The GROW Model, Reviewed: Where It Came From, What the Evidence Supports and Where It Turns Into a Script

What's the most widely taught management conversation in the world? Almost certainly a four-letter one. GROW – Goal, Reality, Options, Will – has been the backbone of coaching training for over thirty years. If a manager anywhere has been on a coaching course, the odds are good they came home with those four letters in their notebook.

The model earns a careful review for the same reason CliftonStrengths did: enormous adoption, warm reputation, and a technical record that gets far less attention than the framework itself. I've spent much of my career around the conversations GROW was built to improve, so this review is written with respect – and with the research open on the desk. Here's where the model came from, what the evidence supports, and the point at which a helpful map becomes a script.

Where it came from

Who invented the GROW model?

Three people, gradually: John Whitmore, Graham Alexander and Alan Fine developed the sequence through their consulting work in the mid-1980s, much of it at McKinsey, translating Timothy Gallwey's Inner Game thinking into a teachable business conversation. The trio agreed no one would claim more credit than the others - an agreement history has not entirely honoured, since Whitmore's 1992 book Coaching for Performance made the model famous and the other two names rarely appear beside it.

The story starts on a tennis court. In 1974 Timothy Gallwey, a Harvard-educated tennis coach, published The Inner Game of Tennis, which argued that performance improves less by instruction than by reducing the interference in a player's own head. The book sold more than two million copies, and its central equation – performance equals potential minus interference – crossed from sport into business with remarkable speed.

One of the people it crossed with was John Whitmore, a former racing driver who had won his class at Le Mans in 1959. Whitmore trained with Gallwey, brought the Inner Game to Britain, and with two colleagues – Graham Alexander and Alan Fine – spent the mid-1980s turning its spirit into a teachable sequence for business conversations. The backstory deserves to be told properly, because it says something about the model. The four-step sequence emerged gradually through the trio's consulting work, much of it at McKinsey, and the GROW label itself was suggested by one of McKinsey's communications specialists, who spotted that the steps could be made to spell something. Fine has said the three of them agreed that no one of them would claim more credit for the basic model than the others – an agreement that history has not entirely honoured, since Whitmore's book made the model famous and the other two names rarely appear beside it.

That book was Coaching for Performance, published in 1992, now in its sixth edition and – by its publisher's count – past a million copies in more than twenty languages. Performance Consultants, the firm Whitmore co-founded, calls it "the world's #1 coaching book" and notes the model's use at organisations including Google, eBay, Barclays and Volvo. Whitmore himself received the International Coach Federation's President's Award in 2007, and by the time of his death in 2017 he was routinely described as the father of workplace coaching. As with Don Clifton, the description is fair. Before Whitmore, "coaching" in business mostly meant remedial instruction. After him, it meant a structured conversation in which the other person does the thinking. That shift deserves the credit it gets.

What it actually is

What does GROW stand for in coaching?

GROW stands for Goal, Reality, Options and Will: what do you want, what is happening now, what could you do, and what will you do. The sequence was built in the 1980s by John Whitmore, Graham Alexander and Alan Fine from Timothy Gallwey's Inner Game work, and the label itself was suggested by a McKinsey communications specialist who spotted that the steps could be made to spell something.

GROW is a sequence for a conversation, and it helps to be plain about how modest – and how clever – that is.

Goal. What do you want? Not the problem; the destination. What would you like to walk out of this conversation with, and what does the longer-term version look like?

Reality. What's happening now? Described specifically, in the person's own words, with the coach resisting the urge to diagnose.

Options. What could you do? Plural by design – the stage exists to get past the first answer, which is usually the answer the person walked in with.

Will. What will you do? Commitment, first steps, timing. (Whitmore's W was Will; Fine prefers Way Forward; Alexander taught Wrap-up. The three surviving versions of the fourth letter are a small monument to the model's shared parentage.)

Two design choices explain its longevity. First, the coach asks rather than tells, which operationalises Gallwey's insight without requiring a manager to read Gallwey. Second, the sequence runs goal-first. That ordering sounds trivial and isn't easily dismissed: starting with the destination rather than the problem changes the emotional weather of the conversation, and it anticipated by two decades the research finding that solution-focused questions outperform problem-focused ones. GROW carries no personality theory, no diagnostic, no score. It is a conversational handrail – which is both its strength and the root of most of what goes wrong with it.

The claims

The publisher's claims for GROW are, by assessment-industry standards, restrained. Performance Consultants presents it as the most widely used coaching framework in the world – plausibly true, and hard to check. The bolder claims tend to come from the training industry around it, where GROW is routinely presented as "proven". So it's worth asking the question this series always asks: proven by whom, and at what?

The evidence - three layers

Is the GROW model evidence-based?

Partly. Coaching in general has respectable meta-analytic support - effects of 0.36 to 0.74 depending on the outcome measured. GROW's ingredients, goal-setting and solution-focused questioning, are well supported. But direct tests of the model itself are rare: one comparison found no significant difference in outcomes with or without it. The fair summary is that GROW is plausible and consistent with the evidence rather than proven by it.

The fair way to read the research is in three layers: the evidence for coaching in general, the evidence for GROW specifically, and the evidence for the ingredients GROW is made of. The three layers say quite different things.

Layer one: coaching works. The meta-analytic record here is respectable. Theeboom, Beersma and van Vianen (2014) pooled 18 studies of workplace coaching and found significant positive effects across the board: goal-attainment strongest (an effect of 0.74), then performance and skills (0.60), work attitudes (0.54), well-being (0.46) and coping (0.43). Jones, Woods and Guillaume (2016) reached a similar reading from 17 studies: a positive overall effect of 0.36, with two findings that deserve more attention than they get. Coaching by internal coaches showed larger effects than coaching by external ones, and adding multisource feedback made outcomes slightly worse, not better. Format barely mattered – face-to-face and blended delivery performed alike. Both papers carry the standard caution: studies with proper control groups showed smaller effects than studies without.

Layer two: GROW itself is barely tested. This is the layer the training brochures skip. The SAGE Handbook of Coaching notes that little research has examined the GROW model directly, and what exists is thin. Scoular and Linley (2006) compared coaching sessions run with and without GROW and found no significant difference in outcome quality. On the positive side of the ledger: a small 2021 Korean study of nursing students found GROW-structured peer tutoring improved study habits (fourteen participants, so a straw in the wind), and a 2022 randomised controlled trial by Okorie and colleagues gave 109 Nigerian school administrators nine weeks of GROW-based coaching and found improved subjective well-being, still present at three-month follow-up. That's encouraging, and it is also nearly the whole file. A model taught to millions of managers rests, in terms of direct evidence, on a handful of studies you could read in an afternoon.

Layer three: the ingredients check out - mostly. GROW is built from goal-setting, structured questioning and a solution-focus, and those components have real support. Goal-setting theory is one of the better-replicated findings in organisational psychology. And Anthony Grant's experimental work found that solution-focused questions outperformed problem-focused questions on every measure tested – more progress towards the goal, higher self-efficacy, less negative feeling. GROW's goal-first, options-forward shape sits squarely on the right side of that evidence.

So the fairest one-sentence summary reads: coaching works, GROW-shaped ingredients work, and GROW itself has rarely been put to a direct test – which means the model is best described as plausible and consistent with the evidence, rather than proven by it.

Where it helps, and where it becomes a script

What are the disadvantages of the GROW model?

Three stand out in the research and in practice. Direct evidence for the model itself is thin - a handful of studies, one finding no difference with or without it. The structure rewards mechanical use: run by the letter, it processes people rather than hearing them. And it brings no knowledge of the person into the room - the same four questions land entirely differently on different people, and the model has no way of knowing which person it has met.

Grant – the researcher who has spent most time on this question – put the risk plainly in a 2011 paper titled "Is it time to REGROW the GROW model?". Session structures, he argued, are maps to guide a conversation, and trouble starts when they get taught as ideologies to be rigidly adhered to. Anyone who has sat through a GROW conversation run by the letter will recognise the symptoms: the coach marching through four stations in order, the "Options" stage produced on schedule whether or not the goal has survived contact with reality, the sense of being processed rather than heard. Grant's own repair was to add Review and Evaluate stages – REGROW – because real progress happens across sessions, and a model with no memory tends to produce conversations with no follow-through.

Whitmore, to his credit, said much the same thing first. Coaching for Performance is insistent that the sequence has little value without the two qualities it exists to serve – awareness and responsibility – and that a GROW conversation conducted without them is just a questionnaire read aloud. The model's originator, in other words, agreed with its best critic: the letters are the least important part.

There's a second, quieter limitation worth naming. GROW structures a conversation; it brings no knowledge of the person into the room. Two colleagues can answer "What's your goal?" in ways so different that the same four questions produce entirely different conversations – one thinks aloud and needs the space to ramble, the other arrives with a finished answer and finds the Options stage patronising. The model has no way of knowing which is which. That gap sits outside GROW's job description, but a coach who walks in with nothing else will feel it by the second session.

What to take from it

Where this lands, then. GROW earns its place: it translated a real insight about interference and self-directed learning into a form a busy manager can hold in their head, it made asking respectable in cultures built on telling, and its goal-first design anticipated where the questioning research landed. Its shared, slightly contested parentage produced something with the durability of folk wisdom – simple, portable, hard to improve without complicating.

Its limits are equally real: direct evidence that is thin to the point of transparency; a structure that rewards mechanical use in exactly the settings most likely to teach it mechanically; no memory between sessions; and no knowledge of the person being coached. The coaching effect sizes in the meta-analyses are respectable rather than miraculous – 0.36 to 0.74 depending on outcome – and none of them can be claimed by GROW specifically.

Use it as its builders used it: a map for a conversation between two people, one of whom is doing the thinking and the other of whom is protecting the space to think in. Retire it the moment it starts running the conversation for you.

The four questions, properly sequenced

How do you use the GROW model well?

One question per stage, sequence intact: what would make this conversation worth having; what is actually happening and what have you already tried; what would you do if the constraint disappeared; what will you do before we next speak, and how will I know. Then the discipline that makes it work: loop back. If Reality reshapes the goal, return to Goal and re-ask it. A conversation that never revisits G is a sign the model is driving, and it should never be driving.

Copy and paste this into your notes for your next one-to-one. One question per stage, sequence intact, plus the discipline that makes it work:

1. Goal: "What would make this conversation worth having?" (Ask before the problem is described, not after.)
2. Reality: "What's actually happening – and what have you already tried?" (The second half saves the most common wasted ten minutes.)
3. Options: "If that constraint disappeared, what would you do?" (Take the first two answers as warm-up; the third is usually theirs.)
4. Will: "What will you do before we speak next – and how will I know?" (A date beats an intention.)

The discipline: loop back. If Reality reshapes the goal, return to Goal and re-ask it. A GROW conversation that never revisits G is a sign the model is driving, and it should never be driving.

Where our thinking sits

Readers of this site will know where Sariio stands: we build the piece GROW leaves out – a current, movable reading of how each person prefers to work, so the questions above land on the person actually in the room. The model supplies the sequence; it has never claimed to supply the knowledge. The two turn out to go rather well together.