The evidence base

Every Number We Use, and Where It Came From.

Vendor research decks tend to blur three different kinds of claim together: what independent research has established, what the vendor has modelled, and what the vendor asserts about its own product. Blurring them is how a business case dies in the second meeting.

So they’re labelled here. If a number carries a Published research tag, an independent source published it and you can check it. If it says Modelled, we calculated it and the assumptions are on the value model. If it says Tradecraft claim, it is a design commitment we have not yet proven at scale — and we say so.

Turnover The manager layer Coordination friction Internal mobility Unused budget How we model What we can’t prove

What Turnover Actually Costs

This is the best-evidenced part of the model and the only tier we would put in front of a CFO unaccompanied.

50–200%

of an employee’s annual salary is the range Gallup gives for the cost of replacing them, varying with role complexity and seniority. This is the multiplier the value model uses, defaulted to 1.0×.Published research

Gallup, This Fixable Problem Costs U.S. Businesses $1 Trillion

$1T

is Gallup’s estimate of what voluntary turnover costs U.S. businesses annually — recruiting, hiring, onboarding, training and the productivity gap during the transition.Published research

Gallup, State of the Global Workplace

52%

of employees who left voluntarily say their manager or organisation could have done something to stop them. This is the single most important number on this page: it means the majority of the cost above is addressable.Published research

Gallup

51%

say that in the three months before they left, neither their manager nor any other leader spoke to them about their job satisfaction or their future with the organisation.Published research

Gallup

What this does not establish. That a coaching platform reduces the number. Gallup’s finding is that the departures were preventable, not that any specific intervention prevents them. The reduction rate in our model is an assumption you set, defaulted deliberately low.

The Layer Everything Is Delivered Through

Every people programme an organisation funds is executed by a manager. The research on that layer is consistently grim, and it has been getting worse rather than better.

15%

of employees say their manager helped them build a career plan in the past six months — down five percentage points year over year. LinkedIn attributes the decline to managers being overwhelmed rather than unwilling.Published research

LinkedIn Learning, 2025 Workplace Learning Report

1 in 5

strongly agree they have talked with their manager in the past six months about the steps to reach their goals.Published research

Gallup

32%

strongly agree their manager helps them set performance goals; 38% say their manager helps them set work priorities.Published research

Gallup

14%

strongly agree that the performance reviews they receive inspire them to improve.Published research

Gallup

~50%

of employees strongly agree they know what is expected of them at work — the most basic precondition for performing well.Published research

Gallup

Why this matters more than it looks. These aren’t engagement-survey soft numbers. They describe a delivery mechanism that is failing, which means every downstream investment — L&D, mobility, retention, performance — is being pushed through a broken channel. Fixing the programme without fixing the channel changes nothing.

The Cost of Not Knowing How Things Work Here

This is the largest pool of value in the model and the one most organisations have never priced. It’s also where you should apply the most scepticism, so the numbers below are the ceiling — not what we claim to recover.

58–60%

of the knowledge worker’s day goes to “work about work” — communicating about work, chasing status, hunting for information, managing shifting priorities — rather than the skilled work they were hired for.Published research

Asana, Anatomy of Work Global Index

209 h

per year, per person, spent on duplicative work — work already done by someone else. Plus 103 hours in unnecessary meetings and 352 hours talking about work.Published research

Asana, Anatomy of Work Index

83%

of teams say they would be more efficient if the right processes were in place.Published research

Asana, Anatomy of Work Index

24 h

per person per year is what the Tradecraft value model assumes can be recovered — roughly 11% of the duplicative-work figure alone, and about 1% of a working year. Of that, only the share you set as redeployed is counted as value. The default is 40%.Modelled

Tradecraft value model

The honest caveat. Asana is a work-management vendor publishing research that supports buying work-management software. The methodology is disclosed and the sample is large, but it is not disinterested, and you should treat it as directional. It is also measuring a broader category than we address: we do not claim to fix your meeting culture or your tool sprawl. We address the slice that is navigational — not knowing who decides, how to raise something, what “good” means here, or whether the answer you’re waiting for is ever coming.

Internal Mobility Is the Lever Nobody Pulls

The research here is unusually consistent across independent sources, which is rare in people analytics.

41%

longer average tenure at companies with high internal mobility compared with companies with low internal mobility.Published research

LinkedIn platform data, Workplace Learning Report

18–20%

more is what external hires are paid than internal candidates promoted into the same role. They also receive significantly lower performance ratings for their first two years and exit at higher rates.Published research

Bidwell, M. (2011), Paying More to Get Less, Administrative Science Quarterly

73% vs 56%

still with the company at three years, comparing employees who made an internal move against those who did not.Published research

Bersin research

24%

of organisations report having a structured internal mobility programme that actually helps people move roles without leaving.Published research

Industry survey data, 2025

How the model uses this. Only the pay premium — 18% of salary on roles shifted from external to internal fill — and only for year one. The tenure and performance effects are real and larger, but annualising them would inflate the line beyond what we can defend in a room.

Money You Have Already Committed and Nobody Spends

Tuition assistance is the clearest example because the utilisation data is published. The same pattern applies to EAP, mentorship programmes and internal learning catalogues.

2–10%

of eligible employees use tuition assistance in a given year. Published estimates vary — Lumina puts it at 2–5%, SHRM-referenced analysis at under 10%, and some surveys report higher. We show the range rather than pick the most flattering figure.Published research

Lumina Foundation; SHRM-referenced analysis

43–57%

of employees are unaware their employer offers the benefit at all, or don’t know its terms. Low usage is mostly a visibility failure, not a demand failure.Published research

Lumina Foundation; industry surveys

~90%

of U.S. employers offer some form of educational assistance. The benefit exists almost everywhere. It reaches almost nobody.Published research

SHRM

Called correctly in the model. Recovering this is not a cash saving — the money was already committed. It is budget converted from unspent to spent on development. It appears in Tier 2 as value, and it is labelled so that nobody presents it to a CFO as money returned.

How We Model Value

Three tiers, never merged. Tier 1 is cash that leaves the business today and stops — auditable against your own finance data. Tier 2 is capacity recovered, grounded in published time-use research but dependent on a recovery rate we assume and you can change. Tier 3 is performance unlocked, which is modelled and unproven.

No hour is counted twice. Recovered hours are valued once, at your loaded cost, and only the share you designate as redeployed. Tier 3 adds only the surplus above cost when that capacity goes into revenue or product work — which is zero unless you explicitly set a return above 1.0×. Double counting recovered time as both a saving and an output is the most common error in vendor ROI models and we have engineered it out.

Effects are counted for one year. The hiring premium persists, tenure effects compound, capability accrues. We count none of that. A three-year NPV would produce a much larger number and a much shorter conversation.

The insight layer carries no dollar value. Knowing where development stalls, which population is quietly at risk, and whether a programme moved anything is the part of Tradecraft we consider most valuable. It changes what you decide to fund. We cannot honestly price a better decision, so it is excluded from the model entirely.

Where we are weak

What We Cannot Prove Yet

You will be asked these questions by your analytics team. Here they are with our answers, before you have to ask.

That Tradecraft reduces attrition. We have no controlled outcome data. The research establishes that most regrettable departures were preventable and that visible internal paths correlate with longer tenure. It does not establish that our product delivers the reduction. That is the pilot’s primary measurement, and we will report it whichever way it lands.

That coaching produces measurable performance uplift. The AI coaching category is too young for anyone to have credible longitudinal outcome data — including the competitors who present it as settled. Our Tier 3 uplift assumption defaults to 0.5% of payroll and should be treated as a hypothesis.

That aggregate insight is reliable in small populations. Our five-person minimum cohort floor is a hard privacy constraint, and it means a small team cannot be reported on at all. In smaller organisations, the insight layer only works with multi-team pooling, and there are structures where it will not work. We would rather tell you that than sell you a dashboard that returns empty.

That capability measurement is precise. We observe capability as a by-product of coaching rather than inferring it from résumés and click behaviour, which we believe is a better method. It is still measurement with error bars, and we state them rather than presenting a clean score.

See How the Pilot Tests All of This