Shifting Ground is a pilot that maps how construction law has moved on nine questions every construction contract has to answer, from who pays for defects to how far adjudication is enforced. It holds 78 authorities (statutes, judgments, reports and standard forms, three of them Scottish for comparison), running from 1838 to 2026 4]. It is an Effectus Construction Services build, and it is published on effectuscs.com at [www.effectuscs.com/insights/shifting-ground.
This piece is not about the law. It is about the build: how a stack of legal documents became a structured, queryable evidence base and an interactive tool, and what that process shows about systemising knowledge that usually sits in textbooks, notes and people's heads.
From documents to records
The first decision was about shape. Each theme became a question with a scale. For defects the question is "Who can recover the cost of putting a defective building right, and against whom?", and the scale runs from "liability contracts" to "liability expands". Every authority then became a record rather than a paragraph. In the dataset, each entry carries its theme, date, kind, citation, court, a one-line holding, its links to other authorities, a score with the reason for that score, and a verification status.
Behind each record sits a case sheet in a fixed form: holding, facts, issue, decision, ratio and obiter. Sections for other reasons for the decision, dissent and later treatment were added in later versions. The fixed form is what makes 78 very different documents comparable. It also has a less obvious benefit. A structured record shows its gaps in a way a paragraph does not: if a field can't be filled from a source, the empty field is visible.
The sourcing rule follows what can actually be checked. Judgments from 2001 onwards were read on The National Archives' Find Case Law service on 26 September 2026, and their summaries were written from the judgment text with paragraph references [1]. Statutes were checked on legislation.gov.uk [2]. Earlier cases, and some Scottish and House of Lords decisions, are not on Find Case Law. For these the summary is checked against later judgments that quote them, ECS's own copies of the reports, or named secondary sources. Each case sheet states the sources checked for that entry, and entries that rest on the reported decision alone are tagged "verification incomplete".
Once the records existed, the same data could be drawn in four ways:
- Pendulum shows each authority's push, by date.
- Swimlanes separate Parliament, each level of court and policy, with curves showing how later authorities treated earlier ones.
- Narrative reads the theme as a sequence.
- All themes compares when each doctrine moved.
All four views read from one dataset. None holds facts of its own, so a correction to a record is a correction everywhere that record appears.
A structured record shows its gaps in a way a paragraph does not.
Checking it, round by round
The first version covered seven themes. It then went through six rounds of review, and every round found something. The findings are kept in a single fact-check log, which exists because one review pointed out that the register listed only selected corrections. That log is now as much a part of the tool as the charts.
In the first round, seven independent reviewers, one per theme, checked every entry against primary sources. They returned 67 findings: 7 errors, 18 overstatements, 36 minor points and 6 items that could not be verified. All errors and overstatements were corrected. One finding was about the page rather than the law: its header said 64 authorities when 60 were distinct, because some authorities appear in more than one theme.
The second round researched the open questions. Three research agents checked the remaining items against my own library first, including scanned law reports read by OCR and copies of the Latham, Egan, Construction Playbook and NEC documents, and then against later judgments and secondary sources. The third round fact-checked two new themes, limitation and latent defects, and collateral warranties and third-party rights. It produced 12 findings, all fixed.
The fourth round was an adversarial stress test by Codex, run without access to the primary sources, so its findings were about internal contradictions and calibration. One finding, on statutory subsection references, was rejected after a check of the legislation itself. Another, that some "ratio" fields held material that was strictly obiter, is why the case sheets gained their sections for other reasons, dissent and later treatment.
The fifth round re-set all 88 scores (some authorities are scored in more than one theme) against a written rubric, using an independent reviewer. It proposed 25 changes, none more than one step and none changing sign. 24 were applied and one was held by editorial override.
The sixth round reviewed the finished page. It found that two views were treating each score as a position when it is a push, so the charts now draw each authority as its own stem from zero and the Narrative labels each by the direction of its own push. It found that averages depended on the order of entries, so an authority in several themes now counts once. It replaced a claim on the page that the courts do most of the moving with a plain description of the collection: 59 of the 75 English-law authorities are judgments. And it found that one judgment in the collection had been given permission to appeal to the Supreme Court [3], a check earlier rounds had not made.
What the scores are, and are not
Each score is direction multiplied by significance, from -2 to +2 on the theme's scale. First-instance decisions, reports, standard forms, obiter points and Scottish decisions in English themes are capped at 1, and every case sheet gives the reason for its score. The rubric makes the scoring consistent and open to challenge. It does not make it objective. The scores remain ECS editorial judgement, and the tool is research commentary, not legal advice. The collection is also a curated set of leading authorities, not a census, so its totals describe the collection and nothing wider.
What the build shows
Most of what a profession knows sits in forms that are hard to check: textbooks, lecture notes, precedent files and the memory of the people who use them. Shifting Ground turns one slice of that into records with fixed fields, named sources and a stated verification status. That is what makes it possible to chart, query, review and correct.
The review history is the strongest argument for doing it this way. Errors were found one field at a time, by reviewers working on one theme at a time, and each fix landed in a single record. A narrative text would have hidden many of those errors inside fluent prose. The structure does not remove judgement, as the scores show. It makes the judgement visible and arguable, which is where professional disagreement belongs.
The same method is not specific to law. Taking the documents an organisation already holds and turning them into records it can query and check is the core of Synaform's Data service.
Sources
- [1] The National Archives, Find Case Law. https://caselaw.nationalarchives.gov.uk.
- [2] The National Archives, legislation.gov.uk. https://www.legislation.gov.uk.
- [3] The Supreme Court, case record UKSC/2025/0156 (Triathlon Homes LLP v Stratford Village Development Partnership), https://www.supremecourt.uk/cases/uksc-2025-0156. Case record re-checked 19 October 2026: status "Hearing listed", hearing 19 November 2026.
- [4] Effectus Construction Services, Shifting Ground: how English construction law has moved, 1838 to 2026. https://www.effectuscs.com/insights/shifting-ground (page header "9 Themes, 78 Authorities, 4 Views", checked 19 October 2026; last updated 26 September 2026).