANNOT TELL YOU
Name the missing input
Estimate a value to fill the gap
The gap itself is the answer —
and SEI names who can close it
The part a web page can't do

These are stills. The live version generates on demand.

Inside our own tooling, any figure in any of the six tools becomes a trace like these in about a minute — generated from the recipe the moment you ask, not pre-drawn. That can't run on a static page, and we'd rather show you than claim it. Come and meet us: pick a number, any number, and watch it draw itself.

We'll gladly show you live when we meet →
Why this exists

Generate on demand. Not just in case.

We didn't pre-build a trace for every number across six tools, whether or not anyone would ever ask. That's the speculative AI spend we built Stonefield to argue against. The logic behind each figure lives as a structured recipe — source, transformation, rule, output — and the picture is drawn only when the question is real.

What's on this page is the exception that proves it: three traces, drawn once, so you can see what a trace looks like before you ask for your own.

In plain English: AI doesn't invent this logic — it executes it. The framework is ours, built over 250+ engagements. The stills are here; the live draw is yours for the asking.