How to Document Business Technology Infrastructure
Here is a test you can run without telling anyone. Imagine you wake up tomorrow unable to answer a phone for two weeks. Who renews the domain name? Who can reset the password on the email admin account? Where is the router's login, and which of the three similar looking boxes in the closet is the router?
If the honest answer to any of those is "only me", you have a bus factor of one. The bus factor is a blunt measure of how many people could be suddenly unavailable before a project or a business stalls. It is not a comment on anyone's health. People take vacations, change jobs, have babies, get sick, and in Puerto Rico they sometimes lose power and connectivity for days after a storm. A business that runs on one person's memory is borrowing against luck.
Documentation is the repayment plan. It does not need to be beautiful. It needs to be findable, accurate enough to act on, and safe.
Decide who the document is for
The most common reason documentation fails is that it was written for no one in particular. Pick your readers and write for them.
- A capable stranger. A contractor or new hire who is technical but has never seen your setup and has to restore service at two in the morning.
- A non-technical owner. Someone who needs to know what exists, what it costs, and who to call, but will not touch a console.
- Your future self. You, eighteen months from now, who has forgotten why that odd rule was added.
Write the main runbook for the first reader and a one page summary for the second. The third reader benefits from both, as long as you record the reasons for decisions and not just the settings.
A runbook structure that holds up under stress
A runbook is a set of written procedures for operating and recovering a system. People reading it will often be tired and stressed, so structure matters more than prose.
I suggest these sections, in this order.
- Start here. One page. What the business depends on, who is responsible, how to reach them, and where everything else lives.
- Inventory. Every system and service, as a table.
- Access and recovery. How to get into each critical system, and how to get back in when the usual route fails.
- Procedures. Step by step instructions for the things that actually happen.
- Incident playbooks. What to do when specific things break.
- Contacts and contracts. Vendors, support numbers, account numbers, renewal dates.
- Change log. A dated list of what changed and why.
Keep each procedure short enough to follow with one hand while holding a phone in the other. Number the steps. Name the exact menu or command. State what success looks like at the end, so the reader knows they are done.
Build the inventory first, because everything else hangs on it
You cannot document what you have not listed. Start with a table. One row per system, and be honest about the ones that are only half known.
| Item | What it does | Owner | Where the admin login lives | Renewal or expiry | Depends on | Notes |
|---|---|---|---|---|---|---|
| Domain registration | Controls the company web and email addresses | Owner | Password manager, shared vault | March each year | Registrar account, billing card | Auto renew is on, card expires in June |
| DNS host | Directs traffic for web and email | IT lead | Password manager | None | Domain registration | Records exported monthly |
| Email platform | Staff mail and calendars | IT lead | Password manager and break glass account | Monthly subscription | Domain, DNS | Rules documented separately |
| Accounting system | Books and invoicing | Finance lead | Finance vault | Annual | Bank connections | Three users plus accountant |
| Office router and firewall | Connects the office to the internet | IT lead | Password manager | Warranty to a set date | Internet line, power | Config backup saved weekly |
Then add the less obvious things that cause outages.
- Domain names and who the registered owner of each is. Many small businesses discover that a former web developer or relative is the registered owner.
- Security certificates and their expiry dates.
- Payment cards attached to subscriptions, because an expired card is a common cause of a sudden cancellation.
- Phone numbers, mobile accounts and who holds the account.
- Hardware serial numbers, locations and warranty dates.
- Each integration and automation that moves data between systems, with its owner and the credential it uses.
A useful discipline is to add a "Depends on" column and fill it in honestly. The inventory then becomes a map of single points of failure. If the domain, the email and the website all depend on one registrar login that only you can reach, that is a finding.
Access recovery is the part everyone skips
Many outages are not technical failures. They are lockouts. Two factor authentication, which asks for a code from a phone in addition to a password, is the right protection, and it also means that losing the phone can lock you out of everything.
For each critical system, write down how access would be restored if the usual person were gone.
- Shared vault. A business password manager with a shared vault for critical credentials, so that access does not depend on one person's memory or notebook.
- Break glass accounts. One or two emergency administrator accounts that are not tied to a person, with long unique passwords, stored offline in a sealed envelope or a safe, and used only in emergencies. Log and review any use.
- Recovery codes. Most services issue one time recovery codes when you enable two factor authentication. Store them in the vault, not in a screenshot on a phone.
- A second administrator. For each major platform, at least two people hold admin rights. A single administrator is a single point of failure by definition.
- Account ownership. Accounts should be registered to a role address such as an admin mailbox controlled by the company, and not to someone's personal email.
- Succession. Name who can authorize recovery if you are unavailable, whether that is the owner, an attorney, or a trusted managed provider, and describe how they would prove their identity to vendors.
Test it. Once a year, have the second person try to gain access using only the documentation. The first time, something will be missing. That is the point of the test.
Keep secrets out of documents
This is the rule that most often gets broken, and it is the easiest way to turn a helpful document into a liability. A runbook full of passwords is a single file that gives anyone who finds it the keys to the business. It also gets copied into email, chat and backups where you can no longer control it.
The rule is that documents say where a secret lives, never what it is.
Write "Admin password is in the shared vault, item named Email Admin" and not the password itself. The same applies to API keys, recovery codes, private keys, database connection strings with passwords in them, and screenshots that happen to show any of these.
Some practical habits.
- Store secrets in a password manager or a secrets manager, and link to the item name from the runbook.
- Check screenshots before pasting them into a document. Blur or crop anything sensitive.
- Keep the runbook itself in a place with access control and a history of changes, such as a shared drive with limited permissions or a wiki, and not in an email thread.
- Keep a printed copy of the Start Here page and the emergency contacts somewhere safe in the office. When the internet is down, a cloud document is no help, and in a hurricane season that is a real scenario.
- If a secret has ever been pasted into a document, treat it as exposed and change it.
Personal data deserves the same care. Document how a process works, not which customer's records went through it. Whether specific records carry legal retention or privacy duties depends on your industry and jurisdiction, and this is not legal advice.
Write procedures that people can actually follow
For each task that matters, write a procedure with the same layout every time. A good template has these parts.
- Purpose. One sentence.
- When to use it. The trigger.
- Who can do it. Role, not name.
- Before you start. What access or information you need, and where to find it.
- Steps. Numbered, concrete, with the exact names of buttons, menus or commands.
- How to know it worked. A check the reader can perform.
- If it fails. The first two things to try, and who to call.
- Last reviewed. A date, and who reviewed it.
Here is an illustrative example for a task that comes up in every small business, a new employee needing email and software access.
Purpose. Create the accounts a new hire needs on their first day. Who can do it. Any administrator. Steps. One, open the admin console and select Add user. Two, enter the name and use the standard address format. Three, assign the role template that matches the job. Four, add the user to the groups listed in the role sheet. Five, issue a temporary credential through the password manager's sharing feature. Six, require two factor setup at first sign in. How to know it worked. The new user can sign in from a different device and reaches only the apps on the role sheet. If it fails. Confirm the license count in the billing page, then check the group assignment.
Writing this took ten minutes, and it spares the next person an hour of guessing. Do the same for offboarding, where the order of steps matters, and for restoring a backup, which you should prove by actually doing it once.
Document the why, and the weird parts
Settings tell the next person what is true now. Decisions tell them what is safe to change.
Every few entries, record why something is the way it is. "The firewall blocks outbound mail on port 25 because the printer once relayed spam." "Backups run at 2 a.m. because the accounting system locks files during the day." "The second internet line is a cellular failover and is not suitable for video calls." One sentence is enough. These notes are the first thing a future reader will wish they had.
Also record the known weaknesses. A runbook that admits "the backup of the shared drive has not been restore tested" is more useful than one that implies everything is fine, because it tells the reader where the next incident will probably come from.
Keep it alive with a review rhythm
Documentation that is not maintained becomes dangerous, because it is trusted and wrong. The fix is a light, regular rhythm rather than a heroic annual project.
| Cadence | Task | Time |
|---|---|---|
| Every change | Update the inventory or procedure as part of finishing the work | A few minutes |
| Monthly | Review the renewal and expiry dates for the next ninety days | 15 minutes |
| Quarterly | Walk through one procedure and one incident playbook as if for real | 1 hour |
| Twice a year | Review access lists, remove former staff, confirm second administrators | 1 hour |
| Yearly | Run an access recovery drill with the second person, and test a backup restore | Half a day |
Make "update the doc" part of the definition of done for any change. Put a review date on every page. If a page has not been reviewed in a year, mark it as unverified in the title, so readers know to be careful.
Hurricane preparedness is a natural review point. Before the season, go through your Start Here page, check that printed copies exist, confirm who has authority to make decisions, and verify that the cloud services you depend on can be reached from a phone on mobile data.
A first week plan
You do not need to document everything. Start where the risk is highest.
- Day one. List every system and service you can think of, in a spreadsheet, with the owner and the renewal date.
- Day two. Find the registered owner of your domain, and fix it if it is the wrong person.
- Day three. Move any credentials from notebooks, chats and spreadsheets into a password manager, and set up a shared vault.
- Day four. Create the break glass account and store its recovery information offline.
- Day five. Write the Start Here page and one procedure, perhaps new user setup, and give them to someone else to try.
By Friday you will have lowered your bus factor more than most small businesses do in a year.
If you would rather not do this alone, Ozcorp InfraOps can help small teams build and maintain this kind of documentation as part of running their systems, and InfraOps Base starts at $500 per month with vendor and license charges billed separately. Whatever you decide, do the first step this week. Your future self, or whoever sits in your chair on a bad day, will be grateful.
Discuss your infrastructure requirements
Review the current service offer or schedule a conversation about the scope you need.
Stay in touch
Follow Ozcorp
Leave your details for occasional notes about new articles and Ozcorp updates.