↻ One cycle per process. The bottleneck moves, which is why it starts again at F.
The FLOWREFY method: eight steps, two halves
The brand name is the acronym. F, L, O and W measure — nothing is changed in that half, things are priced. R, E, F and Y refine — this is where work happens, and only on what the measurement named. This page explains the order. Each step below has a page of its own with the procedure, the criteria for calling it finished, and the mistake typically made at that point.
Why the order is not arbitrary
The common order in practice is the reverse one: a tool gets bought, then someone looks for a process to put in it. The result is an automated workflow that nobody can price — not before, and therefore not after either.
The four FLOW steps produce exactly that missing figure. F collects the candidates, L prices them, O finds the step that actually blocks once the process is under load, and W costs the intervention against its return. Only then is it clear whether anything should be done at all. At that point, “do nothing” is a permitted and occasionally the correct answer.
The four REFY steps are ordered too, by cost. R strips ballast and needs no licence. E clarifies ownership and costs nothing either. Only F touches technology — exactly one lever, because two changes cannot be told apart in Y. Y then re-measures with the same arithmetic L used, and says where the bottleneck has moved to.
After that it starts at F again. Not as a ritual: a bottleneck that has been removed always exposes the next one, because the second-slowest step is now the slowest. One cycle per process, then reassess.
The eight steps
Measure first
Four steps until a number is on the table. Nothing is changed here. Things are understood.
Find — collect the candidates
- Question
- Which processes are worth looking at at all?
- Outcome
- A list of five to fifteen processes, each with a trigger, an output and a rough annual volume. No ranking, no assessment.
Lay bare — price the current state
- Question
- What does each candidate cost per year today?
- Outcome
- An annual cost range in euros per candidate, calculated at fully loaded rates and reproducible on a single page. The list from F becomes sortable.
Observe — find the real bottleneck under stress
- Question
- Which step actually blocks when things get tight?
- Outcome
- A bottleneck probability per step, a lead time distribution (P10, P50, P90) for the process, and a utilisation figure per role. Exactly one step is named as the bottleneck.
Weigh — decide whether the fix is worth it
- Question
- Is the intervention worth it — and from when?
- Outcome
- A reasoned decision: which lever, what it costs, what it saves, when it pays for itself. Or the finding that none of them does.
Then refine
Four steps until the improvement is proven. Only now is anything touched, and only what the measurement named.
Reduce — strip the ballast
- Question
- What can go before anything gets built?
- Outcome
- A process with fewer steps, fewer system changes and fewer waiting loops — with no licence bought.
Enable — make ownership explicit
- Question
- Who decides, who executes, who is merely informed?
- Outcome
- For every remaining step, exactly one accountable role, a named deputy and a decision limit.
Fit — exactly one lever, cleanly wired
- Question
- Which single technical lever gets built, and how is it wired in?
- Outcome
- One implemented change at one point, connected to the systems already running, with documented behaviour on failure.
Yield — prove the return
- Question
- What did the intervention achieve — and where is the bottleneck now?
- Outcome
- A before-and-after in euros and lead time from the same arithmetic as L and O, the deviation from the estimate in W, and the newly named bottleneck.
What the method does not do
- 01It does not replace knowledge of the process. The numbers come from conversations with the people who run it; poor conversations produce a poor measurement.
- 02It does not produce point values. Every result is a range. Anyone who needs two decimal places to decide is deciding on instinct anyway.
- 03It says nothing about processes whose cost cannot be stated in euros. A step that carries reputation, legal certainty or due care is not assessed here — it simply stays.
- 04It is not a maturity model and not a certification. A cycle ends with a changed process and a before-and-after calculation, not with a grade or a seal.
Questions about the procedure
In this order: collect candidates (F), price each candidate per year (L), simulate the most expensive one under load to find the step that actually blocks (O), and work out whether an intervention pays for itself (W). Only then does anything change: strip ballast (R), make ownership explicit (E), build exactly one technical lever (F), re-measure with the same arithmetic (Y). The first four steps change nothing — they produce the figure everything later is measured against.
Because the usual three — analyse, implement, review — collapse precisely the points where projects fail. Between analysis and implementation sit two separate decisions: which step is the bottleneck (O) and whether the intervention pays (W). Between the decision and the technology sit two steps that cost nothing and often deliver more than the technology does: clearing ballast (R) and settling ownership (E). Folded into “implement”, they get skipped in practice.
O can be skipped when the process is small enough that the bottleneck is obvious — though the assumption should then be checked in Y. R and E cannot be skipped: they cost nothing and they decide whether the lever from F lands on a clean process or an overgrown one. Skip L and you have no baseline in Y, which means you will never know whether the intervention achieved anything.
That depends on the availability of the people who run the process, not on the method. FLOW is collection and arithmetic and moves quickly once the conversations happen. REFY depends on implementation: R and E are decisions, F is construction, and Y requires the new state to have run long enough to be routine — measure earlier and you are measuring the changeover, not the process.
FLOW changes nothing. It collects, prices, simulates and costs — it ends with a figure and a decision. REFY changes things and then proves it: first what costs nothing (ballast, ownership), then exactly one technical lever, then the re-measurement. The separation is the point of the method. Collapse it and you change things before knowing what they cost.
If you want to know what your process costs: measure it.
Start with the free diagnosis or download FlowVisual. If you want to speak to someone afterwards, we're reachable.
Remote · fixed price · result in euros