← Guides

How to automate weekly social media reporting

The short answer

Automate a weekly social report in five steps: pull each platform’s numbers through its API on a schedule, save every pull, compare it with the week before, have an AI model draft the commentary, and send the draft to one person to approve. Run it at least 48 hours after the week ends, because Instagram’s own data can lag by that long.

Banner reading “The weekly report that writes itself.” beside a marble statue in an orange suit reclining next to a typewriter, with Instagram, YouTube, Facebook and LinkedIn logo tiles.

Scope

What can be automated, and what should stay with a person?

The collecting, the comparing and the first draft can be automated. The judgement cannot. A weekly report is made of three kinds of work, and only two of them are mechanical.

  • Collecting is opening each platform, exporting numbers and pasting them into a sheet. A scheduled job does this without errors or holidays.
  • Comparing is working out what moved since last week. That is arithmetic once the numbers are saved in one place.
  • Deciding is saying what the team should do differently next week. A model can draft it. A person has to own it.

Most teams have not got this far. In Supermetrics’ 2026 Marketing Data Report, a survey of 435 marketers, 39% said they use AI for reporting and analytics, and 6% said AI is fully built into how they work.

The data

What does each platform’s API give you?

Less than the app shows you, and with rules attached. Every platform decides which numbers it hands over, how fresh they are and who is allowed to ask. Check those rules before you promise anyone a metric.

Platform
Instagram
You can get
  • Account: reach, views, likes, comments, shares, saves, follows and unfollows
  • Each post: reach, views, likes, comments, shares, saves
  • Reels: average watch time and skip rate
Limits
  • Data can be up to 48 hours behind
  • Story numbers are only available for 24 hours
  • Kept for up to 2 years
  • “Impressions” was removed on 21 April 2025; use “views”
Access

A professional account and an app with the insights permissions.

Facebook Pages
You can get
  • Page and post views
  • Engagement on each post
  • Audience breakdowns for 100 people or more
Limits
  • Only for Pages with 100 or more likes
  • 90 days per request, two years in total
  • Most numbers update once every 24 hours
  • Several older metrics were retired by 15 June 2026
Access

A Page token with the read insights permission.

LinkedIn
You can get
  • Company page and post statistics, once your app is approved
Limits
  • Access is for registered legal organisations only
  • Two review stages: Development, then Standard
  • Standard needs a written use case, test logins and a screen recording
Access

An approved Community Management app, verified by a Page super admin.

YouTube
You can get
  • Views and estimated minutes watched
  • Average view duration
  • Subscribers gained
Limits
  • Two APIs: Analytics for specific questions, Reporting for bulk exports
  • Each query is charged against a quota
Access

Sign-in by the channel or content owner.

X
You can get
  • Post and account data through the X API
Limits
  • Priced per request, with no subscription
  • Budget for it before you add it to a weekly job
Access

A developer account with credits.

From each platform’s developer documentation, read on 11 November 2026. Pick a platform. TikTok is left out because we could not confirm its current rules at the source.

Two of these rules shape the whole job. Instagram’s data can be 48 hours behind, which sets the day you run the report. LinkedIn only gives access to registered organisations after a review, so start that application first, because you cannot build around it.

The job

How does the job work, step by step?

Five steps, run in order, once a week. We run this for Instagram Reels on several accounts, and the notes under each step are how our own build does it.

Step
1. Fetch
What happens
A scheduled job asks each platform’s API for last week’s account and post numbers.
In our build
One script per platform. It asks only for the metrics the report uses, so a renamed metric fails loudly in one place.
Stays with a person
Choosing which accounts and which metrics matter.
2. Store
What happens
Every run is saved as it arrived, with the date it was pulled.
In our build
Each run adds one line to a history file per account. Nothing is overwritten, so any past week can be rebuilt.
Stays with a person
Nothing. This step should never need a person.
3. Compare
What happens
This week is set against last week and against the account’s own average.
In our build
Week on week is a comparison of two lines in the history file. Posts well above the account’s average are flagged.
Stays with a person
Deciding what counts as a good week for this account.
4. Draft
What happens
An AI model reads the numbers and writes the commentary: what worked, what to fix, what to try next.
In our build
The model gets the numbers and the brand notes, and nothing else. If the model call fails, the report still goes out with numbers only.
Stays with a person
Checking the commentary says what the numbers support.
5. Approve and send
What happens
One person reads the draft, edits it if needed and approves it. Then it is delivered.
In our build
The report waits for a named approver. Nothing reaches a client or a channel without that click.
Stays with a person
The approval, every week.
Pick a step to see what happens, how we do it and what a person keeps.

Three choices in that build have saved us the most trouble:

  • Save every run and never overwrite. When a platform changes a metric, you can still rebuild any past week from what you saved.
  • Let the drafting step fail. A model call can time out. The report should still arrive with its numbers, and say plainly that the commentary is missing.
  • Give the model the numbers and nothing else. Commentary that can only draw on this week’s figures and your brand notes has less room to invent a reason.
Version
Weekly social reportInstagram, 2 to 8 November
Waiting for approval
  • Views48,210+12% on last week
  • Reach31,940+7% on last week
  • Saves612+31% on last week
  • Follows184-9% on last week
Flagged post

Tuesday’s Reel reached 2.4 times the account’s average and was saved 203 times.

Commentary, drafted
  • Saves rose faster than views, which points to the how-to Reel on Tuesday.
  • Follows fell while reach grew. The week’s posts reached new people but gave them no reason to follow.
  • Next week: make a second how-to in the same format and end it with a reason to follow.
Commentary

The drafting step did not run this week. The numbers above are complete.

What reaches the approver. Example figures, not a client account. Switch to “Numbers only” to see the report on a week when the drafting step fails.

Timing

When should the report run?

Wednesday morning, for the week that ended on Sunday. Meta’s documentation says the data behind Instagram’s metrics may be delayed by up to 48 hours, and that most Facebook Page metrics update once every 24 hours. A report built on Monday morning is built on numbers that are still moving.

Stories need their own schedule. Meta makes Story metrics available for 24 hours, so a weekly job never sees most of them. Run a small job every day that saves yesterday’s Stories, and let the weekly report read from what it saved.

Failure

What breaks, and how do you catch it?

Platforms change their APIs more often than teams check their reports. Each of these has happened, and each has a cheap check.

Dates and limits are from Meta’s developer documentation, read on 11 November 2026.
What breaksWhat it looks likeThe check
A metric is renamed or removedInstagram dropped “impressions” on 21 April 2025. Facebook retired several Page metrics by 15 June 2026.Fail the run when a requested metric is missing. Never fill the gap with zero.
Missing data looks like zeroInstagram returns an empty set, not a 0, when it has no data.Treat empty as “no data” and show it that way in the report.
The login expiresAn expiring system user token from Meta lasts 60 days.Refresh it on a schedule, and alert a person a week before it runs out.
Paid and organic get mixedInstagram’s “likes” and “views” are organic. Its “total” versions include boosted posts and ads.Pick one and label it in the report.
The API version is retiredMeta’s Marketing API v24.0 was available until 10 November 2026.Put each version’s end date in the calendar when you adopt it.

The last line of defence is the approver. A person who reads the report every week notices when reach halves for no reason. Give them a way to say “this looks wrong” that stops the report going out.

The choice

Should you build this or buy a reporting tool?

Buy if a standard dashboard is enough. Build if you need your own definitions, written commentary or an approval step. Reporting tools already hold the platform connections and keep them working, which is most of the maintenance.

Listed prices on each vendor’s own pricing page, checked on 11 November 2026. They change often.
ToolPlanListed priceWhat the plan includes
SupermetricsStarter$59 a month, or $44 billed yearly2 data sources, 1 user, 1 destination
SwydoSingle plan$69 a month, or $62 billed yearly10 data sources, unlimited users and reports
Sprout SocialStandard$199 a seat each month, billed yearly5 social profiles
HootsuiteProfessional$199 a user each month, billed yearlyUnlimited accounts, custom and scheduled reports
DataboxBusiness$399 a month billed yearly, or $499 monthlyUp to 30 data sources and 10 users

A tool gets you charts on a schedule. It will not write commentary in your voice, hold a report until someone approves it, or apply your own rule for what counts as a good post. Those are the reasons to build.

If you build, start small: one platform, six metrics, one person approving, one place the report lands. Add the second platform when the first has run clean for a month. Ourweekly report workflow shows one run from source checks to the decision at the end, and reporting is one of the things we set up in anautomation build.

Questions

Questions people ask

Can you automate social media reporting?
Yes. Collecting the numbers, comparing them with last week and drafting the commentary can all run on a schedule. Deciding what the numbers mean for next week, and approving the report before it goes out, should stay with a person.
What day should a weekly social media report run?
At least two days after the week ends. Meta’s documentation says Instagram insights data can be delayed by up to 48 hours, so a report for Monday to Sunday is safest on Wednesday morning.
Why does my automated report show zero impressions for Instagram?
Because the metric no longer exists. Meta removed “impressions” from the Instagram insights API for all versions on 21 April 2025 and replaced it with “views”. A report that still asks for impressions gets an error or nothing back.
Can I pull Instagram Story numbers into a weekly report?
Only if you collect them daily. Meta makes Story metrics available for 24 hours, so a job that runs once a week misses six days of Stories. Run a small daily job that saves them.
Do I need a developer to automate social reports?
Not for a standard dashboard. Tools such as Supermetrics, Swydo and Sprout Social connect to the platforms for you. You need someone technical when you want your own metric definitions, written commentary or an approval step.
How much does a social media reporting tool cost?
On 11 November 2026, Supermetrics listed its Starter plan at $59 a month, Swydo listed $69 a month, and Sprout Social listed its Standard plan at $199 a seat each month on annual billing.
Can AI write the commentary in a social media report?
It can write a first draft from the numbers. Give it only the figures and your brand notes, and have a person check that each sentence is supported by a number before the report is sent.

Free audit

Want this worked out for your team?

We look at how your marketing work runs today and show you what to build first, what it would cost and what to leave alone. The audit is free.

Keep reading

All media →