-
Step 01
Build your source hierarchy
Auditors, courts and filings rank above academic work, which ranks above consultancy reports, which rank above trade bodies, which rank above vendor content. Not because lower tiers are wrong, but because their incentives are known and their data is not independently checkable.
- The Landscape — every source with what it is for and what it won't tell you
- Pillar 07 · Reading the trade body
- About & method — how this site sources and corrects
DoWrite your own hierarchy, and note where you disagree with ours. Disagreement here is the point.You should now be able to answerFor each of the five source tiers, what is the incentive that bends it, and what would make you move a specific source up or down? -
Step 02
Practise provenance discipline
'$20.5 billion' and '$20.5 billion (MBI/FMI, 2026, US permanent modular, press release; full methodology member-gated)' are different facts. The second can be checked, compared and superseded. Make it a reflex.
- Pillar 01 · Market — 'Why the numbers disagree' and 'Our rule'
- MBI Industry Analysis — the release
- McKinsey 2019 — find the caveat everyone omits
DoTake any recent modular article. For every number in it, write source, date, definition and whether it is primary. Count how many survive.You should now be able to answerHow many figures in the last industry report you read could you trace to a primary source with a definition attached? -
Step 03
Learn the industry's unsourced lore, and design a test for each
Three widely repeated claims have no primary source we can find: the 300–500 mile transport radius, the double-wall acoustic advantage as a number, and per-factory throughput. Each is testable. Nobody has tested them. That is an opening.
- Pillar 04 · Inside the factory — 'Two claims that need better data'
- Pillar 02 — 'Logistics: the 300% surprise' (the radius claim)
- Articles index — 'Coming next' (open calls)
DoPick one. Write the study design: what data, from whom, at what cost, and what result would falsify the claim.You should now be able to answerWhat is the cheapest experiment that would replace one piece of industry lore with a sourced number? -
Step 04
Read announcements for what they omit
A project profile without a cost, a schedule variance against the original plan, or a comparison to site-built in the same submarket is telling you something by omission. 'World's tallest' announcements are unproven until occupied; AC NoMad was 'the world's tallest modular hotel' for six years and is now a condo.
- Pillar 05 · Case studies — 'AC Hotel NoMad' and 'Announcements are not buildings'
- MBI · Alora at Dry Creek profile (Aug 2026) — practise on this one
- The Real Deal — AC Hotel NoMad cancelled, replaced by condo (Sep 2024)
DoFor the Alora profile, list what a complete case study would contain that the article does not. Then find out whether any of it is publicly available.You should now be able to answerWhat five data points would turn a project profile into a case study, and how often does the trade press include any of them? -
Step 05
Follow the filings, not the feed
WARN notices, assignment-for-creditors filings, 10-Qs, bankruptcy dockets, legislative records and auditor reports are where the unconflicted numbers live. The Harbinger story broke in a WARN filing read by a blogger, months before any trade outlet.
- Offsite Straight Talk — Harbinger WARN filing (Jun 2026)
- Boxabl 1H 2026 10-Q summary
- LegiScan — Pennsylvania SB 908 (2025)
- Pacific Legal Foundation — MBI v. Stephenson
DoSet up alerts: state WARN databases in California, Idaho, Pennsylvania and Colorado; SEC filings for BXBL, SKY, CVCO, WSC, MGRC; state legislature trackers for 'off-site' and 'modular' in prevailing-wage bills.You should now be able to answerWhich filing would tell you first that a major factory was in trouble, and are you watching it? -
Step 06
Maintain a tracker with a verification discipline
Our Reciprocity Tracker and Factory Ledger publish their gaps as verified / partial / unverified tags. That is a method, not a weakness: a tracker that hides what it has not confirmed is a rumour table. Build your own, or fork ours, with the same discipline.
- Reciprocity Tracker — read the 'How to read this table' box and the compact note
- Factory Ledger — 'Scope and rules'
- Interstate Compact on Industrialized/Modular Buildings — model legislation (PDF) — an exercise: find the current member states
DoTake three unverified rows from either tracker. Confirm or correct them against the agency's or company's own page. Send us the result; we will credit you and date the change.You should now be able to answerWhat is your rule for when an unverified row is removed rather than left standing, and how would you defend it to someone whose company is in the row? -
Step 07
Watch the three questions that will decide the next five years
Will public buyers commit to pipelines factories can invest against? Will factory work get a wage classification of its own or be regulated as a job site? Will anyone publish the factory wage gap, plant utilisation and whole-life carbon before regulators and lenders force it? Everything else is downstream.
- Pillar 08 · Prevailing wage — 'The missing number' and 'Why public buyers decide this'
- Pillar 09 · Consolidation year — 'What to do with this'
- Terner Center — Potential Pathways (Mar 2026): the seven themes as a research agenda
You should now be able to answerWhich of the three questions could you contribute a sourced answer to in the next twelve months, and what would it take?
Check yourself
If you can answer these without looking, you have finished the trail. If not, the step number tells you where to go back.
- Rank five named sources by incentive and access, and defend one placement someone would dispute.
- Restate three industry figures with full provenance from memory.
- Describe a study design that would test one piece of unsourced lore.
- List the five omissions that most often separate a project profile from a case study.
- Name the filing that would first reveal a factory in distress.
- State your rule for removing unverified tracker rows.
- Pick the one open question you will work on and the first source you will go to.