Most companies have one person who gets asked whenever something goes wrong. She knows why a machine has to run slower with a particular material, which customer needs duplicate delivery notes and where the inspection records from the years before the system changeover are kept. None of this is written down anywhere. It sits in one head, and that head has a leaving date: retirement, a move to another employer, a long illness. Anyone who only starts securing that knowledge during the notice period mainly secures passwords and phone numbers - the rest walks out of the door. This article describes how to capture head knowledge beforehand, write it down in a usable form and put it to the test: as a schedule with fixed dates that runs alongside day-to-day business instead of starting on the last working day.
Key takeaways
- The date is known, yet the handover starts late. A retirement date is fixed years in advance, but in practice the capture of knowledge starts in the final weeks - leaving time for access rights and phone lists, but not for experience.
- Not everything belongs in a document. What matters are workflows that occur rarely, carry high follow-up costs and are handled confidently by only one person today. Daily goods receipt needs no description, an export declaration does.
- A knowledge map with four columns per task is enough to start: frequency, consequences of failure, number of people who can do it and where the paperwork lives. The order of work follows from that on its own.
- The test counts, not the document. Only once the stand-in has completed the workflow without asking, while the experienced colleague is still on site, has the company genuinely taken the knowledge over.
- Three metrics are enough to steer by: coverage per critical workflow, the age of the most recent description and the number of follow-up questions in the first weeks after a handover.
What a company loses when experience walks out
Company knowledge falls into two parts. One part sits in systems: master data, documents, bills of material, prices, deadlines. The other part sits nowhere. It consists of exceptions, back stories and rules of thumb - the knowledge that one particular supplier does not keep his promises at month end, that an inspection rule has been applied more strictly since an incident eight years ago, and that one customer returns any invoice without a purchase order number. In a process analysis this part shows up because the description of a workflow becomes vague at exactly one point and the answer is: that is how our colleague has done it for twenty years. That is where the value sits, and that is where the risk sits.
The trigger is rarely a surprise. By the mid-2030s the number of people of retirement age from 67 will rise from 16.7 million (Federal Statistical Office) to at least 20.1 million (Federal Statistical Office), while the number of people of working age between 20 and 66 will fall from today's 51.2 million by 3.2 million (Federal Statistical Office) even with high net immigration. Behind this stand the high-birth cohorts between 1955 and 1969. The act of leaving follows a well-documented pattern too: the average age at which an old-age pension was first drawn was 64.7 years (German Pension Insurance) in 2024, compared with 62.9 years in 2003. For a company this means the date is known years in advance - and is still treated as a staffing question rather than a question of whether the work can carry on.
Where the management itself is affected, the knowledge question becomes a question of survival. By the end of 2029 around 545,000 (KfW Research) of the 3.87 million mid-size companies in Germany intend to hand over the business, which is roughly 109,000 (KfW Research) successions a year. At the same time 569,000 (KfW Research) owners plan no continuation after their own withdrawal, which would amount to around 114,000 (KfW Research) business closures per year. The average age of owners is over 54 years (KfW Research), and more than 2 million (KfW Research) are already 55 or older. Whoever hands over a business hands over more than machines and customer lists. The technical side - accounts, rights, mailboxes - is covered in the article on account access when staff join and leave; this article is about what no access concept can capture.
Why it only becomes obvious in the final week
The knowledge map: four questions per task
The first step is a list, not a piece of software. Write down which tasks a department actually performs - not the job description, but the weekly and annual cycle. Then give every task four entries: How often does it occur? What happens if it fails or goes wrong? How many people in the company can carry it out on their own today? And where is the paperwork? The fourth column is the uncomfortable one, because the honest answer is often „on a drive nobody else opens" or „only in someone's head". Those are precisely the rows that belong in the coming months.
The map comes together faster when it builds on existing workflow descriptions. Anyone who has already drawn up a process map for mid-size companies simply adds two columns there. Anyone starting from scratch takes one department, and two hours of conversation plus a spreadsheet get them a long way. The boundary matters: the point is not to record every single movement, but to find the places where the company currently depends on one individual. Everything else is a later stage of process documentation.
Task Frequency Impact Able to do Paperwork
------------------------------+-------------+---------+------------+--------------------
Prepare annual accounts | once a year | high | 1 of 4 | scattered
Set up special-purpose machine| 4x per year | high | 1 of 6 | head knowledge only
Create export declaration | 2x per month| high | 1 of 3 | folder in the office
Pricing for custom parts | weekly | medium | 2 of 5 | sheet on a drive
Book goods receipt | daily | low | 5 of 5 | manual, up to date
Order of work: rows with high impact and exactly one capable person first.
The daily workflow with full coverage can wait.The order of work follows from the table on its own. At the top sit tasks with high impact and exactly one capable person, at the bottom the daily workflows everyone picks up anyway. This matters because the available time is limited: a handover runs alongside day-to-day business, not instead of it. Five to fifteen rows from the map are a realistic amount for six months. Anyone who instead has sixty workflows written up at once ends up with sixty half-finished texts and not a single tested stand-in.
What belongs in writing - and what does not
A workflow description is not a manual. It is the answer to the question of how somebody with basic knowledge gets this task finished when standing in. That includes the trigger, the documents and access required, the order of the steps, the exceptions together with their reasoning and the point at which someone else has to decide. The formal requirements of process documentation for audits help here, because they impose a structure that auditors expect anyway - and because both documents serve the same purpose: making it traceable how a case actually runs in the building.
What does not belong in it matters just as much. Click-by-click instructions age with the next software release and only describe the obvious anyway. Recording the reasons behind decisions instead - why this customer needs individual approval, why that item is shown separately - produces something that lasts for years. Complete data lists in the text are equally dispensable: they belong in the system they come from, and the description merely names where to find them. Where files and documents are still scattered, getting started with document management is the more effective first step.
The effort pays off because replacing people from outside is slow and uncertain. Between July 2024 and June 2025, on an annual average, 392,773 (KOFA) of roughly 1.1 million vacancies for qualified staff could not be filled in arithmetical terms, and 71.7 percent (KOFA) of those unfillable vacancies fell to small and mid-size companies. The Federal Employment Agency classifies 163 (Federal Employment Agency) of around 1,200 assessed occupations as shortage occupations and counted roughly 439,000 (Federal Employment Agency) reported jobs subject to social insurance for skilled workers, specialists and experts on an annual average in 2024. Covering a task with one person today effectively means planning for a replacement search that takes months.
Exceptions rather than the standard case
The standard case largely explains itself. What gets described is what deviates: special conditions, customers with their own inspection routine, cases with a statutory deadline. Those are the cases that cost a stand-in the most time.
Reasons rather than click paths
Why a decision goes one way and not the other survives every software change. The route through the input mask ages with the next release and is better placed in the system itself.
Location rather than content
Lists, prices and master data belong in the leading system. The description points to them instead of copying them - a copy inside the text ages unnoticed and gets used regardless.
The handover gets fixed dates
The most common reason for a failed handover is not unwillingness but capacity. Asked about the obstacles to digitalisation, 74 percent (Bitkom) of companies name the shortage of skilled workers and 60 percent (Bitkom) a lack of time. A handover run as a permanent side task loses out to every customer enquiry. It therefore needs the same treatment as a project: a start, fixed dates, named people and an end. Two hours a week over six months is manageable for most companies - sixty hours in one block during the final month is not.
The schedule takes the leaving date as its anchor and works backwards. Six months beforehand the knowledge map is drawn up, then the writing starts, from the third month the stand-in works alongside, and in the final weeks she runs the workflows on her own while the experienced colleague is still available. Where approvals are involved, who signs in future is settled at the same time; how to model that cleanly is covered in the article on digitising approval workflows.
Step 1: Draw up the knowledge map
Two hours per department. Tasks, frequency, impact, number of capable people, location of the paperwork. The result is an order of work, not a documentation set. That order is agreed with management so that nobody questions the selection later on.
Step 2: Write up workflows, one per week
The experienced colleague talks, the stand-in writes. That direction matters: whoever listens and takes notes notices immediately where an explanation assumes knowledge that is still missing. The text stays short and names exceptions together with their reasoning.
Step 3: The stand-in works alongside
From now on the stand-in runs the workflow while the experienced colleague sits next to her. Every question is noted down and goes into the description as an addition. The number of questions per run is the first metric you can rely on.
Step 4: A trial run without that person
For two weeks the stand-in runs the workflows alone, with no option to ask. Whatever goes wrong is the honest remainder of head knowledge. This trial takes place while the experienced colleague is still in the building - afterwards it is no longer a trial.
Step 5: Agree a window for questions
For the first weeks after the departure, an available contact and a framework are agreed, in writing and with payment where it goes beyond collegiality. Every question during that window points to a gap in the description.
Two hours per department. Tasks, frequency, impact, number of capable people, location of the paperwork. The result is an order of work, not a documentation set. That order is agreed with management so that nobody questions the selection later on.
The experienced colleague talks, the stand-in writes. That direction matters: whoever listens and takes notes notices immediately where an explanation assumes knowledge that is still missing. The text stays short and names exceptions together with their reasoning.
From now on the stand-in runs the workflow while the experienced colleague sits next to her. Every question is noted down and goes into the description as an addition. The number of questions per run is the first metric you can rely on.
For two weeks the stand-in runs the workflows alone, with no option to ask. Whatever goes wrong is the honest remainder of head knowledge. This trial takes place while the experienced colleague is still in the building - afterwards it is no longer a trial.
For the first weeks after the departure, an available contact and a framework are agreed, in writing and with payment where it goes beyond collegiality. Every question during that window points to a gap in the description.
Every step needs a named stand-in, and early. Deciding only in the final month who takes over throws away the months in which somebody could have worked alongside. It is normal for those involved to receive this extra task in different ways, which is covered in the article on bringing staff along with new workflows. It helps not to present the capture of knowledge as a form of control, but as what it is: relief for everyone who would otherwise remain the single point of contact. Accompanying training and rollout takes further heat out of the subject.
Standing in is the real test
A written workflow proves nothing. Only a run by a different person shows whether the description holds. That is why holiday cover is the best test instrument available: it is scheduled anyway, it costs no additional time, and it produces real cases instead of constructed examples. Anyone who deliberately schedules cover for the critical workflows from the knowledge map tests every row at least once over the course of a year - without a single separate project. The condition is that questions are not answered by phoning the holiday destination, because that is exactly what makes the test worthless.
| Aspect | Handover with a trial run | Handover via a folder |
|---|---|---|
| Evidence the knowledge arrived | The stand-in has completed the workflow | A document exists, unread |
| When gaps become visible | During the trial, with someone to ask | After the departure, with nobody to ask |
| Effort for the experienced colleague | Spread over months, two hours a week | Bundled into the final weeks |
| How current the description stays | Updated with every trial run | Frozen at the state of the leaving date |
| Robustness in case of illness | The stand-in is already trained | Training starts when it is urgent |
| Prerequisite | A schedule with dates and a named stand-in | Only a place to store files |
The same idea carries beyond retirement. Every task that depends on one person today is equally exposed to illness, parental leave or a short-notice move. It is therefore worth maintaining the map beyond the immediate occasion and updating it once a year. Recurring customer cases are particularly suitable for this because they occur often and are clearly bounded - the workflow described in the article on handling complaints digitally is a typical example of a task that quietly depends on a single person in many companies.
Where deadlines apply, cover is not optional
Three metrics that show where the knowledge stands
A handover project without measurement ends up as a matter of feeling. Three figures are enough. First, coverage: how many people can carry out each workflow marked as critical on their own? Second, the age of the description: when was it last touched during a real run? Third, the number of follow-up questions in the first four weeks after a handover. All three can be maintained from information you already have and need no separate software; the evaluation fits the same approach as other metrics and reporting in the company.
The cut matters. Coverage measured across all tasks says little, because the daily workflows pull it upwards. It only becomes meaningful across the rows of the knowledge map marked as critical. The same applies to the age of the description: what counts is not the date the file was last changed, but the last real run by a stand-in. And with the questions it is not the number alone that is interesting but their distribution - if they cluster on a few workflows, the gap is located and can be closed quickly.
- Coverage per critical workflow: the share of tasks marked as critical that at least two people can carry out on their own. A target value and a deadline are set once and then reviewed each quarter.
- Age of the most recent description: the time since the last run by a stand-in, not since the file was last changed. A text nobody has applied for two years is an assumption.
- Follow-up questions after a handover: the number and distribution of questions in the first four weeks. Each one is taken into the description as an addition, otherwise it comes back at the next change.
- Number of tasks with no named stand-in: the most uncomfortable of the four figures, because it cannot be reduced without a staffing decision. It belongs on the same sheet regardless.
Which lever changes what
Passing on knowledge is a form of training in practice, and training does take place - just rarely in a structured way. 73 percent (Bitkom) of companies train staff on digital topics, but at 62 percent (Bitkom) such offerings exist only occasionally and at merely 11 percent (Bitkom) are they open to all or almost all employees. Only 50 percent (Bitkom) pursue any strategy at all in this area. For a handover this means the framework is usually there, it is simply not aimed at the workflows the company depends on. The knowledge map supplies exactly that target list.
The cost side can be estimated roughly as well. In 2025, one hour worked cost companies in manufacturing and services an average of 45.00 euros (Federal Statistical Office). Two hours of capture per critical workflow, fifteen workflows, plus the same amount of time again for working alongside and the trial run: that is around sixty hours or roughly 2,700 euros of working time, spread over six months. Set against that is a workflow standing still because nobody can run it after a departure. The calculation works out for every workflow whose failure costs more than two days.
Fifteen workflows are a realistic amount
What a company can prepare on its own
The first steps need no outside support. A spreadsheet, two hours per department and the willingness to write down uncomfortable answers are enough for the map. Only once that turns into a schedule spanning several departments is a structured survey worthwhile; how such an approach runs is described in the article on how a process analysis works. What matters is not to start with the tooling question. A place to store descriptions is quickly found; deciding which fifteen workflows come first is the actual work.
- Collect the departure dates for the next three years - from the age structure, announced moves and ongoing succession considerations. Without that list, any prioritisation stays a gut feeling.
- Draw up a knowledge map with four columns per department and mark the rows with exactly one capable person.
- Select fifteen of the marked rows and set up a schedule working backwards from the earliest departure date.
- Name a stand-in per workflow before the writing starts - the stand-in takes the notes, not the departing person alone.
- Set up a fixed weekly slot of two hours and treat it like a customer appointment.
- Schedule a trial run after every workflow written up and count the questions instead of ticking the document off.
The most expensive knowledge is not the complicated kind but the rare kind. A workflow that occurs once a year is not picked up in passing by anyone - and it fails at precisely the moment it is needed again.
One side effect is worth noting: companies that have written up their critical workflows find every later changeover easier. Anyone planning to migrate data from a legacy system needs the answer to which fields serve which business purpose anyway - and by then it is already written down. The same applies to replacing paper filing and digitising documents, or to the question of why a case takes so long, which the article on shortening lead times addresses. The handover is therefore not only insurance against a departure, but the groundwork for every change that comes after it.
Sources and Studies
Related Articles
Handling complaints digitally: deadlines and evidence
Which periods run from delivery, which records should be created on a complaint case, and how to map both digitally without turning it into a large project.
Choosing Business Software: Requirements Come First
How to decide before you buy: measure the volume baseline, write a lean requirements document in a week, score vendors by weight and check the contract terms.
Reading processes from system data: where it snags
How to reconstruct the actual workflow - loops and special paths included - from the timestamps your systems already record, instead of guessing or estimating.