Two numbers appear in almost every vendor pitch for transfer of training. Two numbers appear in almost every vendor pitch for transfer of training: that between 5 and 10 percent of training alone reaches the job, and that between 80 and 90 percent does when coaching is added (Baldwin & Ford, 1988; Joyce & Showers). Both figures are widely cited in the field. Both get quoted far past what the underlying research, where retrievable, can carry.
If you buy training for a mid-market or enterprise workforce, this post is for you. It is a source audit, not another summary. We trace each number to the work behind it, say what was actually measured, and state plainly what the citation cannot support. You will read what transfer of training means and why the definition decides the number, what Baldwin & Ford actually reviewed, what Joyce & Showers supports, why the figures are benchmarks for design choices and not guarantees, why most transfer research is not about people managers, why a 90 day boss scored measure tests typical transfer, five questions to audit a vendor’s claim, and what this means when you are buying. Tandem Academy is $1,000 a year or $99 a month, and enterprise pricing lives on the cohort page. Ask which study a claim maps to before you accept it.
Who This Is For, and Why the Two Numbers Keep Showing Up
This is written for HR, L&D and executive buyers at mid-market and enterprise companies who have to approve a training purchase, and for anyone handed a slide with the 5 to 10 percent transfer figure (Baldwin & Ford, 1988) on it.
Those figures appear in pitch decks, conference talks and procurement documents. The training-alone number is credited to Baldwin & Ford (1988), the 80 to 90 percent coached number to Joyce & Showers. Both are quoted far more often than they are sourced. For what the evidence says to do instead, see Manager Training: Why Most of It Fails.
Every mention here carries its citation, and where a primary source has not been retrieved, this piece says so rather than smoothing it over.
If you run a small business or manage a team without a training budget, the audit method below still applies to anything you are asked to buy. A smaller-budget option and its price appear later.
What Is Transfer of Training, and Why the Definition Decides the Number
The definition decides the number. Baldwin & Ford (1988) define transfer as the degree to which trainees effectively apply the knowledge, skills and attitudes gained in a training context to the job (p. 63).
That definition sets a high bar. A course that changes what people know has not transferred. Transfer is what happens on the job afterward.
Transfer is not one construct. The literature separates maintenance from generalization, near transfer from far, and typical transfer from maximum transfer.
Typical transfer is what happens over time without explicit prompts or evaluation. Maximum transfer is what a study produces under ideal support. A single percentage can be true of one and false of the other.
So the first question about any transfer number is not how big it is. It is which kind of transfer the study behind it measured. That distinction decides what belongs in the curriculum. For the content that follows from it, see this research-grounded content map for new manager training.
The 5 to 10 Percent Figure: What Baldwin & Ford (1988) Actually Is
It is a literature review, not an experiment. Baldwin & Ford (1988) appeared in Personnel Psychology, volume 41, number 1, pages 63 to 105. The paper reviews and critiques existing research on how training design, trainee characteristics and work environment affect the conditions of transfer. The paper’s scope is critique and synthesis; no single transfer rate is described in the ERIC abstract or in the review’s widely cited framework.
The model it sets out has three inputs: trainee characteristics, training design and work environment. Those feed training outputs and conditions of transfer. Check a vendor claim against that model, not against a headline percentage.
Any 5 to 10 percent figure has to come from the individual studies the review synthesized. Those studies used different measures, different populations and different follow-up intervals, so the number carries all of that variation with it.
The review also names research gaps, including the need to test different ways of operationalizing training impact on transfer. So the citation cannot support a claim that a specific program transfers at 5 to 10 percent, or that your managers will. Read it as one design instruction: name the behaviors in advance, the way a management development plan that actually changes behavior does.
The 80 to 90 Percent Figure and Joyce & Showers
The coached figure, 80 to 90 percent transfer with coaching, is credited to Joyce & Showers, and this piece cites that source every time the number appears. The primary source has not been retrieved.
Four details still have to be checked against it: the study design, the sample, the follow-up interval, and what the word transfer meant in that work. Those four decide what the number can support. Until the original is read, the coached figure is not a settled experimental result, and this audit will not treat it as one.
What the two figures describe is a difference in design, not a difference in effort or motivation. Coaching, feedback and follow-up are design variables, the same ones Baldwin & Ford (1988) placed in their model, which treats training design as an input that includes support, feedback and job relevance.
The defensible reading: reinforcement after training is associated with far higher application than training alone. That pattern is the same one behind why change management fails when reinforcement never arrives. The exact percentage inside your organization stays unmeasured until you measure it.
The Numbers Are Benchmarks for Design Choices, Not Guarantees
The 5 to 10 and 80 to 90 percent figures (Baldwin & Ford, 1988; Joyce & Showers) point toward a direction: training alone tends to fade, reinforcement tends to stick. A number without a population, a context and a measurement method is decoration, not evidence.
Baldwin, Ford and Blume make that point from inside the field. They call for richer reporting on trainees, trainers and organizational context, for optimizing transfer rather than only learning, and for broader measurement of transfer outcomes. Writers in this field have described a crisis of relevance in academic work (Hoffman, 2016, cited in Baldwin, Ford and Blume). The distance between what studies measure and what buyers need to know is part of the problem.
So treat the benchmarks as design instructions. Build the reinforcement into the plan, starting with First-Time Manager Training: What to Cover in the First 90 Days (and What Can Wait) rather than buying the event and hoping.
The Manager Gap: Most Transfer Research Is Not About People Managers
McCaffrey, Howe and Burr note that few transfer studies use managers as subjects. Most involves other populations, so applying the 5 to 10 and 80 to 90 percent figures (Baldwin & Ford, 1988; Joyce & Showers) to first-time managers is an inference, not a finding.
Those authors synthesize two management transfer studies and find a complex interaction between training design and work environment. The interaction matters here. A manager who returns from training to a boss who never asks for the new behavior will not apply it, however good the course was.
That gives you a second audit question: does the study population look like the people you are training? If not, ask what makes the translation safe.
It also explains why a ninety-day measure beats an end-of-course score for management training. Coaching for New Managers: Why the First 90 Days Change Everything shows what structured support in that window produces. The work environment is part of the test, not an excuse for a weak result.
Why a 90 Day, Boss Scored Measure Tests Typical Transfer
Typical transfer, as defined earlier, is the condition most training actually faces, and the condition a buyer should want measured.
Who rates the result changes the answer. Self-report tends to run high. A knowledge test shows learning, not application. A manager’s own boss has a reason to be accurate, because that boss sees the behavior at work.
The interval changes the answer. A score taken on the last afternoon of a course cannot speak to transfer. Transfer has not had time to happen. A satisfaction score cannot answer the transfer question at all. It measures reaction to the event, a different construct from application on the job. For more on matching the measure to the question, see what a leadership assessment should actually measure.
One concrete version of this design is the Manager Effectiveness Index. It scores fifteen manager behaviors across five dimensions: clear expectations, coaching conversations, delegation, accountability, feedback and trust. Each manager’s own boss scores them before, at the end, and ninety days later. Behavior is named in advance. The rater is identified. The interval is long enough for the result to mean something. That structure tests typical transfer rather than the reaction to a good event.
How to Audit a Vendor’s Transfer Claim: Five Questions
The Index shows what a defensible measure looks like. Audit a vendor claim the same way. Five questions.
- Which study, and what kind? A literature review, a single experiment and a meta-analysis support different sentences. Baldwin & Ford (1988) is a review, so it supports statements about the conditions of transfer, not a rate for one program.
- What was measured? Ask which type of transfer the claim maps to, near or far, typical or maximum, and expect a specific answer.
- Who rated it? Self-report, a test, a supervisor score or an objective outcome. The less self-reported, the more it tells you.
- At what interval, against what comparison? Same-day scores measure reaction. A ninety-day follow-up with a baseline or comparison group starts to say something about behavior change.
- Would the sample look like ours, and what conditions does the claim depend on? Managers, industry, company size and coaching volume all shape the result. Ask the vendor to name the study design each number maps to, and to state what it cannot prove.
For the wider evidence on programs that fail this test, see supervisor training that actually changes behavior.
What This Means When You Are Buying
The audit ends where the decision starts. Build reinforcement into the plan rather than buying the event and hoping, which is what the benchmarks have always argued.
For mid-market and enterprise buyers, the Manager Performance Cohort covers one capability for up to seven managers, virtual, eight to twelve weeks, measured three times. Enterprise cohorts carry the Ninety-Day Behavior Change Guarantee: if the score does not improve and conditions were met, another coaching cycle runs at no cost. Pricing lives on the cohort page.
The guarantee covers enterprise cohorts only. It does not apply to Tandem Academy, and no behavior-change percentage is claimed for any Tandem program.
For owners and managers at small businesses that cannot buy a $15,000 program, Tandem Academy is $1,000 a year or $99 a month: nine leadership courses, an AI coach available every week of the year, live group coaching capped at ten seats, and assessments.
For association executives, the Academy runs as a non-dues revenue program across the whole network, members plus suppliers, exhibitors and prospects. The association keeps 30 percent, or $300 per person per year, on the first purchase and every renewal.
In every case the measure is the point. Ask what is scored, by whom, and when.
Conclusion: Ask Which Study the Claim Maps To
Both figures are widely cited. Both get quoted far past what the underlying research, where retrievable, can carry.
Baldwin & Ford (1988) is a literature review about the conditions of transfer. It does not report a transfer rate for your program, or for anyone’s. The 80 to 90 percent coached figure (Joyce & Showers) remains unverified until the primary source is retrieved and read directly.
The two numbers earn a place as design benchmarks, and that direction is worth building for. It is not a guarantee.
So when a vendor hands you a transfer claim, ask which study design it maps to, what was measured, who rated it, at what interval, and against what comparison. Then measure your own cohort. Your managers, your raters, your ninety days. That answer is the only one that speaks to your results instead of someone else’s.