I spend a fair chunk of my week reading mail headers, and the thing that still gets me is how often one character decides whether money moves. Last month someone in finance sent an invoice to a client address that was one letter away from the real one. The domain looked fine at a glance, the display name matched exactly, and a reply came back within the hour asking us to update the bank details. Nobody noticed until the actual client called three days later asking where the invoice was.
I’ve noticed this is almost never a training failure. The person who sent it is careful, has done accounts payable for years, and keeps a checklist taped to the monitor. What tripped them up was an address copied out of an old thread, where someone had already replied to the wrong sender and the wrong sender had replied back politely.
What the near misses look like
Once you start looking, the pattern is everywhere. A supplier writes from a hyphenated domain and someone types the version without the hyphen. A logistics partner uses a regional subdomain and a forwarder strips it off. Swap one letter for a lookalike and you get an address that a tired human reads as correct while the mail server quietly delivers it somewhere else.
The ugly part is the follow-up. The fake reply usually arrives fast, sounds calm, and asks for something small. A corrected remittance, a new bank form, a copy of the signed page. It never demands anything dramatic, which is exactly why it slides past a reader who is skimming between two other tasks.
I keep a small folder of these. Half are invoices, a quarter are shipping documents, and the rest are contract drafts bouncing between two people who each think they are talking to the other side. Reading them in a row is a decent reminder of how little it takes. Two of the newest were less than a week old, which says something about how often this happens in an ordinary month.
Why this keeps happening
Most of it comes down to how the address entered the workflow. Someone forwards a thread, the next person hits reply, and the address is inherited without anyone reading it character by character. Autocomplete then cements the mistake. Type the first four letters of the domain and the wrong one now sits near the head of the list because it was used most recently.
There is also the display name problem. Mail clients show a name, not a domain, and on a phone the domain hides behind a tap. So the message says it is from the accounts team at a company you have worked with for six years, and technically it is not.
Then there is the honest one. The real address is ugly. Long, hyphenated, with a regional tag on the end. People shorten it because typing it correctly feels like admin work, and the shortened version resolves to a domain someone registered for exactly this reason. Nobody in that chain made an unreasonable decision. The sum of the decisions was still wrong.
Where consolidation helps, a little
I am skeptical of anyone who says a purchase fixes this. I’ve also seen unified platforms such as Ping64 used to cut down the overhead of juggling many separate security tools. The appeal there is mundane. Fewer consoles to check, fewer places where a warning gets buried under a different warning. That matters more than it sounds, because lookalike domains only get caught if somebody is actually reading the mail flow data on a normal Tuesday afternoon.
Small things that helped
First, put the domain into the display name for your twenty largest counterparties. It looks clumsy for a week, then everyone stops noticing, and a one-letter difference becomes visible without a click.
Second, when a client mails about bank details, call the number on the contract, not the number in the signature. We do this even when it is inconvenient, and it has killed two attempts so far.
Third, ask your mail team for a rule that flags newly registered domains resembling your existing counterparties. It is cheap and it fires rarely enough that people still read the alert when it appears.
Fourth, retype the address from your own records instead of continuing an old thread. It takes ten seconds and it breaks the inheritance chain that started the whole mess.
Wrapping up
We recovered the money, mostly because the client called us. That is not a control, that is luck, and I would rather not run a finance team on luck. The fix was small in the end. One rule, one phone habit, and one conversation with the person who sent it, who felt awful and did not need to.
If you run mail for a company of any size, spend an afternoon on this. Not a project, not a rollout, just an afternoon with a list of your counterparties and someone who can write a mail filter. It will feel boring. Boring is the point.
References
https://attack.mitre.org/techniques/T1583/001/
https://www.cisa.gov/resources-tools/resources/phishing-guidance-stopping-attack-cycle-phase-one
https://csrc.nist.gov/pubs/sp/800/63/b/upd2/final
https://www.enisa.europa.eu/topics/cyber-threats




