How the engine works
A glass-box rules engine.
DCA RouteLab is a deterministic commercial reasoning engine. There is no model, no training data and no hidden scoring. Every flag, label and recommendation comes from a named rule that an officer can read, cite in a board paper, and challenge.
The transparency promise
If an officer asks “why did this flag fire?”, the answer is always one sentence citing the specific inputs and the named rule they matched. The same project inputs will always produce the same flags, labels and counterfactuals. Every output is traceable to a stable rule ID and, where applicable, a section of the supporting white paper.
Anatomy of one rule
Sample rule
surveysIncompletepriceability tree · red · amber“Surveys incomplete but procurement assumes scope certainty”
Why it matters
Unresolved survey information tends to convert into variations, claims or provisional sum disputes once the contract is let.
Paper guidance
“If design maturity is below threshold and programme urgency is high, test two-stage, ECI and framework options.”
Reshape choices
- —Complete or progress critical surveys before tender
- —Build provisional sums with governance protocol
Cited: §3 narrative on survey / design maturity · Rule set v1
See it fire on a project →What the engine takes in
Commercial and project parameters only — no resident or service-user data is required or accepted.
Project profile
Asset type, value band, procurement stage, contract form (JCT, NEC, framework, bespoke) and intended pricing approach.
Sixteen conditions
Officer ratings on a Low / Medium / High scale covering design maturity, inflation exposure, market capacity, client capability and twelve others.
Procurement routes
One or more of nine routes to compare — open tender, two-stage / ECI, framework call-off, dynamic market, alliancing, direct award and others.
Live market context
Optional snapshot of ONS CPI, Bank Rate, materials inflation and build-cost OPI — used to pre-fill condition ratings, never to run a model.
The evaluation pipeline
Five pure functions. No randomness. No external calls at evaluation time.
- 01
evaluate(project)The single entry point. Loops over every named rule, checks its condition, and collects the ones that fire. Then attaches a behaviour label to each selected route and finds the smallest counterfactual change that would improve each route.
- 02
priorityReds(analysis, n)Returns the top-n red flags by priority weight — the gate-review summary an officer sees first. The most serious flags, not just the most numerous.
- 03
stressTest(project)Nudges three key conditions one step in the adverse direction (inflation higher, design maturity worse, market capacity thinner) and reports which routes degrade. A stylised volatility overlay — not a Monte Carlo forecast.
- 04
evaluateCounterfactual(project, change)Clones the project, applies one hypothetical condition change, and re-runs the full evaluation. Powers the “what if design maturity improved?” prompts on each route card.
- 05
ruleFiringForCell(project, route, condition)For any route × condition cell on the heat-map, returns an intensity (high / medium / low) and a plain-prose explanation of why that route is more or less exposed to that condition.
The building blocks
Fourteen self-contained, inspectable modules. Each is a named set of rules an officer can read.
- Routes
Static definitions for nine procurement routes — strengths, vulnerabilities, governance burden, affordability exposure.
- Rules
Around thirty named, numbered rules. Each has a predicate, a severity (red / amber / info), a white-paper citation, suggested clause families, reshaping choices and a “cleared by” hint.
- Exposure
Indicative base / adverse / severe outturn bands in £m using deterministic uplifts. Suitable for a capital board paper. Explicitly not a simulation.
- Delivery
Assesses live post-award events (variation, claim, contractor distress, programme slip, market shock) against the original commercial posture.
- Reshape
Recommends the best-fit pricing mechanism from current conditions, lists mechanisms to avoid, and flags any mismatch with the officer’s chosen approach.
- Scenarios
Base / adverse / severe market overlays, loosely calibrated to Bank of England ACS stress scenarios, applied on top of a live market snapshot.
- Panel
Assembles the printable Commercial Strategy Panel — a structured document summarising routes, flags, protections, exposure and reshape recommendations.
- Conditions
Sixteen commercial conditions with display labels and reverse-scoring logic (e.g. “Low” design maturity means high commercial risk).
- Protections
Contract-aware checklist of pre-launch protections — PCGs, performance bonds, retention bonds, step-in rights — derived from conditions, value band and contract form.
- Benchmark
Named point-in-time snapshots so officers can compare two versions of the same project and see which conditions worsened.
- Value fit
Flags where a route’s overhead or legal exposure is disproportionate to the project’s value band.
- Trees
Tags every flag and route assessment against one of seven commercial fit-test lenses — marketability, priceability, risk transfer, affordability, client capability, optionality & timing, downside protection.
- Market defaults
Translates live market data (CPI, Bank Rate, OPI) into condition pre-fill ratings using asset-specific thresholds.
- Lessons
Captures post-award lessons and links them back to the controlling commercial lens — an institutional learning loop tied to the same rule taxonomy.
Determinism guarantees
Why an officer can stake a board paper on the output.
No model, no randomness
Every rule is a named predicate. The same inputs always produce the same flags, labels and counterfactuals.
Versioned rule set
The constant RULE_SET_VERSION = “1” is stamped on every decision record, so any saved entry is traceable to the exact rules that produced it.
Browser-local custody
Conditions, ratings, decisions, snapshots and reasoning notes are stored in your browser. Nothing is sent to a server unless you explicitly save or export.
What the engine returns
Every output carries a stable ID, a plain-English explanation and a citation back to the rule that produced it.
Route assessments
A behaviour label — Robust, Usable with controls, Fragile, High governance burden, Legally sensitive, or Not enough information — with the prose reason that produced it and the strengths and vulnerabilities behind it.
Triggered flags
Each flag carries a unique ID, a white-paper citation, a severity colour, a five-step posture ladder (from “proceed” to “stop and escalate”), suggested clause families and reshaping choices tagged by leverage.
Counterfactual hints
For every route, the smallest single-condition improvement that would upgrade its behaviour label.
Stress results
A side-by-side baseline vs. stressed analysis with a per-route degradation delta — stylised, not Monte Carlo.
Evidence & provenance
Every market-derived figure carries its source, reference period and retrieval date. The engine never smooths, forecasts or interpolates these values — it reads the latest published print and stamps it.
ONS SPPI / Build-cost OPI
Used as the tender-price inflation assumption in the market-slip readout. Refreshed monthly against the ONS release calendar. The series label (All / Housing / Infra) is selected from the asset type.
Bank of England Bank Rate
Used only as a context input for the finance-exposure narrative. The effective date of the current rate is displayed alongside the figure.
ONS CPI (fallback)
Substituted for build-cost inflation only when the OPI series has not yet been refreshed. Clearly labelled where used.
Officer overrides
Any value an officer changes in the workspace is stamped “officer input” on the Decision Record and displaces the live-market figure for that project only.
How the market-slip figure is derived
For award horizons of 6, 12 and 24 months, the engine compounds the latest published tender-price YoY on the value-band midpoint (or the officer-entered exact value). The low/high bracket applies −1pp / +1.5pp to that rate. No forecast is generated. Replace with the project cost plan and its procurement programme as soon as they are firm.
Read it, then run it
The fastest way to understand the engine is to watch it fire.
This page describes the current rule set (v1) and will be updated as the engine matures. It is a commercial reasoning aid, not legal advice or financial approval; officer judgement and statutory duties remain with the authority.