Why approval processes collapse

4 reasons, in order of how often they kill one.

The client has to log in. This is the big one. A busy owner will not create an account, remember a password and learn an interface to approve 4 captions. They will reply to whatever arrives on their phone.

Nothing happens when nobody replies. A draft sitting in “awaiting approval” forever is a queue with no exit. Silence has to mean something.

Feedback arrives in 3 places. Comments in the tool, a WhatsApp voice note, a correction on the call. Now nobody knows which version is current.

Everything needs approving. If a reply to a comment has to go through the same process as a campaign launch, the process becomes the bottleneck for work that is time-sensitive by nature.

Rule one: approval goes to them

The client should never log in to anything. They receive a link, they see the posts as they will appear, they tap approve or write a note. That is the entire interaction, on a phone, in 90 seconds.

Every hour of friction you add here comes back as a week of delay. The agency logs in. The client taps.

Rule two: silence has a meaning, agreed in advance

Write it into the engagement, not into a policy document nobody reads.

The version that works: “Content goes out for approval by Thursday for the following week. If we have not heard back by Monday morning, we publish as drafted.”

Clients almost never object to this, because the alternative is that their account goes quiet and they know it. What it removes is the week where nothing ships and nobody can say why.

The one exception worth carving out: anything involving price, legal claims or a named third party waits for an explicit yes.

Rule three: one place for feedback

Not because tidiness matters, but because scattered feedback produces published mistakes.

If a client sends a voice note, the person who hears it writes the change into the draft as a comment before doing anything else. The voice note is not the record. The draft is.

This sounds like bureaucracy until the week someone acts on a verbal instruction that was later reversed, on a post that is now live, and nobody can reconstruct who said what.

Rule four: 3 tiers, not one

The most common mistake is treating all content as equally risky.

No approval. Replies to comments, review responses, community management. These are time-sensitive and low-risk, and a reply that needs sign-off arrives too late to be worth sending. Agree the tone once and trust the person doing it.

Batch approval. The standard calendar, approved a week at a time in one sitting. This is 90 percent of the output.

Explicit approval. Campaign launches, anything with a price or a legal claim, anything sensitive, anything naming another company. Individually, in writing, no exceptions.

Getting this split right is what makes the whole thing survive a busy month.

What to do about the client who never replies

Every agency has one. 2 approaches work and one does not.

Works: the deadline rule above, applied consistently from week one. They adapt or they accept what ships.

Works: a standing 15-minute slot. Same time every week, screen shared, approved live. For a certain kind of client this is the only thing that ever works, and it is cheaper than chasing.

Does not work: chasing. Three follow-ups a week trains them that the deadline is soft, costs you the time you were trying to save, and makes you feel like a supplier rather than an adviser.

The record you will be glad of

Keep who approved what and when. Not to win an argument, though it does that.

Keep it because 6 months into an account, the question is rarely “who approved this” and usually “how long does approval actually take here”. If the answer is 4 days, your Thursday deadline is wrong and the calendar should be built a week further ahead.

That is a process fix you can only see if someone recorded the timestamps.

EngageDigital is building approval workflows with client-side review by link, so the person approving never needs an account. It sits on the Business plan. See where it is in the queue or join the early-access list.

What actually blocks you

3 things, every time, and all three are predictable.

The Instagram account is personal, not business. No API access at all until it is converted and linked to a Facebook Page. 5 minutes for the client, a week if nobody tells them.

Nobody knows who owns the Facebook Page. A former employee, a cousin, an agency from 2019. Until someone with admin rights grants access, nothing can be connected.

The Google listing is unverified or claimed by someone else. Verification can take days and sometimes involves a postcard. Start it on day one or it blocks the entire local programme.

None of this is difficult. It is all slow, and it only becomes slow if you discover it in week two.

The pre-flight message

Send this before the kickoff call, not after. It converts a week of back-and-forth into one reply.

Ask them to confirm, in writing:

• Which accounts exist, with the exact handle for each
• Whether Instagram is a Business or Creator account, and which Facebook Page it is linked to
• Who has admin on the Facebook Page and the Business Manager, by name
• Whether the Google Business Profile is verified, and who holds it
• Whether a Snapchat Public Profile exists, if that is in scope
• Which accounts they want us to touch, and which are off limits

That last line saves an argument later. There is always one account someone in the family runs.

Connect by permission, never by password

Do not accept a password. Not for convenience, not because the client offers.

Every platform has a proper permission model: Business Manager roles on Meta, page admin on LinkedIn, channel permissions on YouTube, manager access on Google Business. Use them. The client keeps ownership, you get access, and when the relationship ends they revoke it in one click without changing a single password.

A shared password is also a straightforward security problem, and if you handle EU clients it is one you would have to explain in a breach notification.

Every serious tool connects through the platform’s own login screen and stores a revocable token. If something asks you to type a client’s password into it, that tells you what you need to know about it.

The audit before you post anything

2 hours, before the first piece of content. It is also the best first deliverable you will ever send.

The Google listing, field by field. Hours including holidays, categories, services with prices, attributes, action links, photos from this year. Most listings are half empty and this is the highest-return hour in the whole onboarding.

Unanswered comments and reviews. There will be months of them. Clearing the backlog in week one is visible, cheap, and occasionally recovers a customer.

What actually worked before you arrived. Their best posts from the last year, before you overwrite the approach with your own. Sometimes the thing that works is the owner’s phone camera, and replacing it with polished graphics makes performance worse.

Bios, links and handles. Consistent name, working link, current description. 10 minutes, always wrong somewhere.

The first week, in order

Day one. Send the pre-flight list. Start Google verification if it is needed, because it is the only thing with a queue you cannot control.

Day two. Connect everything that can be connected. Note what is blocked and who has to unblock it, by name.

Day three. Run the audit. Send it as a short document. This is when the client decides you were worth hiring.

Day four. Clear the comment and review backlog. Agree tone for replies while you are in there.

Day five. Build 2 weeks of calendar and get it approved. 2 weeks, not a month: you do not know enough yet, and a month of content built on a wrong assumption is a month wasted.

What to write down while it is fresh

The things you will otherwise ask twice.

Who approves content and how fast they actually reply. Which topics are off limits. The one competitor they do not want mentioned. Whether the owner wants to be in photos. Which of their own past posts they liked.

Keep it in one page per client, somewhere the whole team can see. In a year it is the difference between an account anyone can cover and an account that only one person understands.

EngageDigital connects every account through the platform’s own login, never a password, and imports past posts on day one so the reporting does not start from empty. See how it works for agencies or join the early-access list.

Who this actually affects

Start with the number, because it changes how seriously people take the rest. Somewhere around one in 20 people has a vision impairment significant enough to affect how they use a phone. On an account with 10,000 followers that is several hundred people for whom your post is a blank.

Screen readers announce alt text where it exists. Where it does not, Instagram and Facebook substitute an automatic guess, which produces things like “may be an image of one person and text”. Technically a description. Useless as one.

And it is not only permanent impairment. It is a cracked screen, bright sun, a slow connection that never loads the image, and anyone who has turned images off to save data.

What each platform gives you

Instagram. Alt text per image, including every slide of a carousel up to 10. This is the one most often missed, because it is per slide rather than per post.

Threads. Per image, on up to 10 files.

Bluesky. Per image and, unusually, on video too. Still rare across the industry.

Pinterest. Alt text up to 500 characters, and it does double duty: Pinterest is a search engine wearing a feed, so the description does ranking work as well as accessibility work.

LinkedIn and Mastodon. Per image.

X. Per image, though it is easy to skip in most tools.

Video captions are a separate problem with a bigger audience. Most video on social is watched with sound off, which means captions are not an accessibility extra; they are the only way most people consume it at all.

How to write one that works

Alt text is not a caption and it is not a place for keywords.

Describe what matters, not everything. “A barista pouring a flat white, seen from above” beats a full inventory of the counter.

Read the text in the image. If a slide says “3 ways to cut your reply time”, the alt text says that. An image of text with no alt text is completely lost.

Keep it to a sentence. Around 125 characters is the practical limit before a screen reader gets tedious.

Skip “image of”. The screen reader already said it is an image.

Decorative means empty. A background texture with nothing to communicate should have empty alt text rather than a description of the texture.

The carousel problem, specifically

A carousel is usually a sequence: a hook, then steps, then a conclusion. Each slide is a separate image with separate alt text, and treating them as 10 independent descriptions loses the thread.

Write them as a sequence too. Slide one sets up what the carousel is. Slides two to nine each carry their own point. The last one carries the call to action, because a listener who reaches the end should know what to do next, the same as everyone else.

It takes about 4 minutes for 10 slides. That is the entire cost.

Where AI genuinely helps

This is the one part of social media work where generated text is unambiguously better than the alternative, because the alternative is nothing.

A model reads the image and produces a usable description in a second. It will miss context it cannot see, so skim it and fix the one in 10 that is wrong. But the comparison is not “AI alt text versus excellent hand-written alt text”. It is “AI alt text versus a blank field”, and the blank field is what almost everyone ships.

If your tool can generate it per slide, the 4 minutes becomes 20 seconds and the excuse disappears.

The reach argument, for the sceptics

If accessibility on its own does not move the decision, there is a second reason.

Alt text is text the platform can read. On Pinterest it directly affects whether a pin surfaces for a search. Across the others, it is one more signal about what the post is, in a system that mostly guesses from engagement.

It will not transform your reach. It costs 4 minutes and helps several hundred people per 10,000 followers actually see what you published, which is a better return than most things on the calendar.

EngageDigital exposes alt text per carousel slide rather than one field per post, and can draft it from the image itself. See how publishing works or join the early-access list.

Why the usual report fails

The standard version is a dashboard export with a follower count at the top. It fails for 3 reasons.

It answers a question nobody asked. The client does not want to know their impressions. They want to know whether this is working and what happens next.

It cannot be checked. Blended scores and invented indices sound authoritative and mean nothing. A client who asks how a figure was calculated should get an answer in one sentence.

It has no opinion. A page of numbers with no interpretation transfers the analysis back to the person paying you to do it.

The five-page structure

Long enough to be substantial, short enough to be read. This shape works for almost any client.

Page one: what we did. Posts published, by network, against what was planned. Boring and essential. A client paying a retainer is buying consistency first.

Page two: what happened. The real platform metrics, named in each platform’s own terms. Reach on Instagram, impressions on LinkedIn, views on TikTok. Never blended, never summed into a total that means nothing.

Page three: what worked. The three best posts and the three worst, with a sentence each on why. This is the page clients actually read, and the one that is impossible to automate.

Page four: what we learned. Comments and reviews, response time, the questions people kept asking. 20 comments about delivery is a product page problem, and telling them that is worth more than any impression count.

Page five: what next. Three specific things you will do differently, and anything you need from them. A report that ends without a decision is a newsletter.

Metrics to include

Publishing consistency. Planned versus published. If it slipped, say so on page one rather than hoping nobody notices.

Reach or impressions, per network, unblended. Label whether each is reported or derived.

Engagement rate with its denominator stated. Per impression and per follower are different numbers with the same name. Pick one and keep it for a year.

Response time on comments and reviews. Almost nobody reports this and it is the metric a client can feel.

Referral traffic. From the website analytics, not the platform. It survives every API change and it is closest to money.

Search terms, for local clients. The monthly search terms from the Google listing, next to calls and direction requests, beat every social metric for a business with a shopfront.

Metrics to leave out

Follower count as a headline. It moves slowly, it is easily bought, and it tells you nothing about the month.

Any invented composite score. If you cannot explain the formula in one sentence, it does not belong in a document with your name on it.

Cross-network totals. A TikTok view and a LinkedIn impression are not the same event. Adding them produces a bigger number and a smaller truth.

Anything from a network that reports nothing. X, Discord and Slack give no analytics to any tool. Report activity there, and put a dash where the number would be.

The page that keeps clients

Page three. 3 posts that worked, three that did not, and a sentence of judgement on each.

It is the only part of the report that proves a person was involved. Everything else could have been generated. This is where the client sees that someone looked at their account and formed a view, and it is the reason they renew.

The sentences do not need to be clever. “This one worked because it answered a question 3 people had already asked in the comments” is enough, and it is more than most agencies write all year.

Cadence and format

Monthly for most retainers. Quarterly for a strategic review with year-on-year comparisons, which only exists if someone has been storing history since day one.

Send it as a document, not a dashboard login. Clients do not log in. They forward a PDF to someone who was not on the call, and that person needs it to make sense alone.

Send it on the same date every month. A report that arrives on the third Tuesday is a system. One that arrives when someone remembers is a favour.

EngageDigital keeps full analytics history from the day an account connects and exports per client, so a quarterly review can look back further than a month. See how reporting works or join the early-access list.

Why the gaps exist

Social analytics do not come from measurement. They come from an API returning numbers the platform chose to expose, under rules the platform changes without consulting anyone.

3 things follow from that, and all three surprise people.

The metrics differ by network. Reach on Instagram, impressions on LinkedIn, views on TikTok. They are not the same measurement, and stacking them into one total is arithmetic that means nothing.

The history is short. Most platforms expire their numbers after about 30 days. If nobody stored a snapshot, last March is gone for everyone.

Some networks return nothing. Not less than you want. Nothing.

The three with no analytics at all

X. Reading engagement data now sits behind enterprise pricing. You can publish and you cannot measure. Anyone showing you X engagement numbers has derived them from something else.

Discord. No analytics API exists, for anyone. What you can report is that the post went out, to which channel, when.

Slack. Same. No message analytics are offered to third-party tools at all.

For these three the honest report line is a dash and a footnote, not a zero. A zero implies nothing happened. A dash says nobody can see.

The ones with partial data

Reddit gives post-level metrics — score, comments, upvote ratio — and almost nothing at profile level. No follower analytics worth the name.

Threads covers only content published after April 2024, when its API opened. Older posts return nothing, for everyone.

Bluesky returns follower counts and basic engagement. Impressions are derived rather than reported, which is a calculation, not a measurement, and should be labelled that way.

Mastodon collects very little by design. That is the point of the network, and no tool can give you more than the instance holds.

Google Business Profile has no per-post analytics. Instead it gives location impressions and the monthly search-terms report, which is more useful than post data for a local client but arrives in a different shape.

What a client actually needs to be told

Not at the end of the quarter. At the start of the engagement, in the proposal.

The sentence that works: “11 of the 15 networks we publish to report performance data. Three report none, because their platforms do not offer it to any tool. For those we report activity and outcomes instead.”

2 things happen when you say this early. The client stops asking for a number that does not exist, and they start trusting the numbers that do — because you were the one who told them where the edges are.

Reporting the unmeasurable networks

A dash in a table is honest but unhelpful on its own. 3 things can be reported for Discord, Slack and X without inventing anything.

Activity. Posts published, on schedule, to which surface. A client paying for consistency can see consistency.

Downstream effect. Referral traffic in the website analytics, which the platform does not control. This is the one number that survives every API change.

Qualitative. Screenshots of the thread, the reply, the reaction. Unfashionable, and frequently the most persuasive page in the report.

The history problem, and why it compounds

The 30-day window is the gap people notice last and regret most.

Everything looks fine for the first month. Then a client asks how this September compares with last September, and the answer is that nobody kept it. Not the tool, not the platform, not you.

The fix is unglamorous: a daily snapshot, stored, from the day an account is connected. It is the most expensive thing to promise because the storage never stops growing, and it is the only way a year-on-year comparison exists at all.

If you are choosing a tool, ask exactly one question about analytics: does the history start when I connect the account, or does it roll? The answer tells you whether you can write a proper annual review 12 months from now.

What good looks like

A report that can be defended line by line. Metrics named by their platform’s own term rather than blended into a single invented score. Gaps shown as gaps. Derived numbers labelled as derived.

It is less impressive than a dashboard with a big number on it, and it is the only version that survives a client who reads carefully.

EngageDigital snapshots analytics daily and keeps the full record, and shows a dash where a platform gives nothing rather than filling the space. See how reporting works or join the early-access list.

The short version

Publer is the best value under about 12 accounts. 4 dollars an account for unlimited scheduling and eternal post history is genuinely hard to beat.

Buffer is the most mature product here, with a comment inbox across 10 networks, a browser extension, mobile apps and 15 years of integrations.

Later is the Instagram-first planner, with link in bio and social listening that neither of the other 2 includes.

Each of them is a reasonable choice. Which one depends almost entirely on how many accounts you run and how many people run them.

Pricing, at the same size

All three price by unit, and the units differ.

Buffer. $6 a channel a month on Essentials and $12 on Team for the first 10 channels ($5 and $10 billed yearly). Channels above 10 are cheaper in bands, down to $1 on Essentials past 50. Free tier: 3 channels, 10 scheduled posts per channel.

Publer. $8 an account a month on Business, $4 on Professional, billed yearly, with every tenth account free. Extra team members are charged on top. Free tier: 3 accounts, no X.

Later. Social sets of 8 profiles. Starter $18.75 for one set and one user, Growth $37.50 for 2 sets and 2 users, Scale $82.50 for 6 sets and 4 users. Extra sets $11.25, extra users $3.75.

At 20 clients with 10 accounts each, 200 profiles, that comes out roughly as Buffer $345, Publer $1,440, Later $296 a month.

At 10 accounts and one person, the order reverses: Publer around $72, Buffer around $50, Later $18.75 if they fit into one set.

That reversal is the whole story of this category. Nothing is expensive or cheap in the abstract.

Networks each one reaches

Buffer publishes to 11: Instagram, Facebook, LinkedIn, YouTube, TikTok, X, Threads, Pinterest, Bluesky, Mastodon and Google Business Profile. No Reddit, Discord, Slack or Snapchat.

Later publishes to eight: Instagram, Facebook, TikTok, Pinterest, LinkedIn, YouTube, Threads and Snapchat. No X, Reddit, Discord, Slack, Bluesky, Mastodon, and no Google Business Profile at all.

Publer covers a broad list including several of the community networks. Check their current page rather than ours, because it changes.

If your clients live only on the big visual networks, all three are fine. If one of them runs a Discord server or a subreddit, the list gets short quickly.

Where each one genuinely wins

Buffer wins on maturity and on the inbox. Their community inbox covers 10 networks with AI replies and saved replies, which is more networks than we cover. The browser extension, the mobile apps and the integration list are years ahead of anything launching this year. If you have 5 channels and want something that works today, this is the safe answer.

Later wins on Instagram craft and on the extras. Best times to post, visual grid planning, Reels audio, plus link in bio and social listening with brand mentions and competitive benchmarking. If your work is mostly Instagram and you value planning tools over breadth, they are built for exactly that.

Publer wins on value and on volume. Unlimited scheduling at every paid tier, eternal post history from $4 an account, RSS automations and spintax post recycling. For one brand or a handful of accounts, it is the most software per dollar in this comparison.

Where they struggle

Fairly, and with the same specificity.

All three charge by unit. Accounts, profiles or seats. Every one of them gets more expensive as you succeed, and two of them meter more than one thing at a time.

None manages a Google listing properly. Buffer and Publer post to Google Business. Later does not support it at all. Posting is a fraction of the work: hours, services, menus, attributes, action links and the search terms report are what decide whether a local business appears in the map pack.

The community networks are thin. Reddit, Discord and Slack are absent or partial across the board, and they are where a lot of product audiences now actually live.

Where we sit, stated plainly

We publish to 15 networks including the four that get skipped, manage the whole Google Business listing rather than posting to it, and charge a flat price with no per-profile and no per-seat billing.

And the things we do not have: no link in bio inside the product, though mylink.ie is ours and its Pro tier comes with every paid plan. No social listening, and it is not on the roadmap. No paid advertising management. No 15-year track record, because our beta opens in October 2026.

If you need something that works this afternoon, one of the other three is the right answer and we would say so on a call.

How to choose

3 questions settle it faster than any feature table.

How many accounts will you run in a year? Under 15, per-unit pricing is fine and often cheaper. Above it, model the bill at your target size before you commit.

Do any clients live somewhere unusual? A Discord server, a subreddit, a Google listing that matters more than the feed. That single answer eliminates more options than price does.

Who else needs access? If the answer is 3 people, check whether the tool charges for that, because two of these do.

Figures in this article were read from buffer.com/pricing, later.com/pricing and publer.com/plans on 20 September 2026. They change; check before quoting them. Our side-by-side pages go deeper: vs Buffer, vs Later, vs Publer.

Why any of this is public

2 reasons, neither of them noble.

The first is that we are asking people to join a waitlist for a product that does not exist yet, on the promise of a rate locked for 12 months. That promise is only worth something if the price is defensible, and the only way to show it is defensible is to show where it comes from.

The second is that this category is full of pricing that nobody explains. Per-profile, per-seat, per-bundle, AI credits, analytics tiers. If we are going to argue that a flat price is fairer, we should be able to say why we can afford one.

Where the money actually goes

Running a tool that connects to 15 networks costs money in 4 places. Roughly in order of size.

Platform API access. Most networks do not charge for publishing access, which surprises people. Meta, LinkedIn, TikTok, YouTube, Pinterest, Reddit and the rest expose their APIs under rate limits rather than invoices. What they cost is engineering time, because each one behaves differently and each one changes without warning.

The exception is X, and it is a large one. We come back to it below.

Infrastructure. Hosting, the database, the workers that publish on schedule, and media storage. Media is the part that grows fastest: a video upload is orders of magnitude larger than a post record, and an agency uploading a month of content at once moves real volume.

Analytics storage. This is our own choice and it is not free. Platform APIs expire their numbers after about 30 days, so we take a daily snapshot of every connected account and keep it. Multiply accounts by days by metrics and the row count climbs steadily, forever, for every customer who stays.

It is the single most expensive promise on our pricing page, and we made it deliberately, because “the report you can still run next year” is the thing a per-month-window competitor cannot say.

AI generations. A real per-request cost, paid to our AI provider. A post generated across 6 platforms is one press of a button and one bill. This is why AI is metered per plan rather than unlimited: it is the one feature where a heavy user genuinely costs more than a light one.

The X problem, in detail

X is the only network in the set that charges per action, and the numbers are public: $0.015 for a standard post through the API, $0.20 for one containing a link. Deleting a post through the API also carries a charge.

2 decisions follow from that, and we want them on the record.

We pass it through at cost. Not marked up, not bundled into the subscription. You see the figure in the composer before you publish and it comes out of a prepaid balance.

We do not hide it in the plan price. The alternative would be to average X usage across all customers and raise everyone’s subscription to cover it. That is simpler to explain and it makes light users subsidise heavy ones, which we would rather not do.

There is a third thing worth saying plainly: X also removed analytics access at any price a small company would pay. So on X you can publish and you cannot measure, and no tool on the market is in a different position. Anyone showing you X engagement numbers has estimated them from elsewhere.

What this means for the price

Put the four cost centres together and a shape emerges.

Connected accounts cost us almost nothing. An account is a row, a token and a scheduled refresh. Whether you connect three or 400 barely moves our bill, which is exactly why we do not charge per profile. Tools that do are not passing on a cost; they are choosing a pricing model.

Team members cost us nothing at all. A seat is a user record.

Posts, replies and AI generations do cost. Each one is work: an upload, a call, a retry, sometimes a model request. So those are what we meter, on every plan.

That is the whole logic of the pricing page. Charge for the things that scale with usage, not for the things that scale with your success.

Where the margin sits at 50 customers

At small scale, none of this is a business. 50 customers on a mix of plans is a few thousand dollars a month, against infrastructure that costs a meaningful fraction of that and engineering time that costs far more than all of it.

The honest description of the first year is that the agency work funds the product, which is why this is built by people who already run an agency rather than by a funded startup. It also shapes what we build: features get made because a client needed them, not because a roadmap said so.

The model works at a few hundred customers, not at 50. We would rather write that down now than pretend otherwise and quietly change the pricing in month six.

What could change the numbers

3 things, and we will say so here when they do.

A platform starts charging. X did it. If another follows, it goes on the sub-processor and pricing pages with the real figure, passed through at cost.

Analytics storage outgrows the plan. If keeping full history becomes genuinely expensive at volume, the fix is a longer retention tier rather than silently trimming everyone’s history. We would tell you before, not after.

AI costs move. They have mostly moved down, which is why generations are included rather than sold as credits. If that reverses, the quotas change and the changelog says why.

EngageDigital is built by Proceed Interactive, a Cyprus agency that got tired of paying per profile. The changelog is dated, including the entries where something took longer than we said. Join the early-access list.

The 3 models, and who they suit

Almost every tool in social media management prices one of 3 ways. Knowing which one you are on explains most of what you pay.

Per channel or per profile. Buffer charges $6 a channel a month on Essentials and $12 on Team for the first 10 channels ($5 and $10 billed yearly), then less per channel in bands above 10. Publer charges $8 an account a month on Business, with every tenth account free. The bill is a direct function of how many accounts you manage.

Per bundle. Later sells “social sets” of 8 profiles, one from each network it supports. Scale covers 48 profiles at $82.50, and extra sets are $11.25 each. Cleaner than per-profile, but you pay for the slots whether or not you fill them: 2 clients who only use Instagram consume 2 sets, which is 14 profiles you are not using.

Per seat. Hootsuite charges €99 to €399 per user per month. The account count stops mattering above their entry tier, and the headcount starts mattering instead.

None of these is dishonest. Each one suits a particular shape of customer, and the trouble starts when your shape changes and your pricing model does not.

What it costs at 20 clients

Take a realistic small agency: 20 clients, 10 connected accounts each, 200 profiles, a team of two or three. All figures read from the vendors’ own pricing pages on 20 September 2026.

• Buffer Essentials, monthly billing with its volume bands: around $345 a month.
• Publer Business at $8 an account, every tenth free: around $1,440 a month.
• Later, Scale plus 19 extra social sets: around $296 a month.
• Hootsuite Professional at €199 a seat for 5 people: €995 a month.

Those are not edge cases. 20 clients is a small agency, not a large one.

The number that matters is not the absolute figure, though. It is what happens to it next month.

The incentive problem

Per-profile pricing creates a quiet conflict between your software and your business.

You win a client with 9 accounts. On a per-channel tool that is another $45 to $90 a month. Not fatal, but it is a real cost attached to a good event, and it recurs.

Worse, it shapes decisions. Should this client also be on Pinterest? On a flat plan you try it and find out. On a per-profile plan someone does the sum and the answer is usually no. You end up under-servicing accounts to protect a software bill, which is the wrong thing to optimise.

And it makes churn feel oddly good. Losing a client reduces your costs, which is a genuinely strange incentive to build into a tool you use every day.

Where the lines actually cross

Being fair about this matters more than winning the argument, so here is the honest arithmetic.

Under about 12 accounts, Publer is the best value in this category. 8 dollars an account with unlimited scheduling is hard to beat, and we would not pretend otherwise.

Under about 20 channels, Buffer costs less than our Business plan and comes with years of product maturity, a browser extension and mobile apps we do not have yet.

Under about 30 profiles, if they happen to fit neatly into social sets, Later is competitive.

Above those points, the per-unit models compound and a flat price stops looking like a preference and starts looking like arithmetic.

There is one exception with no crossover point at all: per-seat pricing with a small team. If 3 people need access, a seat-based tool charges you for 3 departments’ worth of software so that 3 people can schedule posts.

The shape we built for

Every tool is built for a shape of customer, whether or not it says so. Ours is specific.

A lot of accounts and a small team. If you have many of one and few of the other, per-unit pricing is working against you in both directions at once.

In practice that is 4 kinds of business:

• The new agency. Five to 20 clients, one or 2 people. The phase where winning work should raise your margin, not your costs.
• The creator across every network. One person, 15 profiles, because being everywhere is the job. Per-profile pricing punishes exactly that.
• The multi-location business. 10 shops, 10 Google listings, 30 social profiles, one marketing person.
• The single brand that keeps adding channels. 3 accounts free, and everything including the Google Business manager stops at $49.

We meter posts, replies and AI generations instead, which are the things that actually cost us money to deliver. Accounts and seats cost us almost nothing, so we do not charge for them.

What to ask before you sign anything

3 questions, whichever tool you end up choosing.

What does this cost at twice my current size? Not today’s price. Model the bill at 40 clients and see whether it is still a number you would accept.

What happens when I hire someone? If the answer is a per-seat charge, that is a tax on a good decision.

What is actually metered? Posts, accounts, seats, AI credits, analytics history, or some combination. A tool that meters 4 things at once is a tool whose bill you cannot predict.

EngageDigital charges a flat price with up to 200 connected accounts per plan and no per-seat billing on any plan. See the plans, compare us with Buffer, or join the early-access list.

Why Reddit is worth the risk

2 things make Reddit different from every other network in a social calendar.

It ranks. Threads surface in search for exactly the comparison queries that sit closest to a purchase: “best X for Y”, “is Z worth it”, “alternatives to”. A thread written 3 years ago is still collecting clicks.

It is trusted. People deliberately append the word to their searches because they want an answer from someone with no incentive. That trust is the asset, and it is also precisely what you destroy if you show up wrong.

So the upside is real and so is the downside. This is the only platform where a bad post costs you more than a wasted slot.

How brands actually get burned

Not usually by being obvious. By being careless in four specific ways.

Missing flair. Many subreddits require a post flair and remove anything without one, silently and immediately. You will see the post live on your own profile and assume it worked.

Breaking a self-promotion rule. Most communities have a ratio, a designated day, or a flat ban on links to your own domain. These are published in the rules and almost never read.

The brand-new account. An account created last week, with no karma, posting a link to its own product, in a community it has never commented in. Every signal says advertisement.

Marketing register. The one that gets you disliked rather than removed. “We’re excited to announce” does not exist as a sentence on Reddit. It reads as someone who wandered in from LinkedIn.

The rules are published, and machine-readable

Here is the part most people do not know: a subreddit’s rules and its flair list are available through the API. You do not have to remember them, and a tool does not have to guess.

A serious workflow fetches both before the post is submitted. The required flair is presented as a field you must fill. A post that would break a stated rule is rejected in your own editor, where the cost is 30 seconds, rather than by a moderator, where the cost is the post and a bit of your standing.

The limitation worth stating: only the rules a subreddit publishes can be checked. The unwritten ones, the tone, the things regulars know, are still your job. No tool solves culture.

What to actually post

3 formats work. Everything else is a coin flip.

The self post that answers a question properly. Text, no link in the body, genuinely useful whether or not anyone buys anything. 40,000 characters are available and the long ones do well. If your product is relevant, it appears once, late, and plainly.

The comment in an existing thread. Higher return than a new post and far lower risk. Someone asks which tool does X, you answer the question, disclose that you make one of them, and say where it is weak. Readers can smell an undisclosed vendor instantly, and they cannot punish a disclosed one.

The profile post. Reddit lets you post to your own profile as well as to subreddits, which most tools ignore entirely. No rules, no removal, and it accumulates into something people can browse.

What does not work: a link to your homepage with a title lifted from your meta description.

A realistic weekly rhythm

You do not need a Reddit strategy. You need about 20 minutes.

Pick 2 subreddits. The ones where your category is actually discussed. Not the biggest ones.

Read the rules once, properly. 10 minutes, and it never has to happen again for that community.

Comment for a month before you post. On other people’s threads, about things you know. This is the part everyone skips and it is the part that decides how the first post lands.

Then one post a month, at most. Reddit rewards scarcity. A brand that posts weekly reads as a brand running a campaign.

Reply to everything. Including the critical comments, especially the critical comments. A vendor who takes a hit gracefully in public buys more credibility than any post.

What the data will and won’t tell you

Reddit exposes almost nothing at profile level. No follower analytics worth the name, no karma breakdown through the API. If you want to measure this channel by a dashboard number, you will be disappointed.

Post-level metrics do exist and are enough: score, comment count, upvote ratio, for the lifetime of the post. The upvote ratio is the one to watch. A post with a high score and a ratio near 50 percent means you started an argument, which may be fine or may be a warning.

The honest measurement for Reddit is different anyway: did the thread rank, and did anyone arrive from it. That lives in your analytics, not in a social tool.

The one-line version

Be a participant who happens to sell something, not a seller who happens to participate. Everything else here is detail.

EngageDigital schedules Reddit posts with the subreddit’s live rules and flair list checked before submission, brings comments back into the inbox, and lets you post to your own profile as well as to subreddits. See how Reddit works in detail or join the early-access list.

The job it is bad at

AI will not find your brand voice. It cannot, because your voice is the accumulation of decisions you have made about what you care about and how you talk about it, and none of that is in the prompt.

Ask a model to write “in a friendly, professional tone” and you get the average of every brand that ever wrote in a friendly, professional tone. It reads like a press release that went to a spa. Your audience has seen 10,000 of those this year and their eyes slide off it.

There is a narrow exception. Give a model five or six of your own real posts, the ones that actually sounded like you, and ask it to match their rhythm and vocabulary. That works, because now the voice is in the input rather than in an adjective. Most people never do this, then conclude AI writing is bad.

It also will not decide what to say. A model handed “write something for Tuesday” produces content-shaped filler. The idea has to be yours.

The job it is genuinely good at

Here is the work that actually eats your week: you have one idea, and it has to become six different posts, each obeying a different set of rules.

A LinkedIn post wants 300 words, a hook in the first 2 lines before the fold, no hashtag soup, and a link in the comments rather than the body. A TikTok caption wants 60 characters and lowercase. A Reddit post wants no marketing register at all and will be removed if it breaks the subreddit’s rules. An Instagram caption wants the first line to work as a preview. Threads wants 500 characters and tolerates a poll.

None of that is creative work. It is translation, and it takes about 40 minutes a time. That is the 40 minutes a model can genuinely take off your hands.

The difference between a useful AI feature and a gimmick is whether it knows those rules before it writes, or whether it writes something generic and then truncates it to fit.

Context is the whole game

The quality of an AI draft is almost entirely a function of what the model can see when it writes.

On a reply, it should have the whole thread. Including the nested replies you have not opened. A reply drafted from a single comment in isolation answers the wrong person half the time, because the actual question was asked 2 levels down.

On a Reddit post, it should have the subreddit’s rules. Fetched live, not remembered from training. Communities change their requirements, and a post written against last year’s rules gets removed.

On a review reply, it should have the star rating and the full text. A two-star review and a five-star review need completely different registers, and the specific complaint has to be named in the reply or it reads as a template.

On a caption, it should have the platform’s real limit. Not a guess. Writing 3,240 characters for LinkedIn and then cutting at 3,000 loses the ending you actually wanted.

This is the unglamorous part of AI product work, and it is the part that decides whether you edit the draft or delete it.

A working method

The teams getting real value from this all follow roughly the same shape.

Write the idea yourself, badly. 3 sentences in a notes app. What happened, why it matters, what you want someone to do. Do not polish it.

Let the model do the translation. One version per network, each against that network’s real constraints.

Edit the first line of every one. The hook is the part that decides whether anything else gets read, and it is the part a model is weakest at, because it does not know what will surprise your specific audience.

Kill one of them. If a version does not work for a platform, do not post it. Six mediocre posts is worse than four good ones and silence on 2 networks.

Alt text, quietly the best use

The AI feature nobody talks about is the one that writes alt text from the image itself.

Almost nobody writes alt text. It is tedious, it is invisible to the person writing it, and on a 10-slide carousel it is 10 separate fields. So it gets skipped, and every post becomes unreadable to anyone using a screen reader.

A model reads the image and produces a usable description in a second. It is not perfect and you should skim it. But the comparison is not “AI alt text versus great alt text”. It is “AI alt text versus none”, and none is what almost everyone ships.

What we do with your content

Worth stating plainly, because the question comes up and the answers in this industry are often vague.

When you press an AI button here, the relevant content is sent to our AI provider, Anthropic, to produce that one draft, and the result comes back to you. It is not used to train any model, by us or by them. It is not retained for training. It is not shared with other customers and never leaves your workspace.

Every provider that touches your data is named on our sub-processor page, with its location and the transfer mechanism we rely on. If a tool cannot tell you that in one sentence, ask why.

The honest position

AI in social media is neither the revolution the marketing pages describe nor the slop its critics assume. It is a very good translator and a mediocre writer.

Give it the translation job and it saves you most of an afternoon every week. Give it the thinking job and it produces exactly the kind of content that makes people scroll faster.

EngageDigital includes an AI assistant on every plan, with no second subscription and no API key of your own required. See what the assistant does or join the early-access list.