The real cost of knowledge loss for a mid-market business.

She’s been with you for six years. She’s the one who knows which vendor invoices need a second look before approval. She remembers that the Tuesday batch run breaks if anyone touches the schedule between 4 and 5 a.m. She has the cell-phone number of the AP person at your third-largest customer because the official email goes nowhere. When her team has a question, she answers. When you have a question, she answers.

Salary: $65,000 a year. U.S. Bureau of Labor Statistics median for an Operations Manager at a small-to-mid business. A fair number. A standard hire. Nothing about her compensation looks like risk.

Then Sarah accepts a job offer somewhere else. Two weeks’ notice. Cordial. She’s leaving for entirely good reasons.

This article is about that first thing.

Four independent research bodies have, in different ways, measured the same problem from different angles. The numbers converge.

1. 42 %

   Knowledge that lives nowhere but in one head

   The share of institutional knowledge — the processes, judgment calls, and shortcuts that keep a business running — that exists only in an individual employee’s memory. Panopto / YouGov · n=1,000+ U.S. workers.

2. 20 h/w

   Hours per knowledge worker spent searching

   Average time looking for information that already exists somewhere inside the company. Coveo’s 3.6 hours per day corroborates within rounding distance. Coveo · Relevance Report, Workplace edition.

3. 58 %

   “New” work that is actually a re-solve

   Share of knowledge-worker time spent re-deriving an answer someone else inside the organization already found — and never captured. Panopto Research · Workplace Knowledge & Productivity Report.

4. 5–12 mo

   Recovery window after one key-person departure

   Time to backfill, ramp the replacement, and recapture lost institutional context. Fully-loaded cost runs 50% to 200% of annual salary. SHRM · Society for Human Resource Management.

These aren’t headlines. They’re consensus figures from four different research bodies pointing at the same root cause.

Take Sarah’s $65,000 salary as the anchor. SHRM’s replacement-cost range puts the fully-loaded cost of a single mid-market salaried departure at 50% to 200% of annual compensation over the six months that follow. Here is what that looks like.

Where the money actually goes.

Five buckets account for the entire range. The first four are visible. The fifth is the one nobody is measuring.

1. 01

   Overtime to cover the gap

   The rest of the team picks up Sarah’s work. Some of it gets done. Some of it gets done at 1.5×. Some of it doesn’t get done, and you find out two months later.

2. 02

   Recruiting and onboarding

   Agency fees run 15–25% of salary at the mid-market layer. In-house recruiter time isn’t free either. Add a referral bonus if you’re lucky enough to hire one.

3. 03

   Ramp time

   Three to nine months before the replacement is functionally Sarah-equivalent — assuming you find one. Until then, output is at 60–80% of where it was.

4. 04

   Lost customers and slipped revenue

   The quiet one. One missed renewal because nobody pinged the customer in week three. One late invoice that pushes a payment past quarter close. Neither shows up on a knowledge-loss line item.

5. 05

   Recreated work

   The 58% re-solve number. Every workflow Sarah owned gets re-derived by the next person, usually badly the first time, and the deltas only surface in a finance review a quarter later.

The four numbers above look like four different problems. They aren’t. They’re four symptoms of the same root cause: your business runs through an individual’s head, not through a system the next person can read.

Watch the chain.

1. → 42 %

   Lives in one head

   Because there is no system designed to capture it.

2. → 20 h/w

   Spent searching

   Because what isn’t in the system has to be found by asking. Asking takes longer than reading.

3. → 58 %

   Re-solving

   Because finishing the search and capturing the answer is one more step nobody’s job description includes.

4. → 5–12 mo

   Recovery

   Because the one person who knew is gone, and the system underneath was the wrong shape to onboard a successor.

Fix the system. The other three numbers come down with it.

The $32K–$130K range above is the operational cost. The bigger number lives elsewhere.

When a mid-market business goes to market — sale, recap, even a serious financing — the diligence team looks at key-person concentration as a first-pass filter. They want to know what happens if your top three people leave. They are going to model the answer to that question regardless of what you tell them.

Two equally well-run businesses with the same revenue, the same margin, and the same customer base will trade at different multiples depending on how the operating knowledge is held. The one whose institutional context lives in systems, runbooks, and crawlable documentation gets the deal. The one whose context lives in three people’s heads gets a discount, or an earn-out, or a structured exit that ties the founders in for three more years.

The two doors of a deal are price and structure. Key-person risk doesn’t close the price door — it closes the structure one.

We don’t have a tidy industry number for this discount. The working assumption among the M&A advisors we’ve compared notes with is roughly a 5–15% key-person diligence haircut, but treat that as folk wisdom until someone publishes it. Every operator we’ve worked with who came out the other side of a transaction tells the same story: the buyer noticed.

The category of work we do at Sauce is named after this problem: knowledge architecture. Three things change when you put it in.

01 · The knowledge stops being a person.

It becomes a place — usually a documented workflow inside a system you already pay for, plus a layer of structured runbooks the next hire can read on day one. Sarah is still Sarah. But the parts of her job that can’t walk out the door stop being able to.

02 · The search time collapses.

When the institutional context is queryable — whether by a person, a teammate, or an AI agent — the 20 hours a week of search becomes 2 to 4 hours of read. Coveo’s data and the IDC follow-ups converge on this: it’s the architecture, not the model, that determines the answer.

03 · The exit story changes.

When a buyer’s diligence team asks what happens if your top three people leave, the answer becomes here’s the operating manual, here’s the access map, here’s the runbook for the work that used to live with them. The discount disappears.

At Sauce we call this the cure for the 42% problem. The page where we lay out the full research and methodology is here. The pillar of our practice that addresses it most directly is Data Architecture. Most engagements pair it with Software Integration — because the knowledge has to live somewhere your team already opens.

1. [ 1 ]

   Panopto / YouGov · Workplace Knowledge & Productivity Report

   Survey of 1,000+ U.S. knowledge workers on institutional-knowledge capture and loss. Source of the 42% “lives in one head” figure.

2. [ 2 ]

   Coveo · Relevance Report — Workplace edition

   Knowledge workers spend an average of 3.6 hours per day searching for information — roughly 18–20 hours per work week. Comparable figures published by IDC over the prior decade.

3. [ 3 ]

   Panopto Research · Workplace Knowledge & Productivity Report

   Re-solving previously solved problems as a share of total knowledge-worker time. Source of the 58% figure.

4. [ 4 ]

   SHRM · Society for Human Resource Management

   Replacement-cost ranges for salaried key personnel: 50–200% of annual salary, with 5–12 month recovery windows for institutional context.

5. [ 5 ]

   U.S. Bureau of Labor Statistics · Occupational Employment & Wage Statistics

   Median salary for Operations Manager (SOC 11-1021) used as the modeling anchor.