The Metrics That Actually Matter: How to Measure Social Media ROI the Right Way
Written byPorcha Kent • Head of Creative & Operations
Strategy - March 2025
The reason most brands cannot prove their social media works is that they are measuring things that were never connected to the business in the first place.
I have sat in a lot of quarterly reviews where a very confident deck reported that impressions were up 40%, engagement rate had improved, and follower growth was tracking to plan, while nobody in the room could say whether any of it had produced a single dollar. That is not a measurement problem. It is a strategy problem wearing a measurement costume.
Impressions and reach. These tell you how many times content appeared on a screen. They do not tell you whether anyone registered it, cared about it, or acted on it.
Reach is the easiest number to grow, which is precisely why it dominates reporting. You can buy it, you can manufacture it with bait, and it always goes up if you spend more. A metric that reliably improves regardless of whether the work is good is not a metric. It is a comfort.
Follower count. A vanity metric with an unusually strong grip on executives.
Followers are a stock, not a flow. A large disengaged following is worth less than a small active one, it costs nothing to accumulate through bad tactics, and it cannot be spent. The number of brands that have grown a following of a million people who buy nothing is considerably larger than anyone admits.
Engagement rate in isolation. Better than the previous two, and still unreliable on its own, because engagement without intent is entertainment.
Plenty of accounts have delightful engagement and sell nothing. Meme accounts run by brands are the clearest example: the content performs, the audience loves it, and the audience has no relationship whatsoever with the product.
Likes. The cheapest possible action a person can take, and therefore the least informative. Directionally useful in aggregate, individually meaningless.
None of these are useless. They are diagnostic, meaning they help explain why something happened. They are terrible as objectives, because optimizing directly for them produces content that generates the metric and nothing else, and it will do so with impressive efficiency right up until someone asks what it was all for.
Saves and shares. These cost the viewer something socially or cognitively, which makes them a far better proxy for genuine interest.
A share carries social risk: the person is putting their own credibility behind your content. A save is an admission that the content is worth returning to. Both are expensive actions in a way that a like is not, and expensive actions are informative. A high save rate on a product video usually means people are bookmarking it to buy later, which is about as close to purchase intent as an organic signal gets.
Click-through to owned properties. The moment someone leaves the platform for your site, they have expressed real interest and entered a place where you can actually measure them.
Track it with proper UTM parameters so you can see which content, which creator, and which format drove it. A brand that cannot answer "which post sent us traffic last month" has instrumented nothing.
Conversion rate from social traffic, and its cost. How many of those visitors did the thing you needed them to do, and what did it cost to get each one. This is the number that ends arguments and it is the number most social reports carefully avoid.
Assisted conversions. Social rarely closes the sale by itself, and last-click attribution will therefore make it look worthless.
This is the single most common way good social work gets defunded. The channel creates demand, someone searches the brand name a week later, search takes the credit, and the social budget gets cut. Look at the paths that touched social somewhere along the way, and be honest that this attribution is imperfect rather than pretending it is not happening.
Branded search volume. One of the most underused indicators available and it is free in Google Search Console.
If your social work is doing its job, more people should be typing your brand name into Google over time. That is demand you created. It appears nowhere in your platform dashboards, it moves on a lag, and it is one of the only accessible proxies for whether any of this is building anything.
Repeat purchase rate by acquisition channel. The one almost nobody tracks and the one most likely to overturn your entire investment case.
If customers who found you through creator content have a lower lifetime value than those from paid search, that changes everything about where the next dollar goes. If they have a higher one, you have just built the argument for tripling the budget, and you would never have known.
One primary metric per campaign, tied to what the campaign is actually for. We set it before launch and we resist the urge to grade the campaign on everything else.
An awareness campaign should be measured on reach quality and branded search lift, not on conversions it was never designed to produce. A conversion campaign should be measured on cost per acquisition and essentially nothing else. A community campaign should be measured on depth of interaction and returning audience.
Judging all three against the same scorecard is how good work gets killed and bad work survives, because the awareness campaign will always look worse on conversions and the conversion campaign will always look worse on engagement, and whichever one the reporting favors will get the budget regardless of what the business needed.
A reporting rhythm that a non-marketer can read. Weekly, the diagnostic numbers, so we can steer. Monthly, the business numbers, so we can decide. Quarterly, the strategic question, which is whether this channel deserves more, less, or different investment.
If your monthly report cannot be understood by the person who signs the checks, it is not a report. It is a defense.
The hardest discipline in this work is admitting what you cannot know.
Social attribution is genuinely difficult. People see a video on their phone, think about it for a week, mention it to a friend, and buy on a laptop with an ad blocker running. No dashboard captures that cleanly. Anyone who tells you their attribution model is airtight is either selling the model or has not looked closely at it.
We would rather tell a client that we are confident about the direction and uncertain about the magnitude than hand them a precise number we quietly know is fiction. Precision that is not accurate is worse than an honest range, because a false number gets planned against, and the plan fails for reasons nobody can trace back to the number.
The brands that measure well are the ones willing to hold that uncertainty and act anyway, using several imperfect signals that point the same direction rather than waiting for a certainty that does not exist in this medium and never will.
Porcha leads creative and operations at Proach Media, where she runs influencer programs, content strategy, and production for the agency's brand clients.