Decisions in a DM Do Not Exist

A decision made in a direct message has not been made. Two people hold a version each and nobody else can search it. We run one shared channel per client relationship, and anything that changes what gets built is said there.

Why do decisions made in DMs get made twice?

Because the record is private and partial. The two people in the thread each remember a version, and neither sits anywhere the designer or the person covering next week can read. When the work reaches them, they make the call again on their own judgement.

Nothing looks broken while it happens. The page stops matching what the client asked for, and the correction arrives later as feedback.

Where each kind of message belongs, and how it fails somewhere else:

What is being said	Where it belongs	Why	Failure mode elsewhere
A yes or no that unblocks work	Shared channel	It is a decision, not a chat	Work restarts on somebody’s memory of it
Copy or design feedback	Shared channel	The next editor needs the reasoning	The same note given twice, differently
Scope or price change	Written confirmation outside chat	It changes the agreement	Two versions of what was agreed
Weekly status and blockers	Shared channel, weekly	Everyone reads the same thing at once	Only whoever asked gets told
Feedback about a person	A call, then a DM	It is about a person	A correction with an audience
Anything a new joiner needs	Channel or the project doc	They cannot search your DMs	Onboarding by interrogation

Why does everyone keep using DMs anyway?

Because they skip the audience. Writing in a channel means phrasing the thing for people who lack the context, which takes the sender longer than the DM would have. That saving is real, and the cost it creates lands later, on somebody else.

Our team does not share a working day. An answer typed at the end of one person’s evening arrives while the people who need it are asleep, and by morning it sits in a thread they have no way to search.

What does one channel per relationship look like?

One channel named for the relationship and not for the project, with both companies sitting inside it. We run 8 client accounts across 11 brands.

Where one owner runs several products, each brand is treated as its own account, because each one needed its own voice and its own ICP. An approval on one brand’s copy tells you nothing about the next one.

What goes in the channel: the ask, the answer, the blocker, the link to the thing that shipped. Every account gets delivery every week, and the questions that come out of that week go to the channel rather than to whichever of us the client was already talking to. The test is whether somebody who joins the account later would need the thread to understand why the work looks the way it does.

What still belongs in a private message?

Anything about a person. Performance, pay, a complaint about somebody’s work, the personal reason behind a missed deadline. Those start on a call and continue in a DM. Moving one of them into a shared channel gives an audience to something that needed a conversation.

Commercial negotiation sits in the same bucket until it is agreed, and then the agreed version moves somewhere permanent. Retention is a setting somebody else controls, so chat history is not a contract.

A decision in a DM does not exist for anyone else.

What do you do when a client keeps sending DMs?

Answer, then post the answer in the channel and tag whoever it affects.

The people this protects are not in the thread at all. Whoever covers the account next week reads the channel, and so does whoever joins the account later, because neither of them can search a conversation they were never in.

Related reading: what belongs in a client update and running a team across time zones.