A common pattern after a small business is hit: a mailbox gets taken over, a fake invoice goes out, someone notices, the password is reset and everyone goes back to work relieved. A few months later the same thing happens to a different person, through the same gap, because nobody ever worked out what the gap was.
Breach analysis is the step that gets skipped. It means looking back at an incident, or a near miss, and answering four plain questions:
- How did they get in?
- What did they touch, and for how long?
- Why didn’t anyone notice sooner?
- What do we change so it doesn’t work next time?
Large companies have security teams for this. A small office can do a useful version with its IT provider in an afternoon, as long as the evidence is still there.
Before anything else: keep the evidence
The first instinct after an attack is to wipe everything and start clean. Hold off for a moment, because the traces of how the attacker got in are exactly what a quick cleanup destroys.
- Disconnect, don’t switch off. Unplug an affected computer from the network or turn off its Wi-Fi, but leave it powered on. Some evidence only lives in memory.
- Don’t reimage yet. Wiping and reinstalling a computer is often the right end point, but only after someone has captured what’s on it.
- Call your IT provider and your cyber insurer early. Many policies want to hear about an incident quickly, and some expect you to use investigators they approve. Ring them before you hire anyone.
- Talk about it on a clean channel. If email might be compromised, use phones.
- Report it. The FBI takes reports of cybercrime and fraud at ic3.gov. If money was wired, call your bank at once: the sooner it knows, the better the chance of stopping the transfer.
Logs don’t last forever
In most small offices the best evidence lives in Microsoft 365, and it expires. Microsoft keeps audit records for 180 days on most business plans (Microsoft Learn). Sign-in logs in Microsoft Entra ID, the part of Microsoft 365 that handles accounts, last 30 days on the paid plans and only seven days on the free tier (Microsoft Learn).
If you suspect something, have the logs exported now, not after the meeting about it next week. And if your plan keeps sign-in logs for only a week, that’s worth fixing before you need them.
Building the timeline
The core of the analysis is a timeline: what happened, in order, with times. Everything else hangs off it. For a typical small-office incident the sources are:
- Sign-in history for the affected accounts: when, from where, from which device, and whether MFA was used.
- Mailbox changes: new inbox rules, forwarding, and apps the user approved.
- The audit log: files opened, downloaded or shared, and settings changed.
- Endpoint alerts from the antivirus or endpoint detection tool on each computer.
- Firewall and DNS filtering logs, which show where infected machines tried to connect.
- People. Ask the person who clicked what they saw and when. They’re the best witness you have, and they’ll only tell you if they aren’t afraid of being blamed.
Then put the first sign of the attacker next to the moment someone noticed. The gap between those two times is the most useful number in the whole review. Mandiant’s 2026 report put the worldwide median at 14 days between an intruder getting in and being found (M-Trends 2026). Whatever your gap was, the review should make it shorter.
Finding the real cause
“Someone clicked a phishing link” is where most reviews stop, and it’s rarely the cause. People click links. The useful question is why a click was enough. Keep asking why until you reach something you can change:
- Why did the click give them the account? The password was typed into a fake page.
- Why wasn’t MFA enough? It was a phone prompt, which a fake page can relay. Or MFA was off for that account, because of an old exception for a scanner or an app.
- Why did nobody notice for two weeks? Nothing was set to alert on new inbox rules, and sign-in alerts went to a mailbox nobody reads.
Each answer points to a fix: phishing-resistant sign-in for staff who handle money, closing the old exception, an alert on inbox rules, and alerts sent to someone who will act on them. Our post on Microsoft 365 account takeover goes through those settings.
Who you may have to tell
In Florida, the Florida Information Protection Act (section 501.171, Florida Statutes) applies when personal information about Florida residents is involved. In broad terms:
- Affected individuals generally have to be told no later than 30 days after you determine a breach happened, or have reason to believe it did.
- If 500 or more Floridians are affected, the Florida Department of Legal Affairs has to be told within the same 30 days.
- If more than 1,000 people are notified at once, the nationwide consumer reporting agencies have to be told as well.
- A vendor holding data for you has to tell you within 10 days of finding a breach.
Medical practices, card payments and some contracts bring their own rules on top. The analysis is what tells you whether personal information was reached at all, which is why it matters for this part. Notification is a legal decision, though, so take the facts to your attorney and let them decide what applies.
Turning findings into changes
A review that ends in a report nobody acts on is wasted. Finish with a short list, and give every item an owner and a date:
| Finding | Change | Owner | By |
|---|---|---|---|
| MFA off for the scanning account | Replace scan-to-email login with a supported method, turn MFA on | IT provider | Two weeks |
| No alert on new inbox rules | Alert on rules and outside forwarding | IT provider | One week |
| Bank change approved by email | Phone check on a known number for every bank change | Office manager | Now |
Those rows are examples of the kind of thing a review turns up, not a template to copy. Then check back in a month to confirm each change was made.
The gap between those two times is the most useful number in the whole review.
Do it for near misses too
The best time to run this review is when the attack failed. Someone reports a fake sign-in page they nearly used, or a filter catches a payment request from a lookalike domain. Nothing was lost, the logs are fresh and nobody is stressed, so you learn the same lessons for free. A quarterly 30-minute look at what was blocked and reported is a good habit.
For the attacker’s side of the same story, our post on the anatomy of a cyberattack walks through the seven stages an attack moves through and where each one can be stopped.
When it happens to one of our customers
For our Managed IT customers, this is part of how we handle an incident. We help you contain it, lock down the affected accounts and devices, work out what happened and restore normal operations from backups we’ve tested, then we sit down with you to agree what changes. The day-to-day protection that makes the next incident less likely is on our Cybersecurity page.
If you’re dealing with something now, call (386) 252-2292.