How to Tell If Your Social Media Agency Is Working.

The honest test of a social media agency is whether the constraint you hired them to fix has moved.

Not how many posts went out, and not whether the monthly report looks tidy. Volume and reporting are the two things easiest to produce without producing a result, which is exactly why a coasting retainer optimises for both.

It is harder to apply than it sounds. Most brands never wrote the constraint down, so six months in there is nothing to measure against and the conversation drifts back to output.

Start there. An agency that cannot name what it is fixing cannot show you that it is working.

Write down the constraint before you judge anything

Every engagement is bought to fix something specific: reach that stalled, a channel that never launched, a team drowning in production, a feed that looks fine and sells nothing.

If you cannot say in one sentence what you hired them to fix, you cannot fairly judge whether they fixed it. Neither can they.

Write it down now, even if you are eight months in, and ask the agency to write their version separately. If the two sentences do not match, you have found the problem before you have found a villain, and it is usually a briefing failure.

Whether to have an agency at all is a separate question, settled by the economics of agency versus in-house.

Volume is not a result, and a tidy deck is not either

A report that leads with what was made rather than what was decided is built to reassure, not to inform. Forty posts went out, here they are, here is the engagement on the best three.

The tell is whether anything in it changes what happens next month. If next month’s plan is last month’s plan with new dates, the reporting is decoration. Ask what the report caused.

Benchmark against the platform, not against last month

The month on month comparison is the most misleading line in agency reporting, and it misleads both ways.

Rival IQ’s 2025 benchmark study, which analysed more than four million posts across fourteen industries, found engagement rates fell on every major platform year on year. Facebook was down 36%, TikTok 34%, Instagram 16%, and X took the biggest hit at 48%.

Read that properly. On a platform where the median dropped by a third, an account that held flat had a good year, and an agency reporting a small decline was telling you the truth.

The gap between sectors is just as wide, with higher education topping Instagram and health and beauty at the bottom. Your benchmark is your sector on that platform, not your own last quarter.

The questions a good agency will enjoy answering

None of these is a trick. An agency doing real work answers them quickly, and has usually volunteered the answer already.

• What do you believe the constraint on this account is right now, and how has it changed?
• Which of the last ten things we published would you not publish again, and why?
• What is the median in our sector on this platform, and where do we sit against it?
• Which numbers on this report are the platform’s current metric, and which came from a template?
• What have you asked us for and not received?
• What did you try this quarter that did not work?
• Who actually does the work, day to day?

The fourth catches more agencies than you would expect. Instagram made views its primary metric across formats and Meta deprecated impressions in its API in April 2025, so a report still using the old labels is a template nobody has revisited.

What a weak answer sounds like

A weak agency answers a question about outcomes with a number about outputs. Ask what moved, and you are told how much shipped.

The patterns worth naming:

• The explanation for a good month and a bad month is the same explanation.
• Every problem is the algorithm, and no problem is a decision anyone made.
• Nobody can name a single thing the agency got wrong.
• The commentary in your report would fit any brand if you swapped the logo.
• You are always the first to notice something has gone wrong.

The last one matters most. An agency that brings bad news early is doing the job. One that waits to be asked is managing you, not the account.

How long to give it before you judge

Judgement has a floor. A new agency needs access, an audit, a plan and a stable rhythm before anything it does can be measured. That is a first month, not a first fortnight.

• 30 days: access sorted, an audit delivered, a plan you recognise, output going out reliably. Missing this is a process failure, raise it now.
• 90 days: the shape of the output has changed and they can tell you which formats work. Signal, not revenue.
• Six months: the constraint has moved, or there is a specific and credible reason it has not.
• Twelve months and beyond: compound results, and a system you can point at.

The good relationships run long. The 2025 tenure report from the 4As and the ANA put the average client and agency relationship at around seven years, more than double the 2016 figure. Churning every twelve months is its own kind of failure.

A bad month is not a bad engagement

Every account has bad months. A launch slipped, a format burned out, the whole market had a rough quarter. None of that is a reason to open a search.

The difference is diagnosis. In a bad month with a good agency the explanation is specific, it names a decision somebody made, and next month’s plan changes because of it.

In a bad engagement nothing is specific. The plan does not move, and the account has no memory of what has already been tried.

If you cannot tell which you are looking at, get someone outside the relationship to read the account. That is what an independent audit is for, and it is far cheaper than a search.

Some of it is genuinely not the agency’s fault

This is the part most brands skip, and skipping it is how you fire a competent team. Plenty of accounts underperform for reasons that sit entirely on the client side.

• A sign-off chain that adds a fortnight. Social is a timing business, and a post approved two weeks late is a different, worse post.
• A brand that will not approve anything with an edge on it. If every draft comes back sanded down, you are buying the safest version of the work, and safe is invisible.
• No raw material. No footage, no access to people, nobody willing to be on camera, no story anyone outside the building cares about.
• A moving brief. Awareness in January, leads in March, recruitment in June, and one judgement in September against all three.
• Being judged on work you did not buy. A production partner hired to make video is not failing because nobody set the channel strategy.

If two or more describe your side, the agency has not had a fair run, and changing agency will reproduce the same result with new faces.

The handover test

Here is the test we would be happy to have applied to us. If the relationship ended tomorrow with no ill will, what would you still have?

A real answer includes admin access in your own name, the raw files and not just the exports, documented content pillars, a calendar you could keep running, and a record of what has been tested.

A weak answer is a folder of finished videos and twelve monthly PDFs.

Be fair about the boundary. An agency is entitled to keep its own tooling and process, because that is its product. What you are owed is your assets, your access and your knowledge, and resistance there protects a retainer, not intellectual property.

What good looks like

A working relationship has a texture you can feel on the monthly call.

• The call opens with a decision to make, not a chart to admire.
• The plan changed since last month, and they can tell you what changed it.
• You know how much of the senior person you met is actually on the account.
• The list of things that are broken is shorter than it was six months ago.
• Approvals have a named owner and a deadline, so nothing sits waiting on a maybe. If yours do not, a simple approval process template fixes it.

None of that requires an expensive agency. It requires one running a system rather than filling a calendar.

If you decide to run a search anyway

Sometimes the answer after all of this is still to leave. Count the cost first.

The ANA and 4As study of the pitch process, which surveyed 329 brand and agency executives, found a review consumes roughly a business quarter: two months when the incumbent defends, more than three when a new agency is appointed. A third of marketers reported disruption to daily work, a quarter a delayed launch.

Those figures are from the large end of the market, but the shape holds at every size, and a new team needs ramp time on top.

Better to have the conversation first. Put the questions in writing, agree a window with a named outcome, and judge that. If you still leave, you leave with a written diagnosis, the most useful thing you can hand whoever comes next.

How NBK thinks about judging an agency

NBK is built around social operations, so the deliverable is a system: positioning, packaging, cadence, workflow, approvals and a reporting definition that tells you something. Content is the output of that system, not the product.

That belief has an uncomfortable consequence and we accept it. If the system is genuinely yours, you could take it in house tomorrow and keep it running. An agency whose value evaporates the moment it stops posting was never building anything.

Running always-on accounts past 600 million views a month teaches one thing quickly. Whether a month went well is decided by whether the constraint moved and whether anyone can say why, almost never by how many posts went out.

Next step

Take the constraint sentence and those seven questions into your next monthly call. Most of the time the answers are better than you feared, and the relationship gets sharper for having been asked.

If they are not, you have a specific list instead of a feeling, and that is worth more than a change of supplier.

If your social output feels busy but not effective, NBK can help find the constraint in the system.