Social media ROI, the four numbers that make the calculation honest and the engagement figure worth taking out of the report

Social media return on investment is the revenue you can attribute to social activity, minus the full cost of producing and distributing it, divided by that cost. The reason most reported figures are useless is the cost side rather than the revenue side: the labour, the production and the paid media are three separate costs and most reports include only the third. Fix the cost side first and the calculation becomes something a finance reader will accept.

Measured against the India geo target alone on 22 October 2026, social media roi draws 70 searches a month in India, and against the United Kingdom target alone on the same day it draws 50 a month in the United Kingdom. Two readings of two countries, kept separate.

The four numbers, and where each one hides

1. Labour. Hours spent planning, making, publishing and replying, times a loaded hourly cost. This is the largest cost in almost every social programme and the one almost never counted, because it is already in a salary.
2. Production. Shoots, editing, design, music licences, props, anything bought to make the thing that gets published.
3. Media. What the platforms charged. The only one most reports include.
4. Attributed revenue. Revenue you can trace, over a window long enough to contain your actual sales cycle.

Run it with all four and a programme that looked profitable on media alone often turns out to be break even, which is useful rather than disappointing: it tells you the cheapest improvement available is to the production cost, not to the targeting.

The attribution window is the decision, not a detail

A social post read in March that contributes to a purchase in June shows up in a thirty day window as nothing. Set the window to match your own sales cycle, measured from your own records, and state the window every time you state the figure. A return percentage without its window is not comparable to anything, including last quarter version of itself.

Where the attribution model itself is concerned, the honest approach for most businesses is to stop trying to split credit between touchpoints and instead track one thing well: ask on your enquiry form where somebody heard about you. The self reported answer is imperfect and it is far better than a model nobody in the room understands.

The number to stop reporting

Engagement rate. It is a ratio of reactions to reach, it moves when the platform changes how it distributes, and it has no denominator in money. Two further problems make it actively misleading in a board pack. A post can earn a high engagement rate from exactly the wrong audience, and a small account will routinely show a better rate than a large one, which makes any comparison between them meaningless.

Keep it as a diagnostic for creative. Use it to decide which of two openings worked. Do not put it next to a revenue line.

What to report instead, in order of usefulness

The underlying arithmetic is the same calculation every other channel is judged on, and getting the finance accepted version right once makes every channel comparable. The marketing return on investment formula, in the version finance accepts is worth agreeing with your finance team before you report a social figure, because the argument you avoid is about the formula rather than about social.

A worked example, with made up but internally consistent figures

Take a quarter where a team spends 60 hours a month on social at a loaded cost of a hundred units an hour, buys two thousand units of production a month and spends five thousand units a month on paid media. The quarterly cost is 18,000 for labour, 6,000 for production and 15,000 for media, so 39,000 in total. If attributed revenue over a window matching the sales cycle is 70,000, the return is 31,000 on a 39,000 cost, which is about 79 per cent.

Report it on media alone and the same quarter reads as 70,000 on 15,000, which is 367 per cent. Both numbers are arithmetically correct. Only one of them would survive somebody asking what the team was paid.

Engagement, when it does matter

There is a real mechanism behind the metric, which is why people cling to it. Replies and shares extend distribution, and distribution is free reach. So the useful version of the question is not how to raise the engagement rate, it is how to raise replies from people who could buy. Three things do that reliably: asking a specific question rather than a rhetorical one, taking a position somebody could disagree with, and answering every reply within the hour.

Questions about calculating social media return

What if we cannot attribute any revenue to social at all

Then report cost per qualified conversation and the share of customers who mention social on your enquiry form. Both are real readings, both are cheap, and both are better than a return figure built on an attribution model you cannot defend.

Should organic and paid social be measured together

Separately, then together. They have different cost structures and different decay, so a combined figure hides which half is working. Report both and the total.

How long before a social programme shows a return

Judge paid inside a month, because it is a media buy. Judge organic over two quarters, because its output accumulates and its early months carry the setup cost of finding the format that works for you.

Is follower growth worth tracking at all

As a diagnostic for whether your subject matter is landing, yes. As a reported outcome, no, because followers are not a cost and not a revenue and the number can be bought.

How do we count labour when social is part of somebody job

Ask them to log hours for two typical weeks and use the average. It is not perfect and it is accurate enough to change the decision, which is the only standard that matters here.

The return nobody has a line for

Every figure above assumes the chain runs post, click, visit, purchase. A growing number of buying decisions skip all four: somebody asks an assistant who to use, gets two names, and contacts one of them directly with no social touch to attribute. That arrives in your figures as a customer who heard about you from nowhere. At AI Knows Us we measure whether the five assistants name a business when that question is asked, dated each time, which turns the unattributable part of the chart into something you can actually read.