Well-structured training proposals lose all the time. The five causes, the three readers every proposal has, and the clause most consultants refuse to write.
The proposal is the most expensive document you produce and the only one you cannot invoice for. It is also, in most practices, the least examined. When it wins you move on to delivery; when it loses you get a short, polite email that explains nothing, and the next brief is already in the inbox.
So the same proposal goes out again, slightly reformatted.
What follows is not a structure guide. There is no shortage of those, and most losing proposals are correctly structured — executive summary, objectives, approach, modules, timeline, investment, next steps, all present and in the right order. Structure is not what is failing. What is failing is that the document answers a question the reader was not asking, and it usually stopped being fixable several weeks earlier, in a conversation.
Almost every losing proposal is a faithful record of an incomplete discovery call.
If you left that call with a list of desired outcomes and no understanding of what is currently preventing them, you have nothing to propose except activity — and activity is the one thing the buyer can price-compare. A proposal built on "they want their managers to have better conversations" can only differentiate on days, materials and rate, because it contains no claim about cause. The competing proposal will be cheaper.
A proposal built on "their managers avoid the conversation because the last person who raised it was overruled in front of the team" can differentiate on everything, because it is making a claim nobody else in the shortlist has made.
Whether you can write the second kind is decided in the room, not at the desk. What a leadership training request actually sounds like, and where it goes wrong covers that conversation in detail. Everything below assumes you got what you needed from it, and addresses the failures that happen after.
The person who signs is going to have to justify the spend to somebody — a finance partner, a board, an executive sponsor, or just themselves in six months when someone asks how it went. Your proposal is the raw material for that defence. If it gives them a schedule of modules and a day rate, you have handed them a receipt and asked them to build an argument out of it.
The fix is a paragraph most proposals do not contain: a plain statement of what is currently happening, what it costs the organisation, and what specifically will be different. Not an outcomes list — a causal sentence. If you cannot write it in one sentence without the words "enhance," "empower" or "unlock," you do not yet have a diagnosis, and no amount of module detail will cover for it.
This one runs against the advice in every proposal guide on the first page of search results, so it is worth being careful about the evidence.
The TalentLMS 2026 L&D Benchmark Report finds that "difficulty measuring learning ROI is seen as a challenge by only two in ten HR managers," that "75% of HR managers say their company's L&D strategy is aligned with KPIs," and that "76% of HR managers satisfied in 2025, up from 61% in 2022" on budget.
State the sample honestly, because it is small: 101 US HR managers, alongside a separate sample of 1,000 US employees. That is not a basis for a strong claim about the whole market, and anyone telling you otherwise from the same report is overreaching.
What it is good enough for is a warning. If even a substantial minority of your buyers believe their measurement is already fine, then opening with three paragraphs proving that training ROI can be measured is, for those buyers, a solution to a problem they do not have — and worse, it implies they have been doing it wrong. You have spent the reader's first two minutes, which are the only ones you are guaranteed, on an argument that reads as a mild accusation.
Measurement still belongs in the proposal. It belongs in the second half, framed as how you will know the programme is working rather than as proof that measurement is possible. The difference is not cosmetic: one is a claim about the field, the other is a commitment about this engagement.
Here is the single change with the highest return, and it is the one most practices resist.
In "Testing for Competence Rather Than for 'Intelligence'" (American Psychologist, January 1973), David McClelland argued against inferring capability from proxies. "The best testing is criterion sampling," he wrote, and illustrated it plainly: "If you want to know how well a person can drive a car (the criterion), sample his ability to do so by giving him a driver's test. Do not give him a paper-and-pencil test."
To be clear about what is ours and what is his: McClelland was writing about personnel selection, not about proposals. He never mentions them. The application below is our inference, and you should treat it as one.
A proposal is a paper-and-pencil test of your design capability. It asks the buyer to infer, from a description of a module, how good the module will be. Every competitor is submitting the same kind of inference, in similar language, and the buyer has no way to discriminate between them — so they discriminate on price, or on who they already know.
Sample the criterion instead. Put one real page in. One module specified to the level you would actually deliver: the behaviour it targets, the practice reps, the observation instrument, the named source the design rests on. Not an appendix of everything — one page of the real thing.
The objection is obvious and worth taking seriously: you are giving away the work. Two things about that. First, one module out of twelve is not the design, and a buyer who takes it to build the other eleven themselves was never going to sign. Second, it is the only move in the whole document that cannot be matched by a competitor who does not actually have a method. A generic proposal can imitate any section of yours except the page where the work is visible.
This is the failure that costs most, because it usually survives the proposal stage and kills the renewal.
The Kirkpatrick model is the common language here and most L&D buyers already know it. Its fourth-generation form is explicit about a dependency that proposals routinely omit. In An Introduction to The New World Kirkpatrick® Model, Jim Kirkpatrick and Wendy Kayser Kirkpatrick define "Level 3 Behavior is the degree to which participants apply what they learned during training when they are back on the job", and they define the mechanism that makes it happen: "Required drivers are processes and systems that reinforce, monitor, encourage, and reward performance of critical behaviors on the job."
Read that as a contractor. The required drivers are not yours. Reinforcement, monitoring, reward — those sit with the client's managers and the client's systems. You can design for them, you can specify them, you cannot supply them. A proposal that promises Level 3 behaviour change without saying so is promising an outcome that depends on work the buyer has not agreed to do, and has probably not been told about.
The same document defines what you can commit to observing: "Critical behaviors are the few, specific actions, which, if performed consistently on the job, will have the biggest impact on the desired results," and "leading indicators are defined as short-term observations and measurements that suggest that critical behaviors are on track to create a positive impact on the desired results."
That gives you a proposal paragraph with a defensible shape: name two or three critical behaviours, name the leading indicators you will observe and when, and name the required drivers the client supplies. The mechanics of building that chain are in how to design a Kirkpatrick Level 3 measurement plan.
The last one is the most expensive to win.
Some briefs describe a problem that no programme resolves, because the behaviour is not absent through lack of skill — it is absent because the system punishes it, or because someone would have to give something up. Proposing a twelve-week programme against that produces a well-run engagement, good reaction scores, and nothing observable at the end.
You can decline it, or you can reshape it. What you cannot do is accept the brief as written and quietly hope. If you want the test that separates the two cases before you write anything, it is in is training the answer?
Before the proposal goes out, run one pass over it with a specific question: who is this paragraph for?
Almost every training proposal has three readers, and they are not reading for the same thing.
| Reader | What they are reading for | What loses them |
|---|---|---|
| Your L&D contact — usually the person who briefed you | Whether this makes them look competent for having found you, and whether it is deliverable inside their constraints | Anything that implies their existing framework, vendor list or measurement approach is wrong |
| The budget holder — the business leader whose problem this is | Whether the thing that is currently costing them will stop costing them | Module lists. They do not care what happens in week seven |
| The sceptic — the one who has bought training before and watched it change nothing | Whether this proposal is honest about what it cannot do | Confidence without conditions. The absence of any caveat is itself the warning sign |
Mark each paragraph with the reader it serves. Most proposals, marked this way, turn out to serve the first reader almost exclusively — because the first reader is the one you have been talking to, and their questions are the ones fresh in your mind. A paragraph that serves nobody comes out. A proposal with nothing for the sceptic is the one that loses quietly, weeks later, with no reason given.
Write the conditions of success into the proposal.
Concretely, a short section — six or seven lines — that says what the client has to supply for the programme to produce what it is supposed to produce. Something in this shape:
Conditions this design depends on. This programme is designed to change three specific manager behaviours. It will produce those changes only if three things are true on the client side: participants' own managers are briefed on what is being practised and expected to ask about it; the two observation points in weeks six and twelve are scheduled before the programme starts; and the decision-rights question raised in our discovery call is resolved by week four. If any of these is not in place, the design still runs, but we should reset expectations on the behaviour measures rather than report against a target we knew was unreachable.
Four things happen when this is in the document.
It reframes the price. A proposal that lists only what you will do invites a comparison of your day rate against someone else's. A proposal that lists what both sides must do describes a joint undertaking, which is a harder thing to discount.
It disqualifies the wrong client early, which is a saving even though it looks like a loss. A buyer who will not commit to briefing line managers has told you, before you have spent anything, that the behaviour outcome is not really available.
It gives the sceptic the thing they were looking for. They have watched a programme fail before. The one signal that distinguishes you from the people who sold them that programme is your willingness to say, in writing, in advance, under what conditions yours will not work.
And it protects the engagement you actually want, which is the second one. Most losing proposals fail at the pitch. Most losing practices fail at the renewal, when a programme that ran well is judged against an outcome it was never resourced to produce. The conditions clause is the only part of the proposal that is still working for you six months later.
The reason most consultants will not write it is that it looks like hedging at exactly the moment you want to look confident. It reads that way only if it is vague. Written specifically, with named conditions and named dates, it reads as the opposite: as the only document in the shortlist produced by someone who has done this before and knows where it breaks.
Worth saying plainly, since this piece asks you to change how you write a commercial document.
The Kirkpatrick definitions are the authors' own, quoted from their own publication, and they are definitional rather than empirical — they tell you what the model means by a term, not that the model is correct. The survey figures come from a single vendor report with 101 HR manager respondents, which is a small sample and a self-selecting one; they are strong enough to warn you off an opening move, not strong enough to build a strategy on. McClelland's paper is a landmark argument in personnel selection and is more than fifty years old, and the application of it to proposals is ours, not his.
The rest — the three readers, the conditions clause, the criterion-sample page — is practitioner reasoning, not research findings. It should be argued with, and it is described as reasoning rather than dressed up as evidence.
Try it on a brief you are writing now. The Diagnostic Conversation Coach is free, needs no signup and no email address. Describe the gap in your own words and it will work back to the causal claim your proposal needs, the behaviours worth committing to, and the conditions you should be naming. If it does not sharpen the way you run a discovery call, nothing else here will interest you.