QUANTIFYING ORGANISATIONAL DRAIN AND ROI.

How to read this paper.

Every number in this paper belongs to one of three categories, and we will tell you which, every time.

Published research. Figures from independent, named sources — Crucial Learning, Leadership Circle, Gallup. We cite them; we did not produce them.

Your observed data. Figures that come from your own organisation: what you count when you watch your leaders for two weeks. We cannot know these. Only you can.

SSC's modelled assumptions. Figures we chose — deliberately, conservatively, and openly — to connect the first two categories into an estimate. The most important of these is our 70% recovery assumption, and we will be explicit about what it is and is not.

Most ROI papers blur these three categories into one confident number. That is precisely why sophisticated buyers stop trusting them. We would rather show you a smaller number you can defend in front of a CFO than a spectacular one you cannot. Everything that follows is built on that principle.

1. The invisible line item.

Somewhere in your organisation this morning, a meeting ended on time, everyone nodded, and the one thing that needed saying went unsaid.

Nobody logs it. No system records it. It does not appear in any dashboard. But the decision that meeting existed to make will now take three more weeks, two more meetings and one escalation — and the person who saw the problem coming will watch it arrive, silently, exactly as they predicted.

A defensive behaviour is any moment a leader's reflex stops a conversation that needed to happen — avoiding it, deflecting it, controlling it, or smoothing it over. Each one is a conversation that did not happen, and will not happen tomorrow unless something changes at the pattern level.

Individually these moments are small. That is what makes them invisible. But they are not rare, they are not random, and they are not free. They repeat — because they are driven by an automatic pattern, not a conscious choice — and repetition is what turns a moment into a line item. This paper is about putting a defensible number on that line item, and about what a single, focused intervention can recover.

2. The seven behaviours.

SSC works with a taxonomy of seven defensive behaviours. Each one stops a specific kind of conversation, and each unstopped conversation has a specific organisational cost.

Two things matter about this list. First, these behaviours are observable. You do not need a psychometric instrument to count them — you need two weeks and a scorecard, which is exactly how our measurement ladder begins (Section 7). Second, they are patterned. A leader does not smooth occasionally and at random; they smooth in the same situations, with the same triggers, at roughly the same rate. That regularity is what makes annualisation legitimate — and what makes a pattern-level intervention possible.

3. What the published research establishes.

Three independent strands of research support the model. It is worth being precise about what each one establishes — and what it does not.

Strand one: an avoided conversation has a measurable cost. VitalSmarts — now Crucial Learning — has investigated this twice. An earlier study of roughly 650 people found employees waste an average of $1,500 and a full working day for every crucial conversation they avoid. Their 2016 Costly Conversations study of 1,025 people found every failed or avoided conversation costs an organisation an average of $7,500 and more than seven working days — with one in five respondents putting their worst single avoidance above $50,000. The direction is consistent with the large bodies of work from Gallup and McKinsey on the cost of poor workplace communication. SSC's model defaults to $1,500 — the earlier, lower published average, roughly a fifth of the more recent figure — converted to your currency at live rates.

Strand two: reactive patterns in leaders are the norm, not the exception. Leadership Circle's normative research — one of the largest leadership-assessment databases in the world (Anderson & Adams) — indicates that 70–80% of leaders lead primarily from reactive rather than creative structures: behaviour driven by automatic self-protective patterning rather than conscious, chosen leadership. This is a prevalence figure — SSC's model uses 75%, the midpoint of the published range. It tells you how many of your leaders are likely to be generating defensive behaviours at a meaningful rate. It says nothing about recovery — that distinction matters, and we return to it in Section 4.

Strand three: disengagement is a parallel, independent measure of the same underlying drain. Gallup's State of the Global Workplace (2026) reports that only 20% of employees globally are engaged — 64% not engaged, 16% actively disengaged — at an estimated cost to the global economy of around $10 trillion in lost productivity, roughly 9% of global GDP. Gallup's methodology has nothing to do with counting conversations. That is exactly why we use it: as an independent cross-check that arrives at the same conclusion — the cost of human behaviour in organisations is large — by a completely different route (Section 6).

What this evidence does not establish. None of these sources measured SSC. They establish that the ingredients of the model are real: avoided conversations carry real costs, reactive patterning is widespread, and behavioural drag on organisations is enormous by any methodology. They do not — and cannot — establish what percentage of that cost a single SSC session recovers. That number is ours, it is an assumption, and we label it as one throughout.

4. The model, step by step.

The model multiplies five quantities. Each one is named, sourced, and — in the live calculator — editable, because an assumption you can inspect and override is worth more than one you have to take on faith.

Step 1 — Observe. Count blocked conversations across a sample of your own people leaders for two working weeks, using the seven-behaviour scorecard. This produces a rate: blocked conversations per person per fortnight. (Your observed data.)

Step 2 — Annualise. Multiply the fortnightly rate by 22, on the assumption that the rate you observed holds across 220 working days per year. If the fortnight you watched ran hotter than a normal one, lower the factor. (SSC assumption, editable.)

Step 3 — Scale to reactive leaders. Apply the rate not to your whole leadership population but to the 75% likely to be operating reactively — the midpoint of Leadership Circle's published 70–80% range. We deliberately cost only these leaders. (Published research, editable.)

Step 4 — Price the consequence. Multiply the annual volume of blocked conversations by $1,500 — the earlier of two published averages ($1,500 and $7,500) for the downstream cost of a single avoided conversation. (Published research; the lower average is our conservative choice; editable.)

Step 5 — Apply the recovery assumption. SSC works 1:1 with each leader in a single 90-minute session to recode the one keystone behaviour driving the pattern; the downstream behaviours quieten once it stops being triggered. We model the recoverable share of the annual cost at 70% — recovery that any pattern-level intervention could claim; SSC's route is a single session per leader, embedded through the organisation's existing executive coaching and diagnostics rather than around them. (SSC assumption, editable — see below.)

75% and 70% are different numbers doing different jobs

These two figures sit close together and are easily conflated, so let us separate them explicitly. 75% is a prevalence figure: the proportion of leaders likely to be operating reactively. It is the midpoint of Leadership Circle's published 70–80% range. 70% is a recovery assumption: the share of the modelled annual cost we expect a single-session intervention to remove. It is SSC's own planning figure. It is not published research, it is not derived from the 75%, and the two are not interchangeable. We chose 70% deliberately — below what the single-session literature and our own case experience would suggest is possible — because a conservative assumption that later proves low is a much better problem than the reverse. Section 7 describes how we are converting this assumption into measured evidence.

5. A worked example.

Take an organisation of 1,000 people with 250 people leaders — consistent with the typical ratio of one leader for every three to five employees — and an average salary of $65,000. An observer watches 8 people leaders for two weeks and counts 18 blocked conversations.

Two honest observations about this result.

First, $13.96M is a large number against a $65M payroll — roughly 21% of it. That is not a payroll claim. The $1,500 figure does not price the salary time of the conversation itself; it prices the downstream consequences of its absence — the delayed decision, the rework, the resignation, the project that slips. Consequence costs routinely exceed payroll-share intuitions, which is exactly why they stay invisible on a P&L. Even so, if the number strains credulity in your context: halve it. Halve it twice. The point of Section 8 is that the case survives.

Second, the per-leader figure is the one that matters, and it is worth noticing what drives it. The $51,975 recovery per reactive leader depends only on the observed rate, the annualisation factor, the consequence cost and the recovery assumption. Headcount and leader count cancel out entirely. Your organisation's size scales the total; it does not inflate the per-leader economics. That is what makes the per-leader number the honest unit of decision.

6. The independent cross-check.

The same example organisation, run through Gallup's methodology instead of ours: applying the 2026 global split to 1,000 employees gives 640 not-engaged people at a 5% salary productivity loss (SSC's conservative weighting) and 160 actively disengaged at 18% (Gallup's published figure) —

These two numbers — $13.96M and $3.95M — are not designed to match, and it would be a red flag if they did. They measure different things by different methods: our model prices consequence-per-incident concentrated in your reactive leaders; Gallup's prices productivity loss spread across your entire workforce. What the pair establishes is the thing that matters: the cost of human behaviour in your organisation is significant by any methodology you choose. One method is ours. One has nothing to do with us. They agree on the conclusion.

Expect the gap, and read it correctly. The defensive behaviour estimate will typically run three to four times the Gallup figure, and the reason is structural: Gallup's measure is bounded by payroll — productivity drag can never exceed a share of salary — while consequence costs have no ceiling; one blocked decision can cost more than the salaries of everyone in the room. Read Gallup as the floor; read the defensive behaviour figure as the estimate. A buyer who trusts only the floor is still looking at $3.95M a year in this example. And for calibration in the other direction: pricing the same conversation volume at the 2016 published average of $7,500 would put the defensive behaviour estimate at $69.8M — the $1,500 default is a deliberate four-fifths discount on the most recent published research.

Per head, the two measures are not even the same species: Gallup prices roughly $3,952 of drag per employee; the defensive behaviour estimate prices $74,250 of consequence per reactive leader. And notice where the methods meet. Halve the estimate's two key assumptions — the observed rate and the consequence cost — and its $3.49M lands beside Gallup's $3.95M floor. Two unrelated methodologies converge at the conservative end; they diverge only in how much upside they can see. (At very low observed rates the defensive behaviour estimate can even sit below the floor — the floor still stands, as the independent minimum.) The calculator and report now present this as a single range — from deliberately hostile to the newest published research — so a reader meets one honest continuum rather than two competing claims.

The 5% in context

The most contestable-looking number in the Gallup comparison is the 5% productivity weighting we apply to the not-engaged majority — so it is worth showing where it sits against every independent attempt to measure the same loss.

Every full-phenomenon estimate in the table lands above our figure; the only entry below it is a deliberately single-cause subset. We assume 5% anyway. A comparison built on the lowest defensible weighting cannot be accused of engineering its own conclusion — and if the true figure is higher, as the evidence suggests, the floor is understated, not the estimate.

7. From assumption to evidence: the measurement ladder.

A model is only as trustworthy as its path to being checked. SSC's measurement design has three rungs, in ascending order of evidential weight.

Rung 1 — Estimate. A two-minute structured self-estimate across the seven behaviours, on time-anchored scales (never / monthly / fortnightly / weekly / daily). An honest first pass, nothing more.

Rung 2 — Observe. The two-week scorecard: one observer, real counts, real leaders — initials noted against marks where possible, because concentration identifies the pilot cohort. This is where the worked example above begins — counts replace recall.

Rung 3 — Corroborate. Four or five colleagues observing independently and averaging. Inter-observer agreement is what turns a count into a figure you can take to a board.

Beyond the ladder sits the piece that converts our 70% assumption into evidence: longitudinal measurement of the intervention itself. SSC pilot engagements are instrumented at three points — baseline before the session (T0), two months after (T2) and six months after (T6) — measuring behaviour frequency against the same scorecard taxonomy. This is the study that can confirm, raise or lower the 70% figure, and we will publish what it shows either way. The instrument need not be ours alone: where an organisation already runs 360s, engagement surveys or coaching diagnostics, those join the baseline — the recovery assumption is then tested against measurement the organisation already trusts. Organisations willing to run a measured pilot cohort are, frankly, the collaborators we most want: you get the intervention at the front of the evidence curve; the field gets the data.

8. Sensitivity: the honest range.

This paper does not publish a single ROI ratio. A single ratio is false precision: it multiplies one observed number by three assumptions and presents the product as a fact. The honest statement is a range — and what matters about a range is its floor.

The table below stress-tests the per-leader economics from the worked example. Each row makes the model's assumptions progressively more hostile.

Read the bottom row carefully, because it is the most important number in this paper. Assume the observer overcounted by more than half. Assume the true consequence cost is half the lower published average. Assume the intervention recovers only half of what we model. Apply all three penalties simultaneously — and the modelled annual recovery is still $5,775 per reactive leader, every year, against a one-time 90-minute engagement. At organisational scale, the same hostile case for our 1,000-person example still recovers roughly $1.1M annually, against $9.8M in the default case.

We could publish the ratio the default case implies. We choose not to, for the same reason we default to the lower published average: a number that survives hostile assumptions is worth more than a spectacular one that does not. The specific investment conversation belongs with your SSC director, against your own observed data — not in a white paper's headline.

9. Why a single session can move a pattern.

The economics only matter if the intervention works, so it is fair to ask: how can 90 minutes change a behaviour a leader has run for twenty years? The answer sits in two bodies of evidence — and it is important to be precise about what they validate.

The mechanism. Defensive behaviours are not knowledge problems; a leader who takes over already knows delegation matters. They are prediction problems. The behaviour is generated beneath deliberate thinking: the brain carries an emotional learning — a wired association between a class of situation and a threat — and produces the protective behaviour automatically when the situation appears (Damasio's somatic marker research; Porges's work on threat physiology; Friston's predictive-processing account of the brain). This is why awareness-based development so often fails to change behaviour: it trains the deliberate system while the automatic one keeps firing. The clinically significant finding is that these emotional learnings are rewritable. Memory reconsolidation research (Ecker, Ticic & Hulley, Unlocking the Emotional Brain) shows that when a stored emotional learning is activated and simultaneously met with an experience that mismatches its prediction, the original learning itself updates — not suppressed by willpower, but rewritten at the level where the behaviour is generated. Not awareness. Change. This is the mechanism an SSC session is engineered around: surface the pattern, decode the prediction driving it, and resolve it in a single concentrated arc.

The format. Independently, four decades of single-session therapy research (Talmon, Single Session Therapy, 1990; Hoyt & Talmon, Capturing the Moment, 2014) establish two findings that surprised the field that produced them: the most common number of sessions any client attends is one, and a majority of single-session clients report durable improvement at follow-up. Depth of contact matters more than duration of contact. The multi-month default in coaching is a business-model convention, not an evidential requirement.

What this evidence validates — and what it does not. These two literatures validate the mechanism (patterns held as emotional learnings can update in a single, well-structured experience) and the format (one session is a legitimate, evidenced dose). They do not, by themselves, validate SSC's specific 70% recovery figure in corporate settings. No published study could — the studies were not measuring our intervention. That validation can only come from the T0/T2/T6 measurement described in Section 7, which is precisely why we run it. We hold the distinction openly: the science says the approach can work in one session; our own longitudinal data will say how much it recovers.

10. Limitations.

A model you can trust is one whose authors can name its weaknesses. Ours has five.

Observer effects. A single rung-2 observer can overcount or undercount, and knowing one is observed can change behaviour. This is why rung 3 exists — independent observers, averaged — and why the calculator warns when an entered rate exceeds anything in the source research.

The annualisation assumption. ×22 assumes your observed fortnight was typical. Quarter-end, restructure season or crisis periods run hot; the factor is editable downward for exactly that reason.

Transferability of the consequence cost. The $1,500 and $7,500 averages were researched in North American workplace populations. Consequence costs plausibly vary by sector, salary structure and decision velocity. Defaulting to the lower average absorbs much of this risk; it cannot eliminate it.

The prevalence instrument. The 75% reactive figure comes from one (very large) assessment methodology. Other instruments would draw the reactive/creative line elsewhere. The figure is editable, and the sensitivity table shows the model does not depend on it being exact.

The recovery assumption is not yet longitudinally validated. Stated throughout, and worth stating once more plainly: 70% is our planning assumption, chosen conservatively, pending our own published T0/T2/T6 results. Anyone who tells you they have a validated recovery percentage for a leadership intervention without showing you their longitudinal instrument is selling you the blur this paper was written to remove.

11. Constants and sources.

All research figures are held in USD and converted to local currency at live exchange rates in the calculator. The interactive model, with every constant above exposed and editable, is at singlesession.online — the live calculator and this paper carry the same figures by design (calculator v2.9.8).

Revision note.

This second edition replaces single-point estimates with ranges, attributes every constant to its source and status (published research, observed data, or SSC assumption), separates the mechanism evidence from the recovery assumption throughout, corrects internal figures for consistency with the live calculator, and withdraws the single-ratio ROI presentation of the first edition in favour of the sensitivity analysis in Section 8. This printing (2.1) also corrects the conversation-cost attribution to its primary published sources and aligns the estimate-scale description with the live calculator. Rev. 2.2 adds independent-evidence context for the 5% engagement weighting (Oxford–BT; GCC; Deloitte; ADP).

References.

ADP Research Institute (2019). The Global Study of Engagement. ADP, LLC. 19 countries; 19,000+ workers; 16% Fully Engaged.

Anderson, R. J. & Adams, W. A. (2015). Mastering Leadership. Wiley. Authors' published estimate: 70–80% of leaders lead Reactively.

Bellet, C., De Neve, J-E. & Ward, G. (2019). Does Employee Happiness Have an Impact on Productivity? Saïd Business School Working Paper 2019-13, University of Oxford.

Crucial Learning (formerly VitalSmarts). Conversation-avoidance research (n≈650; $1,500 average) and Costly Conversations study (2016; n=1,025; $7,500 average). cruciallearning.com.

Damasio, A. (1994). Descartes' Error: Emotion, Reason, and the Human Brain. Putnam.

Deloitte (2024). Mental Health and Employers. Deloitte LLP. Poor mental health: £51bn a year to UK employers; presenteeism £24bn.

Ecker, B., Ticic, R. & Hulley, L. (2012). Unlocking the Emotional Brain: Eliminating Symptoms at Their Roots Using Memory Reconsolidation. Routledge.

Friston, K. (2010). The free-energy principle: a unified brain theory? Nature Reviews Neuroscience, 11, 127–138.

Gallup (2026). State of the Global Workplace. Gallup Press.

GCC Insights (2016). Presenteeism study validated against the WHO Health and Work Performance Questionnaire (n≈2,000). Global Corporate Challenge.

Goetzel, R. Z. et al. (2004). Health, absence, disability, and presenteeism cost estimates of certain physical and mental health conditions affecting US employers. Journal of Occupational and Environmental Medicine, 46(4).

Hemp, P. (2004). Presenteeism: at work — but out of it. Harvard Business Review, 82(10).

Hoyt, M. F. & Talmon, M. (eds.) (2014). Capturing the Moment: Single Session Therapy and Walk-In Services. Crown House.

Porges, S. W. (2011). The Polyvagal Theory. Norton.

Talmon, M. (1990). Single Session Therapy: Maximizing the Effect of the First (and Often Only) Therapeutic Encounter. Jossey-Bass.