Yesterday’s export shows twelve orders ready to go. Since then, one customer has changed their address and another has cancelled. The warehouse can open the file, but it still needs to know which instructions are current. Meanwhile, new orders are coming in and someone has to decide where to record them.
When a small team grows, the founder often stops seeing every change to an order. That makes recovery a team job. Someone needs to know what information is available, how much of it can be trusted and which commitments can still be made.
Find out what you can actually restore
On 31 January 2017, an engineer working to repair database replication at GitLab.com accidentally deleted data on the primary database server. The scheduled database backups had been failing and the emails meant to flag those failures had been rejected. GitLab recovered using a snapshot taken about six hours before the outage. Restoring service took over 18 hours. The company identified a gap in its own review: nobody had been responsible for regularly testing the backup procedure. GitLab’s incident postmortem
A useful recovery test answers two different questions. How long does it take to get the information back? How much work might have happened since the last usable copy? In the warehouse, that second answer tells you how far back to look for cancellations, address changes and orders already sent.
Agree what can continue
For the warehouse, the order owner might release a shipment once they can confirm the address, payment and dispatch status from reliable records. Orders with missing or conflicting information stay on hold. Name a backup who can make the same call if the order owner is away, and agree who tells customers about delays.
Give the team one place to record work done during the outage. Use the order number, the action, the time, the evidence used and the person who authorised it. When the system returns, assign someone to match those entries against the live records before normal dispatch resumes. An order that still says “ready” may already be on a van.
Give someone the job of testing it
Ask the person who would cover on a busy day to run the test. Can they access the records with their usual permissions? Can they tell how old the information is? Do they know who can approve an exception? If they have to call the founder to find the file or interpret it, you’ve found a dependency worth fixing.
In an Operations & Intelligence Scan, I would start with the workflow where a lost record or an incorrect commitment would cost the team most. In Stewardship, I would check the recovery routine again after a tool migration, a change of owner or a new handoff. Each can change what the team is able to recover.