How to Offboard a Managed Service Provider
Ninety days is a realistic plan to offboard a managed service provider. Compress it and the revocation step is what gets cut, leaving agents installed and credentials live for months.

In a study of 133 popular open-source projects, 34 percent had a bus factor of one and 65 percent had a bus factor of two or fewer. How to count your business's bus factor function by function, and raise it with runbooks and cross-training instead of new hires.
In a study of 133 popular open-source projects, 34 percent had a bus factor of one, meaning a single person held enough of the knowledge that the project would stall if they walked away, and 65 percent had a bus factor of two or fewer (Avelino, Passos, Hora and Valente, 2016). The bus factor is the minimum number of people who would have to disappear before work grinds to a halt. Most small businesses have never counted theirs, and the number is almost always lower than they hope.
The name comes from a blunt thought experiment: what happens if the one person who understands the payroll export, the DNS records, or the client history gets hit by a bus? The morbid framing is the point. It forces a question owners avoid because the answer is uncomfortable. I study behavioral psychology because that avoidance is predictable, and because naming the risk calmly is the first step to reducing it.
The bus factor, also called the truck factor or lottery factor, is defined as the minimum number of team members who have to suddenly leave before a project stalls for lack of anyone who knows how it works (Wikipedia: Bus factor). The term traces to 1994, when a Python contributor asked what would happen to the language "if Guido was hit by a bus."
A bus factor of one is the danger zone. It means there is exactly one person who can do a critical thing, and no written trail, no second set of hands, and no way to recover the knowledge quickly if they are out sick, on vacation, or gone for good.
What makes this measure useful is that it is specific. Instead of a vague worry about "being too dependent on Sarah," it gives you a countable number, per function, that you can raise deliberately. Payroll might be a one. Deployments might be a one. Knowing which client prefers which invoice format might live entirely in one inbox.
The two terms overlap, and people use them interchangeably, but they come from different fields. Key person risk is the older business and insurance idea: the financial or managerial damage if a founder, a top salesperson or an executive is suddenly gone. The software term is narrower. It concerns the technical experts who are both key and irreplaceable, the people whose knowledge of how something works exists nowhere else (Wikipedia: Bus factor).
For a small business the distinction is useful because it splits one vague fear into two lists. Key person risk asks who brings in the revenue, holds the relationships and signs the checks. The software version asks who knows how the systems work: the DNS settings, the backup schedule, the payroll export, the order in which month-end steps have to happen. An owner can be both lists at once, which is common in firms under ten people. A key person can often be covered by a partner or by key person insurance. A technical process that only one person understands usually can't be covered by money, because the missing thing is knowledge.
Both are a form of single point of failure. The software version is simply the one you can count function by function and raise with documentation and cross-training, which is why the rest of this article focuses on it.
Concentration of knowledge is the natural result of getting work done fast rather than a failure of discipline. The person who built the system understands it best, so the next related task goes to them too, and the gap between what they know and what everyone else knows widens with every efficient decision.
The research bears this out. The same open-source analysis found that fewer than 10 percent of systems had a bus factor above ten, so high concentration is the norm even in projects with many contributors (Wikipedia: Bus factor). If well-resourced software teams drift toward single points of failure, a two-person or five-person business will drift there faster.
There is a psychological pull underneath it. Being the only one who can do something feels like security. It looks like job protection and like value. The uncomfortable truth is that indispensability is fragility that happens to look flattering, and it is fragile for the owner as much as for the specialist. This connects to a pattern I have written about in what I stopped outsourcing and what I will never build in-house: the goal is to place capability where it survives instead of hoarding it.
The figures in the opening come from a 2016 paper by Guilherme Avelino, Leonardo Passos, Andre Hora and Marco Tulio Valente, who estimated the truck factor of 133 popular GitHub projects from their commit histories. The paper reports that "45 systems (34%) have TF = 1" and that 42 systems (31%) have a truck factor of two (Avelino et al., arXiv 1604.06766). Together that is the 65 percent with two or fewer. The highest value belonged to the Linux kernel, with a truck factor of 57.
Two details from the paper matter for a business owner. First, the authors asked the developers whether their projects would be in trouble if the named people left. Developers of 24 systems (39%) agreed outright, and some of the answers are worth reading: "If both of us left, the project would be kind of unmaintained." Second, the paper describes a project that had already lost its only truck factor author. One GitHub employee accounted for 78 percent of the source files of a mobile app, and when GitHub stopped maintaining it, the repository went almost inactive.
These are open-source software projects, not accounting firms or marketing agencies, so the percentages don't transfer directly to your business. What transfers is the shape. Even in public projects with dozens or hundreds of contributors, most of the working knowledge settled on one or two people, and nobody planned it that way. Concentration is what happens by default when work flows to whoever is fastest.
Most owners picture bus-factor risk as a sudden crisis: someone quits, and the wheels come off. The more common cost is quieter and continuous. When knowledge lives in one head, everyone else spends time waiting for it or recreating it.
Panopto's Workplace Knowledge and Productivity Report, a study of more than 1,000 workers, found that 42 percent of the knowledge needed to do a job is unique to the individual who holds it, not shared with anyone else (Panopto, 2018). The same research found workers lose an average of 5.3 hours a week waiting on information from a colleague or reconstructing something that already exists but was never written down.
Gartner reached a similar place from a different angle, reporting that 47 percent of digital workers struggle to find the information they need to do their jobs (Gartner, 2023). A bus factor of one works less like a rare catastrophe you insure against and more like a tax you pay every ordinary week, and it compounds.
The mechanics are simple to draw. On one path, a task goes to one person, that person becomes the silo, and the silo becomes a single point of failure. On the other path, the same task is documented and shared, so the knowledge lives in the system rather than in one person, and the loss of any individual is survivable.
The distinction matters because the fix is not "hire more people." Adding headcount to a business that documents nothing simply creates more silos. The lever is where the knowledge lives, not how many people you employ. We run Kief Studio as a two-person studio, and the way we protect against key-person risk is described in the two-person studio model: systems carry the load that a larger org would spread across staff.
This is also why data governance is the real prerequisite for AI. You cannot automate or delegate a process that only exists as intuition in one person's head. Distribution has to happen at the level of documented, structured knowledge before any tool can help.
The single highest-return move for raising a bus factor is writing things down, and it is undervalued precisely because it feels like overhead. The cost of documentation is a few hours now. The cost of its absence is those 5.3 hours a week, spread across everyone, forever, plus the tail risk of a departure you cannot recover from.
Good documentation is a runbook for the three things that would hurt most if the person who does them were unavailable (how money moves, how the site stays up, and how a client is served), rather than a binder no one reads. Start there, not with an exhaustive wiki. The point of institutional memory that survives turnover is that the memory lives outside any single skull.
Documentation also decays, which is why writing it once is not enough. This is the same discipline behind the idea that automation maintenance is the job, not the launch: a runbook that describes last year's process is a trap, not a safeguard. A documentation-first habit, close to what async-first working demands, keeps the written record current because the written record is how work actually gets coordinated.
At Kief Studio we run managed technology services as a documented service, so the knowledge of how a client's system runs never sits with one person.
Documentation raises the floor and cross-training raises the ceiling. A written runbook lets a second person recover a process under pressure; a second person who has actually run it can do so calmly. The two together move a function from a bus factor of one to a bus factor of two, which is the difference between fragile and merely thin.
You do not need to duplicate every role to do this. The usual advice for raising a bus factor fits any size: reduce complexity, document processes and keep that documentation current, and encourage cross-training (Wikipedia: Bus factor). In a small business, cross-training can be as light as pairing on a task occasionally, reviewing each other's work, and rotating who owns a recurring job. A quarterly "you run it this month" rotation on invoicing or deploys spreads knowledge without adding a salary.
Managed operations are one way small businesses buy a higher bus factor without hiring, which is much of what LTFI exists to provide: the person who knows how the system runs is not your single point of failure when the system is run as a documented service. The same logic informs how we think about building systems before you need them, because the calm time to raise a bus factor is long before anyone is missing.
None of this requires fear, and none of it requires stopping to build bureaucracy. It requires one honest inventory: list the handful of things that would genuinely stall if a specific person were unreachable for two weeks. The list is almost always shorter than owners expect, which is the reassuring part.
For each item, pick the cheaper of two moves: write the runbook, or teach one other person. Neither takes long, and both convert a hidden liability into a boring, documented, survivable process. Secure and resilient operations are a byproduct of building this way, not a separate project you bolt on later.
The bus factor is worth measuring because it turns an anxiety you cannot act on into a number you can raise on purpose. You are not trying to make anyone replaceable. You are trying to make sure that the business, and the person, are both free to take a real vacation.
You don't need software or a consultant to measure this. You need a list, a table and an honest hour with the people who do the work. Here is the version I'd use for a business of two to twenty people.
Write down every process that would stop, or quietly break, if its usual owner were unreachable for two weeks. Most small businesses land somewhere between eight and fifteen items. Typical entries:
For each item, count the people who could do it unaided tomorrow, with the access they already have. "Could figure it out with the old owner on the phone" doesn't count. A function with a written runbook and nobody who has ever followed it gets a half point at most. Anything scoring one goes to the top of the list.
For each function that depends on one person, choose either a runbook or a second trained person, and give it a date. Three or four fixes a quarter is a realistic pace. The Qurtoo playbook for turning one person's know-how into an SOP is a practical way to get the first draft out of someone's head: they talk through the task, the transcript becomes a draft, and someone who has never done the task tests it.
Illustrative example: take a hypothetical six-person bookkeeping firm in Worcester. The audit turns up eleven functions. Seven are covered by two or more people because the staff already cover each other's clients. Four depend on one person, and all four sit with the office manager: the domain registrar login, the backup checks, the payroll provider's admin account and the process for adding a new hire to email. None of them takes more than an hour a month, and small, infrequent tasks like these are the ones nobody else ever learns. Writing four short runbooks and adding the owner as a second admin on each account takes about a day, and it moves the firm's weakest point from one person to two.
Repeat the audit once a year, and again whenever someone joins, leaves or changes roles. Coverage drifts back toward one person on its own.
A runbook doesn't help if the only person with the password is the one who is gone. Many of these problems in small businesses are access problems: one person is the sole admin on the domain registrar, the bank portal, the cloud console or the company's social accounts, and the recovery email is their personal address.
The fix is boring and quick. Put every shared credential in a business password manager with a shared vault, add a second administrator to every critical account, and make sure recovery emails and phone numbers belong to the business rather than an individual. I've written about why a password manager is the best security investment a small business can make, and key-person coverage is one more reason.
In Massachusetts this also lines up with the law. Under 201 CMR 17.00, any business that holds personal information about Massachusetts residents must keep a written information security program, and the first required element is "Designating one or more employees to maintain the comprehensive information security program" (201 CMR 17.03(2)(a)). The same section requires "Preventing terminated employees from accessing records containing personal information." A security program that lives in one person's head, with access nobody else can revoke, is hard to square with either line. The official regulation page on Mass.gov has the full text, and my guide to Massachusetts WISP requirements under 201 CMR 17 walks through the rest.
Departures are where access problems surface. Brian's employee offboarding checklist for IT access lays out which logins to shut off and in what order, and it doubles as a list of the accounts that should never have had a single admin in the first place.
Outsourcing a function moves the risk to the vendor's side rather than removing it. If one freelancer built your website and is the only person who knows how it is hosted, or a small IT provider has one technician who understands your network, that function still depends on one person. You just can't see it from inside the business.
Ask any provider who supports a critical system a few direct questions:
The last question belongs in the contract itself. Exit terms that require documentation, credentials and a transition period are what keep a vendor's single point of failure from becoming yours. I cover the specific clauses in vendor contract exit terms, and Kief Studio's knowledge base explains how vendor lock-in builds up when accounts and documentation sit with the provider. Massachusetts businesses holding personal information also have a duty to oversee service providers under the same section of 201 CMR 17.03.
Keeping records in one shared place helps here too. When customer history, project notes and site details live in a single system that more than one person can open, as with an operating system for small teams that keeps records in one place, a departure or a vendor change costs far less to recover from.
The bus factor is the smallest number of people who would have to leave, all at once, before a project or a business function stalls because no one left knows how it works. A bus factor of one means a single person holds knowledge that exists nowhere else. Higher is safer, because the knowledge is spread across more people and, ideally, into written documentation.
Go function by function rather than trying to score the whole company at once. For each critical process (payroll, deployments, client onboarding, the website staying up), ask how many people could carry it out unaided if the usual person vanished for two weeks. If the answer is one, that function has a bus factor of one. The list of ones is your priority order.
It is, though the cost is usually quiet rather than dramatic. Research found that 42 percent of job-critical knowledge is unique to one person and that workers lose about 5.3 hours a week waiting on or recreating information (Panopto, 2018). The everyday drag of concentrated knowledge often costs more over a year than any single departure.
The cheapest step is documentation, done narrowly. Writing a short runbook for the two or three processes that would hurt most if their owner were unavailable costs a few hours and immediately makes recovery possible for someone else. Cross-training a second person raises it further, but a written record is the lowest-cost first step.
No. Cross-training is about moving knowledge, not adding headcount. Occasional pairing, reviewing each other's work, and rotating who owns a recurring task all spread knowledge across existing staff. Managed operations can also raise your effective bus factor by running documented systems as a service rather than leaving them dependent on one internal person.
A bus factor of 1 means one person is the only one who knows how to do something critical, so the function stops if that person is unavailable. It is the definition of a single point of failure. In the 133-project study, 34 percent of the projects were in exactly that position (Avelino et al., 2016).
A low bus factor means knowledge and access are concentrated in very few people, usually one or two. The business works fine day to day, but vacations, illness and resignations cause delays, and the people who hold the knowledge become bottlenecks that everyone else waits on.
"Hit by a bus" is the darkly humorous shorthand behind the term. It stands for any sudden departure, including a new job, parental leave or a change of career, not only an accident (Wikipedia: Bus factor). The syndrome is the habit of letting critical knowledge sit with one person and only noticing when they are gone.
For most small-business functions, a bus factor of 2 is a reasonable target. It means one person can be away without the work stopping. For the handful of things that would do real damage if they failed, such as payroll, backups and the domain name, pair the second person with a written runbook so a third person could recover the process if both were out.
In software engineering, the bus factor (often called the truck factor) is the number of developers who would have to leave before a codebase can't be maintained. Researchers estimate it from commit histories by counting how much of the code each developer authored, as the 2016 study of 133 GitHub projects did.
Yes. Lottery factor, truck factor and circus factor are all synonyms for the bus factor (Wikipedia: Bus factor). Lottery factor is the gentler version: it asks what happens if your key person wins the lottery and quits.
The phrase is usually traced to 1994, when Michael McLay asked publicly what would happen to the Python language if its creator, Guido van Rossum, were hit by a bus. "Truck number" appeared in the 2004 book Organizational Patterns of Agile Software Development, and the idea spread through engineering and open-source communities from there.
Ninety days is a realistic plan to offboard a managed service provider. Compress it and the revocation step is what gets cut, leaving agents installed and credentials live for months.
A single disruption to lean, just-in-time production cost the global auto industry an estimated $210 billion in lost revenue in 2021, per consulting firm AlixPartners. An antifragile business reads a shock like that differently: not as a catastrophe to survive, but as information to...
Only 20 percent of executives say their organizations excel at decision making, and most report that much of the time they spend on it is used ineffectively. A large share of that waste comes from treating reversible decisions as if they were permanent. Amazon has a name for the...
Work With Us
Kief Studio builds, protects, automates, and supports full-stack systems for businesses up to $50M ARR.
Newsletter
Strategy, psychology, AI adoption, and the patterns that actually compound. No spam, easy to leave.
Subscribe