MARKET INTELLIGENCE — DATA INTEGRITY
Bad market data can feed you wrong numbers without warning.
The dangerous failure is never the system that breaks loudly. It is a competitor redesigning their website, the collector silently returning nothing, and a quiet month reading as no movement in the market. This is what decides whether the data can be trusted at all.
Free · 20 minutes · no commitment
WHO THIS IS FOR
Anyone about to price against a data feed they did not build and cannot inspect.
THE PROBLEM
By the time anybody notices, the decisions taken on the bad data are already out in the world — quotes sent, capacity committed, prices held. Silent degradation is worse than an outage precisely because nothing prompts anyone to check.
WHAT IT LOOKS LIKE
Market intelligence · Integrity report
EXAMPLE DOCUMENTWeek of September 28
PALLET STORAGE RATES, GTA · SENT MONDAY, OCTOBER 5, 6:00 AM
Sources
Operator A · rate page
Read every 6 hours. 28 rates, all inside their usual range.
HealthyOperator B · rate page
Went from 31 rates to none on Wednesday after its site changed. Quarantined at 2:10 pm; its figures are marked degraded, not read as zero.
QuarantinedListings site
Read hourly. 170 listings across the GTA; 6 buildings listed twice, each counted once.
Healthy
SOURCES BREAK · THE POINT IS THAT YOU FIND OUT
Rejected before the model
One figure for every size and building
- Rejected
- 12
- What it caught
- A placeholder page, not real rates
Under $8 a pallet a month
- Rejected
- 3
- What it caught
- A weekly rate read as monthly
Over three times the median
- Rejected
- 1
- What it caught
- $195.00, a misread $19.50
| Rule | Rejected | What it caught |
|---|---|---|
| One figure for every size and building | 12 | A placeholder page, not real rates |
| Under $8 a pallet a month | 3 | A weekly rate read as monthly |
| Over three times the median | 1 | $195.00, a misread $19.50 |
Flagged, not averaged
Operator A’s Milton building is $18.50 a pallet a month on its rate page and $21.00 on the listings site. Neither enters the median until the two agree.
- Rate page
- $18.50Check
- Listings site
- $21.00Check
- Never used
- $19.75, the middle of the two
Does the signal predict your enquiries?
TESTED EVERY MONDAY AGAINST YOUR OWN STORAGE ENQUIRIES
- Ambient, GTA west, 38 weeks: it holds. The signal moves 2 to 3 weeks ahead of your enquiries. Recommendations on.
- Ambient, GTA east, 21 weeks: weak this quarter. Information only until it holds again.
- Frozen, 9 weeks: too few weeks to tell either way. Information only.
No recommendation is made in a market until the signal is shown to predict your demand there.
What you get.
- Validation rules that reject implausible data before it reaches the model.
- Automatic quarantine when a source’s yield collapses, so it stops instead of reporting zeros.
- Confidence scoring across sources, so conflicting data is flagged, not averaged into a plausible middle.
- A correlation harness measuring whether the signal predicts your demand, which says so plainly when it does not.
How it works.
Priced against the scope, after a Diagnostic that is small on purpose. The shape of the work we can tell you now.
- THE SHAPE
- Not sold on its own. It is the substrate under the other three, and the reason they can be trusted.
- WHAT WE NEED FROM YOU
- Your demand history, which is the thing the signal gets tested against.
- WHAT YOU KEEP
- The validation rules and the full correlation record, including the unflattering parts.
What this does not include.
Published because you should know before the call, not after the invoice.
- A promise that every source stays available. Sources break; the point is that you find out.
- Recommendations on a route or market where correlation has not yet been demonstrated.
QUESTIONS
What happens when a source breaks?
It is quarantined, it stops feeding the engine, you are told, and the affected figures are marked degraded, not reported as if nothing changed.
What if the signal turns out not to predict our demand?
Then the system says so, and recommendations stay informational on that route until it does. A system that can report its own premise failing is the only kind worth pricing against.
MORE UNDER MARKET INTELLIGENCE
WHERE THIS APPLIES
The industries whose page names Market Intelligence as one of the places we would start. Each one is written for that trade, with its own numbers.
Bring us one bad week.
Book a pilot call. Tell us where the week goes wrong, and we’ll tell you what an Operations Diagnostic would map and measure.
The first paid step has a clear scope and an end date, and the pilot after it is one workflow — not a platform, and not an open-ended retainer.
Free · 20 minutes · no commitment