Consulting & Management

A study was delivered six months ago. It is sound, and nothing has moved since. What is missing is not another analysis — it is a dated decision, with a name against it.

A recommendation only matters if it can become a decision — and then a delivered project.

Structured decision

Why a sound recommendation may never be adopted

A clean slab floats above a field of habits already in place: the organisation as it is, full. The recommendation is placed on top — the field sags, nothing was removed, and the slab loses its colour. Rewind: a few habits hollow out and fade, a space opens, and that is the move nobody wants to make. The slab settles into it, takes on the same material as the rest, and the system moves again — then three markers light up: who decides, who does, who checks. A mechanism of adoption, not a case: no real organisation, no names.

Consulting & management

The following Monday

A clean slab floats above a field of habits already in place — the organisation as it is, full. The recommendation is placed on top: the field sags, nothing was removed, and the slab loses its colour. Rewind: a few habits hollow out and fade, a space opens. The slab settles into it, takes on the same material as the rest, and the field starts moving again — then three markers light up: who decides, who does, who checks.

A management team that has already paid for a sound recommendation, and watched it become one more layer — because nobody ever said what would stop.

Watch the slab go out when it is merely placed, then come back to life once a space opens.

Three.js · WebGL · 63 gestes instanciés · affaissement à l'ajout · place ouverte par retrait · onde de reprise · récit en 5 chapitres · repli sans WebGL

THE MECHANICS

What orbits a decision before it is taken

A file's documents, risks and deliverables, each on its own orbit. The decision is taken at the centre, once the three arms line up — not before, and rarely on a hunch.

The documentsSpecifications · Administrative clauses · Bill of quantities · Drawings · Reports · Schedule
The risksDeadline · Firm price · Interfaces · Security · Currency · Payment
The deliverablesTake-off · Execution schedule · Technical memorandum · Site layout · Clarifications · Decision

Hover a point: each arm carries a family, and each has its colour.

Answering a tender is consulting — the moves are the same

Consulting is the firm's original trade, and answering a tender is its most demanding case: the deliverable is dated, binding, and marked by a third party who does not know you. The four moves below are those of a consulting engagement; they are found one by one, unchanged, inside a bid file.

Frame — decide whether to go at all

The first question in a file is not how to answer, it is whether to answer at all: what decision is expected, who takes it, by when, and the two or three facts that will tip it. A reasoned decline is a deliverable, exactly as an answer is.

It is the first piece of consulting in the file, and the only one that, taken the wrong way, costs everything else: no amount of writing recovers a bid that should never have been started.

Diagnose — read the documents, not the summary

An inventory of every document, what the buyer is actually asking for, the contradictions between documents, and what the file leaves unsaid. It is an organisational diagnosis, applied to a document instead of a company.

A contradiction between two documents is not a matter of form: it is a delivery risk, and one has the right to have it cleared in writing before committing.

Organise — steer the response like a project

Who produces what, in what order, with which internal deadline. Critical path, milestones, one owner per document, and a cross-review scheduled before submission — never after.

The deadline is not the milestone: the milestone is the day before. A submission is always decided in the final hour, and that is the hour when portals fail.

Decide, and leave the trace of the decision

Every trade-off — a price, a variant, a joint bid, a withdrawal — is dated, reasoned and attributed to a person. The technical statement is not a writing exercise: it is the written proof that the buyer's own project has been understood.

A statement that describes the company rather than the buyer's project is a catalogue. It is read once, marked low, and nobody can quite say why.

Deciding, then proving the decision rests on evidence

The first screen decides before the team is committed; the second ties every requirement to its evidence, its owner and its status. Both also state what they refuse to promise.

ADVISORY · DECIDE

Decide before committing the team.

An explicit recommendation, its caveats and the evidence behind it.

The cost of a poor bid starts long before submission.

GO · CONDITIONAL GO · NO-GO · INSUFFICIENT EVIDENCE

Method

Weighted multi-criteria matrix, assumptions separated from facts, with an explicit escalation rule.

Capability / requirement fit
Contractual and insurance exposure
Dependencies, exclusions and caveats
Available evidence burden
Assessment confidence: established · partial · N.A.

Our recommendation informs the decision. Legal review and approval authority stay with you: that is what makes the advice usable.

ADVISORY · ASSURE

Every requirement has evidence, an owner and a status.

Compliance becomes a verifiable chain, not a decorative checklist.

A caveat found on the eve of submission becomes a certain risk.

COMPLETE · REMEDIATE · BLOCKING · N.A.

Method

Requirement → evidence → version → reviewer matrix, with deviation log and submitted-version freeze.

Binding requirement
Evidence and exact location
Author / reviewer
Gap, remediation, closure date
Submitted-version fingerprint

An “approved” status carries its reviewer, version and review scope. That is what makes it defensible.

A consulting engagement, week by week

The durations and volumes below are orders of magnitude observed on a typical engagement, not a measurement of the firm: a wider scope stretches them, it does not change their order. What matters is the sequence — nothing is modelled before it is framed, nothing is settled before the file is worked up.

Weeks one and two — frame

On a typical engagement, eight to fifteen short interviews, one per function involved, and a framing note held to a single page: the decision expected, who takes it, by when, and what is explicitly out of scope.

A framing note that does not fit on one page has not framed anything: it has listed. It is the space constraint that forces the scope to be settled.

Weeks two to five — work up the file

Internal documents are taken back to their source, never to the deck that quotes them. Whatever could not be obtained appears in the deliverable, named, with its reason — a declared gap beats a filled one.

A figure lifted from a deck inherits its framing without inheriting its method: it is right in its original context and wrong in the new one, and nothing says so.

Weeks five to seven — settle

Three costed scenarios on identical assumptions, plus the do-nothing case: four columns, never twelve. Criteria weights are settled before a single score is known.

Weights set after the scores no longer weigh anything: everyone adjusts the weight of the criterion that serves their option, and the table becomes a mirror.

Then — carry

A steering meeting every two weeks, an agenda carrying decisions to take rather than topics to mention, and a register of closed decisions: the reason, the date, the person who decided.

A decision with no name against it is not closed: it will be reopened at the next meeting, by someone who was not in the room.

Four situations we know how to take on

Situations from the trade, not client references: each states the method used, the deliverable handed over, and the difficulty that had to be cleared. Assignments actually carried out, names removed, are just below.

A sound study, delivered six months ago, and nothing has moved

We take the study back to extract the decision it calls for, rewrite it in one sentence with a decision-maker and a date, then work up the two or three facts that were missing to settle it. Deliverable: a one-page decision note, and the register where it is dated.

What was hard: the study was not wrong, it simply called for no decision. It had to be accepted that one more analysis would have unblocked nothing.

In-house or outsourced, with no common basis to decide

Both scenarios are worked up on identical assumptions — fully loaded costs, risks, transition time — and the do-nothing case is costed like the rest. Deliverable: a decision table where every criterion carries its weight and its anchor.

What was hard: settling the weights before any score was known. It is the only sequence that stops the intended answer from being manufactured.

A drifting project nobody can locate

We rebuild the real position from evidence: what is delivered, what is accepted, what is paid — three distinct things that progress reports blur together. Then the schedule is rewritten from the end, and what can still be held is separated from what cannot.

What was hard: stating the gap between declared and accepted progress without putting a person on trial. A recovery that hunts for someone to blame stops receiving information altogether.

Governance where everyone approves and nobody decides

For every recurring decision, we name the authority that settles it, the one consulted, the one merely informed. Bodies that take no decision are withdrawn, and the information they carried is provided another way. Deliverable: a governance model and the list of meetings removed.

What was hard: removing a meeting is harder than creating one. Its members must be given back what it brought them — information — without being given back the power to block.

Files framed on evidence, told without the names

Framing a file, writing its interfaces before they become disputes, dating every decision: the firm's assignments seen from the method side. Told as they happened, names removed, volumes kept.

TRACK RECORD · REAL ASSIGNMENTS, NAMES WITHHELD

01
A finishing-trades contractor, around ten staff, northern France. No in-house estimating department.
When · Three consecutive years, between 2017 and 2021.
What we did
We ran its entire bid desk, exclusively: spotting the consultations, inventorying the documents, quantity take-off, unit-price build-up, technical submission, form of tender, filing. It had no estimating department — we were it.
What it produced
111 priced bills of quantities, 68 specifications and tender regulations analysed, 54 signed forms of tender, 48 execution programmes, 47 technical submissions, 27 drawing sets. Up to a hundred consultations handled in a single month.
What was hard
The pace. At a hundred files in a month, what breaks is never the pricing: it is the reading. We held it by industrialising the document inventory and the hunt for contradictions between documents — so that engineering time went to the line items carrying most of the money, and nowhere else.
02
A local healthcare facility, new build, all-trades tender file.
When · File from the 2017-2021 period, analysed in 2026.
What we did
We went through the entire tender file — fourteen packages, from external works and services to signage, taking in structure, waterproofing, joinery, cladding, heating and ventilation, electrical works and lifts — to draw out a usable schedule of works, then costed it blind, line by line.
What it produced
1,063 all-trades line items named and classified by package and family of works. 75 % of them covered by a median price drawn from our own bills. The remaining 25 % are listed as gaps, package by package — that is the part of the document we show first.
What was hard
The file's bills were blank: no prices, no quantities, forty-four priced lines out of one thousand and sixty-three. The take-off was left to the bidders — meaning the client handed every contractor the act that determines the price, with nobody checking that they were all measuring the same thing.
03
Five competing contractors, one and the same package, one and the same tender file. Measured from our own archive.
When · Bids from the 2017-2021 period, measured in 2026.
What we did
We took the five bids submitted on the same works and traced back, for each one, not the price shown but the quantity it had measured before setting it.
What it produced
From 281 to 505 square metres measured for the same works, and from 9.73 to 25 euros per square metre: a one-to-four-point-six spread on a single line of the bill. The full comparison, quantity by quantity.
What was hard
Proving the gap came from the take-off, not the price. A bill of quantities does not say so: it shows an amount that looks firm. We had to rebuild each bidder's quantity to show that five serious contractors had simply not been pricing the same building.
04
The technical submissions actually filed by a finishing-trades contractor, across three years of consultations.
When · Documents from 2017-2021, re-read in 2026.
What we did
We opened and read them one by one, to measure what they actually contained — not what a template claims they ought to contain.
What it produced
31 submissions opened and read, none illegible. A drafting frame in thirteen sections, ranked by actual frequency: staffing, plant, programme, safety, comparable project references and product data sheets appear in twenty-nine out of thirty-one. Twelve drafting rules come out of it, each drawn from a document we opened.
What was hard
The first count announced forty-seven. Only thirty-one really were: the rest were duplicates and files whose name carried the word without the content. We wrote the discrepancy down rather than keep quiet about it — an announced figure you correct beats an announced figure you defend.
05
French contractors and public buyers, on consultations straddling the switchover.
When · Watch maintained through 2026.
What we did
We went back to the text for two things: what changes, and above all the rule that decides which of the two regimes applies to a given file.
What it produced
Since 22 August 2026, every French public contract carries at least one award criterion taking the environmental characteristics of the bid into account (French public procurement code, article L. 2152-7, as enacted by law no. 2021-1104 of 22 August 2021, article 35). For scale: according to the 2024 census of the French public procurement observatory, published in 2025, an environmental criterion appeared in 25 % of contracts and a social criterion in 12 %, across 223,383 contracts worth 233.3 billion euros — of which small and medium-sized businesses took 60 % of the count and 25 % of the value.
What was hard
The regime attaches to the date the consultation was launched, not to the award: a file started before the switchover and awarded after it stays under the old rules. And the two tiers of the text do not switch on the same day — one day apart between regulation and statute. A note that conflates them builds an environmental section for nothing, or omits it where it is being scored.
06
All-trades projects, private and public clients alike.
When · Method tested on files from 2017 to 2026.
What we did
We write every seam between two packages as a dated deliverable — who does what, hands over what to whom, under whose check — into the upstream package's specification, and we make it a chapter of the technical submission.
What it produced
A grid of ten interfaces, each with its control point and its reference defect: finished level and upstands between structure and waterproofing, builder's work openings signed off before the pour, fire-stopping attributed to a named party, screed moisture measured before the floor finishes, base tolerances written into the mason's package at the value the carpenter requires.
What was hard
Two contractors can each comply with their own standard and the works still be rejected between them. Five millimetres of flatness defect at the foot become a centimetre out of plumb at the head. And on the industry's most frequent dispute — laying a floor finish on a screed that is still damp — two standards disagree: one reasons in elapsed time, the other in measured moisture. A waiting period met is not a dry substrate.
07
Organisations where several teams work the same subject, and our own assignments.
When · Practice installed and measured since 2026.
What we did
We keep three separate registers on every assignment: what is done and its evidence, what remains, and what is settled and will not be reopened. Each is updated the moment the thing happens — never at the end, because by then the people have gone and only the document remains.
What it produced
A register of closed decisions carrying, for each one, its reason, its date and the person who took it; an automated check that verifies every evening that no status document still says "to do" about something already done; and the measurement that triggered the method — the word "consolidation" appeared sixty-two times in our own work log before this list existed.
What was hard
A written rule protects nothing: it waits to be read, and nobody reads two hundred rules before acting. What holds is the check that runs by itself. That is what we install with a client — not one more method note.

No client name is published, with or without their consent. We publish no win rate: it is not measured, and an unmeasured figure is not a reference.

What we do

Tender responses run as consulting engagements: a reasoned go/no-go first, a reading of the documents next, then production steered like a project. This is our original trade, and its most demanding case — the deliverable is dated, binding, and marked by a third party who does not know you.
Diagnosis: what is actually blocking, not what gets said in the room.
Costed, compared scenarios, with their dependencies and their risks.
Operating model and governance: who decides, who delivers, who controls.
Change management: what changes in each person's work the following Monday, who announces it, and what stops being done — an arrangement that removes nothing is not adopted, it is merely added.
Owner's advisory on sensitive operations: re-reading the documents on the client's behalf, pricing variations before signing them off, keeping the claims register and its notification deadlines.
Costing the "change nothing" scenario — waiting is an option, and it is almost always the one that wins by default, for want of ever being costed.
Taking back control of a project that has drifted: where it actually stands, what can still be held, what cannot.
Support through delivery to the deliverable, not just to the final presentation.

What we deliver

Framing note & scope of engagement

Reasoned go/no-go note — the written reason for a decline counts as much as an answer

Written owner's-advisory mandate, settled before the first work order

Compared-scenarios file, on identical assumptions

Roadmap & dated milestones

Governance model & steering bodies

Risk management plan

Steering dashboard, with the previous reading kept

The method

Frame

Write in one sentence the decision expected, who takes it, by when, and the two or three facts that will tip it. Most files fail right here.

Instruct

The file is built on evidence: data, constraints, contradictions — and what is not known gets said, instead of filled in.

Arbitrate

Costed scenarios, one recommendation, and a decision that stays yours — plus the way back, while it still exists.

Carry

Monitoring, steering bodies, accountability through to the final deliverable. An agenda carries decisions to take, not topics to mention.

WHAT YOU WILL HAVE IN HAND · The Robust Decision Table

Change one assumption: the option that holds, the one that breaks and the reversible decision are told apart.

Each cell holds or does not — none is grey.

We do not deliver a recommendation detached from its conditions. We organise options, facts, unknowns, effects and the cost of waiting so leadership can decide and know what would invalidate that decision.

The table first shows the choice that survives the broadest set of assumptions, the point that has to be settled and the next action. It also preserves the way back: an irreversible decision is never presented as a simple button.

Test the assumptions

Options, evidence, sensitivity and reversibility

Case for change, theory of change and BAU: the problem, the expected mechanism of change and the do-nothing counterfactual.

Objectives and critical success factors: the outcomes sought and the knock-out criteria.

Longlist to shortlist: options opened up before a documented narrowing.

Five Case Model: strategic, economic, commercial, financial and management cases.

MCDA: anchored scoring scale, independent criteria and no double counting.

Swing weighting: weight reflects the gap between the worst and best level, not the criterion's abstract importance.

Veto and gate: an unviable option cannot re-enter the ranking on the strength of a good average score.

Sensitivity analysis and switching value: the change in assumption or weight that reverses the ranking.

Optimism bias, risk cost and contingency: historical bias, specific risk and residual provision are kept distinct.

Decision rights and RACI: sponsor, decision authority, accountable owner, contributors and control.

Benefits map: outcome, benefit, indicator, baseline, owner and verification date.

Stage gates, RAID log and decision log: risks, assumptions, issues, dependencies and versioned decisions.

The table informs the trade-off; it replaces neither the decision authority that validates criteria, anchors and weights, nor the commitment of those who own the decision.

THE QUESTIONS THAT SET THE PRICE

What we will ask you before we start.

Four questions we ask on every file in this domain. Each comes from an incident someone paid for — us, or a client before us.

Which decision must come out of the assignment, and who takes it?

Before any diagnosis, we write in one sentence the decision expected, the name of the person who takes it, the date by which it must fall, and the two or three facts that will tip it. A deliverable that lights no decision is one more report — and one more report gets judged by its weight.

If nobody asks it — Three months of workshops, an eighty-page report, and the same question still in front of the same committee.

What is already settled, and not to be reopened?

We keep a register of what is closed: the decision, its date, its reason, and the person who took it. Reopening a settled matter costs more than settling it, and every new arrival on the team reopens it if it is written down nowhere. The register is updated the moment the decision falls, never at the end of the assignment.

If nobody asks it — The steering committee replays in March the debate it had in January, with different people and the same outcome.

How will we know the assignment succeeded — by which measure, taken when?

Every objective carries its measure, its starting value, its target and the date of the reading — and the previous reading is kept. A single value cannot be wrong; two can: it is the gap between them that raises the alarm, and a dashboard that overwrites yesterday's reading is blind to any drift.

If nobody asks it — A transformation called successful that nobody can quantify in front of the board.

Who holds the client-side assistance when the designer and the contractor disagree?

The client's assistant is neither the designer nor the builder: they re-read the documents on the client's behalf, price the variations before signing them off, keep the claims register with its notification deadlines, and prepare the decision rather than comment on it. We set that mandate down in writing before the first work order.

If nobody asks it — The client arbitrates alone a technical dispute it has no means of reading, between two parties who can read it perfectly well.

CAPABILITY SCENARIO — NOT A CLIENT REFERENCE

When we get the call

A leadership team is torn between bringing an activity in-house and outsourcing it, with no common basis to decide. We build both scenarios on identical assumptions — fully loaded costs, risks, transition time — and document what separates them. The decision is made in one session, on evidence.

Related disciplines

A file rarely draws on a single discipline.

Engineering & Construction

Executable project

Legal & Compliance

Secured framework

All disciplines in this practice

Planning something in Consulting & Management?

Write to us. We respond with an assessment, not a brochure.

Write to us