A multichannel social strategy you can actually sustain

Every social plan starts as a list of platforms. The list is always longer than the number of platforms anyone is still posting to by the end of the year.

That isn’t a discipline problem, and a content calendar won’t fix it. It’s a capacity problem written down as a marketing decision: the channel count got chosen before anyone asked who would feed it on a bad week.

So here is the sequence we’d argue for, and the order matters more than any individual step in it.

Start with an objective you could fail

Not “increase brand awareness”. You cannot fail that, which means you cannot act on it either.

An objective worth writing down names the thing that changes and the way you’d know: qualified enquiries from a named segment, applications for open roles, support volume moving off the phone, product feedback from people who actually use it, a named account list becoming reachable. Make it specific enough that six months from now someone can say plainly that it didn’t happen.

Be strict about it, because the objective decides the channel, and when it’s vague, every channel looks equally justified. That’s how you end up with five.

Audience first, platform second

The platforms are downstream of a question about people: where does the person you’re trying to reach already spend attention, in the mode you need them in?

That last clause does the work. Your buyer may well be on three networks. They are only in a buying frame of mind on one of them, and they are only reachable in a professional capacity on some. A B2B audience scrolling short-form video in the evening is a real audience in a useless mode.

Research it rather than assuming it: ask your last ten customers where they first heard of you, check which networks already send you traffic, and look at where competitors get replies rather than where they post.

The platforms, described by what they are

Descriptions of social platforms date fast when they’re written as trends. Written as structural properties, they hold up:

• LinkedIn is the only large network where people are logged in as their job. Professional identity is the unit of participation, which is why B2B works here and why anything that reads as consumer advertising doesn’t.
• Instagram is visual-first and rewards things with a physical, spatial or craft dimension: product, place, process, people.
• TikTok and short-form video generally distribute by interest rather than by follower count. That’s why a new account can reach people, and why a follower number means less here than anywhere else.
• YouTube behaves like a search engine more than like a feed. It’s the most demanding format to produce and the only one where a single piece of content keeps working long after you published it.
• Facebook is broad, skews older, and its durable value is groups and local reach rather than a page feed.
• X is real-time and reactive. Narrow, fast, and useful mainly if your audience genuinely lives there.

Pick two to four. Two if the work is being absorbed by someone who already has another job.

And be clear-eyed about what you’re building on. All of it is rented, and the distribution can change without notice; it has, repeatedly, on all of them. Some of the programme’s output should end up somewhere you own: an email list, your own site, a resource people come back to. The owned channel is the destination; the social channels are the route.

Content: one idea, several native forms

The failure here is cross-posting: the same image and caption pushed to five networks by a scheduler, including the two where it reads as an obvious import.

The alternative isn’t five times the work. It’s one idea expressed in the form each platform actually rewards: a short video, a carousel, a written post, a clip pulled out of a longer recording. Same argument, different shape. Consistency of message, not consistency of asset.

For a company nobody has heard of yet, the format that consistently earns attention is the specific story: how something is actually made, a decision you got wrong, what a customer’s real problem looked like before you were involved. Generic industry commentary is invisible because everyone can produce it.

Consistency is an operating problem

Two posts a week forever beats a launch burst and then silence, and this is the point at which most programmes quietly end. A profile that was last updated a long time ago is worse than no profile: it’s a visible signal that the company started something and stopped.

So make it an operation. A calendar with owners and dates, a batch-production habit so you aren’t writing on the day, and a realistic floor: the cadence you can hit in your worst month, not your best one. Publish above the floor whenever you can.

Automate the calendar, not the conversation

Scheduling tools (Buffer, Hootsuite, Sprout Social and their equivalents) genuinely earn their place. Queue posts across accounts, pull performance into one view instead of six dashboards, and monitor mentions and keywords in a single stream. This is ordinary tier-one tooling in the sense we use in our guide to what to automate first: a solved category, buy it, don’t build it.

What should not be automated is the part where a human being is expecting a human being. Automated replies to comments and DMs are detectable, and being detected is the entire cost: the interaction that was supposed to build trust demonstrates instead that nobody is home.

The nuance is that a well-bounded assistant on a messaging channel is a different thing from a canned auto-reply. Social DMs turn into a support queue whether you planned for it or not, and the same rules apply as anywhere else: a bounded scope, an explicit handover to a person, and someone able to see the transcripts. That argument is made properly in our piece on AI assistants for customer contact.

The three ways automation embarrasses people

Over-automation. A feed that is entirely scheduled reads as a broadcast, and audiences are good at spotting it. Keep a proportion of posting reactive and unscheduled; that’s where replies come from.

The queue nobody stopped. This is the one that does real damage. When something unexpected happens (in the world, in your industry, or to you), a scheduled promotional post publishes anyway, into a room that has changed. Every programme needs a named person with authority to pause the queue and a written procedure for doing it, in place before you need it, because you will need it at an inconvenient moment.

Vanity metrics. Followers and likes are the easiest numbers to report and the least connected to anything. They’re what a dashboard produces by default, which is precisely why they end up in the monthly deck.

Engagement is the part that pays

Posting is the cheap half. Replying is where the channel converts from a broadcast into a relationship, and it’s the half that gets cut first when the person running it is busy.

Reply to comments, questions and mentions, including critical ones. Set an internal expectation for how fast, not because a platform requires it, but because an unanswered public question is a worse look than never having posted. Ask things that can be answered. And treat the replies as a research channel: the objections, the misunderstandings and the repeated questions are a free description of what your positioning isn’t doing.

Listening is the same discipline pointed outwards. Track your name, your product names, your competitors and the two or three phrases people use to describe the problem you solve. Most of that stream isn’t actionable. The part that is tends to be worth more than the analytics.

Measure something you’d act on

Define one thing that counts as success on this channel (the same discipline we argue for in choosing an e-commerce platform) and instrument it before you change anything, or you’ll have no baseline.

Then be honest about attribution. Social is a poor last-click performer almost everywhere, and last-click is the model most analytics tools hand you by default. That doesn’t mean it isn’t working; it means the measurement is answering a different question. Tag links consistently so you can at least see the sessions it produces, and watch the slower signals alongside them: direct traffic, branded search, and how often “I’ve been following you” appears in a first sales call.

Review on a fixed cadence rather than continuously. Look at which content earned replies rather than impressions, which platform produced anything resembling your objective, and what you should stop doing. Stopping is the decision people skip.

Doing it in-house, or not

Some of this can be handed to someone else and some of it can’t, and the split is fairly clean.

Hard to outsource: the point of view, the domain knowledge, approval on anything sensitive, and ownership of a public problem when one happens. An outside team can draft a response to a complaint. It should not be the last authority on it.

Reasonable to outsource: production and editing, scheduling, listening setup, reporting, paid distribution, and the design work.

If you do bring someone in, ask about method and control rather than about their showreel. Does the proposal start from your objective, or from a content calendar? Who owns the accounts, the ad account and the analytics property: you, unconditionally, from the first day? What’s the escalation path when something goes wrong publicly? What do you keep if the relationship ends?

Account ownership is the one nobody asks about and the one people regret. Everything else is recoverable.

The short version

• Write an objective you could fail.
• Choose the audience before the platform, and the platform before the calendar.
• Two to four channels. Two if this is somebody’s second job.
• One idea, expressed natively per platform. Never cross-post.
• Set a cadence you can hit in a bad month, then beat it.
• Automate scheduling and listening. Do not automate the conversation.
• Name who can pause the queue, before you need them to.
• Measure one thing you’d act on, and ignore follower counts.

Social sits inside our marketing practice alongside development, e-commerce and AI work rather than in a separate contract, which is usually why it comes up: the content is easy to arrange and the plumbing underneath it isn’t. On Deren the analytics and SEO tooling went in as part of the migration rather than afterwards, so there was no measurement gap at exactly the moment it mattered most. We also run bounded assistants on messaging channels where DMs have become a support queue, and we operate SEOaaS, our own AI content and SEO platform. If your channels are producing activity but nothing you can act on, tell us what you’re trying to make happen.