Social media client onboarding: the first seven days

Almost every difficult client relationship was decided in the first week. Not by the work — by what was never asked, never written down, and never agreed. Three months later the argument is about a post, but the cause is that nobody defined who approves, what the brand refuses to say, or what happens when the client goes quiet for nine days.

Onboarding is the cheapest leverage in the whole engagement. Here is a seven-day version that front-loads the decisions.

Day 1 — Access, before anything else

Access requests take longer than anyone plans for, so they go first and run in parallel with everything else. Send one message with the complete list rather than discovering items one at a time over two weeks:

• Publishing access to each channel, granted the right way — as a role on the company page or a delegated account, never a shared password. Personal-profile logins break the moment the client changes their password, and they are an offboarding problem you do not want.
• Analytics access, read-only, for whatever they already have.
• Brand assets: logo files, fonts, colours, image library, existing templates.
• Anything already written: old posts, newsletters, website copy, sales decks, support macros. This is the single most valuable thing you will receive and clients rarely think to offer it.
• One named person who can answer a question same-day. Not a group inbox.

Set the expectation now that missing access is a blocker with a date attached, not a background task.

Day 2 — The interview that determines the voice

An hour on a call, recorded. This is not a discovery call about goals; it is raw material extraction. Adjectives are useless here — "professional but approachable" tells you nothing two writers would agree on. Ask for specifics:

• "Show me three posts you were proud of, and three you disliked." Then ask what was wrong with the second set. The dislikes are more informative than the likes.
• "Who specifically are you talking to?" A job title and a situation, not a demographic.
• "What do you refuse to say?" Claims they cannot make legally, comparisons they will not draw, words that are banned internally. This list prevents more damage than any style guide.
• "What's true that your competitors can't say?" This is the positioning question in a form clients can actually answer.
• "I or we?" And whether that ever changes by channel.
• "What went wrong with your last agency?" The most useful question on the list. It tells you the failure mode you are being hired to avoid.

Turn the recording into a written voice definition the same week. If you cannot write a post from it without guessing, it is not finished — the standard for a usable definition is in keeping several brand voices straight.

Day 3 — Facts and boundaries, written down

Voice tells you how it sounds. Facts tell you what it is allowed to say. Separate them, because they change at different rates and different people own them.

Collect and store, per client:

• Numbers you may cite, with the date they were true. Customer counts, funding, results, retention. Undated numbers become wrong silently.
• Named customers you may mention, and whether by name or anonymously.
• Claims requiring legal or compliance review before publishing.
• The forbidden list from day 2.

This is the file that stops the expensive mistake: a plausible, well-written post containing a number nobody verified. It is also the file that makes an AI-assisted workflow safe rather than dangerous, because generated drafts fill gaps with invention unless something holds the real values.

Day 4 — Agree the approval mechanics

Two agreements, both in writing, both boring, both worth more than the rest of the contract.

Who approves, and who is the tiebreak. One name. If the founder and the head of marketing both comment and disagree, you need to know in advance whose answer wins, because discovering it during a disagreement is how relationships fray.

What happens on silence. "No response by end of day Thursday" has to mean something specific: either the post auto-proceeds because it is low-risk, or it slips to next week and the client owns that. Ambiguity here is what turns one-day approvals into one-week ones. The full argument is in the approval workflow that doesn't lose the client.

Agree also when they see work: approving angles for the week is fast and cheap, approving a finished scheduled post feels like being handed a bill. Move the approval earlier and the whole engagement speeds up.

Day 5 — The first draft batch, deliberately early

Write three posts and send them on day five, not day fifteen. They will be somewhat wrong, and that is the point: nothing extracts a real voice brief like a client reacting to concrete text. An hour of line-by-line feedback on three drafts is worth more than another discovery call.

Frame it explicitly as calibration so nobody thinks it is the finished standard. Ask for edits in the text rather than comments about the text — how they rewrite a sentence tells you more than what they say about it.

Day 6 — The plan, one page

Now that you have heard them react, write the first month:

• Cadence per channel, honestly scoped to what you were paid for.
• Three to five recurring content types you will run. Named, so both sides can talk about them.
• What you will measure, and what you will not. Agree now that you are not reporting on follower count if follower count is not the goal.
• The one thing to prove in month one. A single objective beats a list.

Send it for approval as a plan, not a proposal. It is the thing you will point at in month three when the request arrives to do something entirely different.

Day 7 — Set the standing rhythm

The engagement needs a metabolism, agreed now rather than negotiated weekly:

• A weekly material drop. Five minutes from the client: what happened, what launched, what a customer said. Without this you will be writing from the website by month two, and it will show.
• A fixed approval window. Same day each week, both sides know it.
• A monthly review, thirty minutes, on what worked and what changes.

Then start. Do not extend onboarding into week three chasing polish; the remaining unknowns resolve faster through published work than through more meetings.

The one-page checklist

Where this stops scaling

At one or two clients this runs on notes and attention. At five, the artefacts multiply — five voice definitions, five fact lists, five forbidden lists, five approvers, five silence rules — and they stop living anywhere reliable. That is the point where onboarding quality quietly degrades and every new client gets a slightly worse version of the process.

SelfSM is built around that: voice, facts, constraints and approval trail stored per brand, so onboarding a sixth client is the same process as the first rather than one more thing to hold in your head. The client approves through a link scoped to their brand and never sees the rest of your book.

FAQ

What should I ask a new social media client? Ask for three posts they liked and three they disliked, who specifically they are talking to, what they refuse to say, what is true that competitors cannot claim, and what went wrong with their last agency. Specifics beat adjectives — "professional but approachable" is not a brief.

How long should onboarding a social media client take? About a week of elapsed time, with the first calibration drafts sent by day five. Longer onboarding rarely produces better output; the remaining unknowns resolve faster through real drafts than through additional discovery calls.

What access do I need from a social media client? Role-based publishing access on each channel (never a shared password), read-only analytics, brand assets, everything they have already written, and one named person who can answer questions same-day.

How do I stop client approvals from dragging? Name one approver with a tiebreak, agree in writing what happens when nobody responds by the deadline, and get approval on the week's angles before drafting rather than on finished posts afterwards.

What should be in a brand voice document? Three real sample posts, first person or brand, the words they use and the words they refuse, one line on stance, and a separate list of factual claims they may make with the dates those numbers were true.