The Discovery Study
Thirty days to find out where the margin actually goes.
Not a strategy deck. A ranked, costed list of the work that is quietly eating your operating margin — built from how your business actually runs, not from what anyone remembers in a meeting.
Why this exists
Nobody in the building can tell you where the time goes.
Not because they're hiding it. Because the work is spread across a dozen systems, a hundred habits, and a few people who have quietly absorbed the mess for years. Ask them and you get the version they can describe. The expensive part is the version nobody thinks to mention.
How the work runs
Five stages. You get three of them before you commit to a build.
The Discovery Study is the first three. The gate sits where it does on purpose — you have the maps, the register and the plan in hand before either of us decides whether there is a program here at all.
Stages one to three are the 30-day study, and the subject of this page. Four and five run across the 120 days after the gate, in cycles, only on what survived stage three.
01 · Observe
Instrumented where it can be. Observed where it can't.
Three lenses, because no single one sees the whole job. The capture isn't the product — having enough of the truth to rank what matters is.
Where your systems will tell us — ERP, job cost, estimating, purchasing, billing — we take the record of what actually happened. Volumes, cycle times, rework, how often a thing gets touched twice.
We sit with the people doing the work and watch the job go past. Most of the expensive friction lives in the gaps between systems, where no log is written and no ticket is ever raised.
Then we ask everyone, by team, where the work actually hurts. Not a satisfaction survey — a map of the friction people feel daily and have stopped reporting because nothing came of it last time.
Anything we deploy runs in accounts held in your name. Your data is not used to train anyone's model, and the vendor list is approved before anything touches it.
02 · Distill
Three lenses, one register.
Observation is cheap and endless. The work is turning it into something a business can actually act on — a map of how the job really runs, and a single ranked list rather than nine competing ones.
03 · Rationalize
Payoff × feasibility × adoption.
Three scores, multiplied — not averaged. A high-payoff item nobody will adopt scores near zero, which is the correct answer and the one most transformation plans refuse to give.
Buy before build: where a vendor already solves a piece, it gets evaluated before anything is written.
How value gets counted
Three categories, against a baseline taken before anything changed.
Hours returned
Time given back, multiplied by the loaded labour rate.
Hires avoided
Capacity created instead of bought.
Margin recovered
Rework removed and change orders captured that were leaking.
Every report separates the capacity that has been returned from the amount that has actually been realised in your financials. They are not the same number, and anyone who reports them as one is flattering themselves.
Day 30 · the gate
What's in your hands when you decide.
Six things. All of them yours, whether or not you carry on.
- Governing principles — agreed with you in week one, and used to decide what gets picked up and what gets left alone
- As-is process maps for your top job families, as the work is actually done
- The heat-map survey results, by team
- A scored opportunity register — every opportunity rated one to five on payoff, feasibility now and adoption likelihood, ranked by the product of the three
- A 120-day value capture plan with named owners, for your sign-off
- Your own workflow, already improved — the executive sponsor's operating system is worked on from day one, not after the study
Then either of us can stop. By email, on or before day 30, no penalty and no further fees. We would rather find out in thirty days that there isn't three to five points of EBITDA sitting in your back office than spend five months proving it slowly.
Worth a conversation?
Bring your call volume, your job count, and the part of the week you dread. Half an hour is usually enough to tell whether there's anything here.