Social media KPIs and what each metric actually proves
Most social reports are built from whatever exports cleanly. Here is which metric evidences which objective — and which evidence nothing.
A social report is easy to fill and hard to make honest.
Follower count exports in one click. Watch-through time does not. So reports drift towards easy figures that answer no question anyone asked.
This page does one thing: it ties every common metric to the objective it can legitimately evidence. For the strategy above the measurement, start with our guide to social media marketing. For KPIs across every channel, not just social, see marketing KPIs.
The three layers a social metric can belong to
Every social metric sits in one of three layers. Borrowing from one layer to prove something in another is where reporting goes wrong.
- Layer 1 — Reach / awareness. Did people see it, and how many distinct people? This layer answers distribution questions only.
- Layer 2 — Engagement / consideration. Did the people who saw it do something that costs them effort? Effort is the signal, not approval.
- Layer 3 — Action / conversion. Did they take a step towards you — a profile visit, a link click, a DM?
The rule: a Layer 1 metric cannot evidence a Layer 2 objective. Reach going up does not mean the content got better — usually it means the platform pushed it to more people. If the objective is brand awareness, reach is the right KPI. If the objective is consideration, reach is only context.
Three denominators, three rates.
The social media KPI table
The part worth saving. Each row names what the metric genuinely evidences, what it cannot, and which layer it belongs to.
| Metric | What it genuinely evidences | What it does not evidence | Layer |
|---|---|---|---|
| Reach (accounts reached) | How many distinct accounts saw it at least once | Attention, interest or audience quality | 1 — Awareness |
| Views | Total delivery volume, repeats included | How many people — one account views many times | 1 — Awareness |
| Non-follower share of reach | Whether distribution is escaping your existing audience | Whether those new people liked it | 1 — Awareness |
| Follower count | Cumulative history of past acquisition | Current reach, demand, or anything this month | None |
| Follower growth rate | Whether the content is still recruiting new people | Revenue, intent, or audience relevance | 1 → 2 |
| Likes | Low-cost approval — the cheapest action available | Consideration, recall or intent | 2 — weakest |
| Comments | Enough interest to stop and type something | Sentiment, unless you actually read them | 2 — Consideration |
| Saves | Intent to come back to this content later | Intent to buy anything | 2 — Consideration |
| Shares | Willingness to attach your content to their own name | The reach that resulted, unless measured separately | 2 — Consideration |
| Watch-through / avg watch time | Whether the content held attention past the hook | Whether the message landed | 2 — Consideration |
| Profile visits | Curiosity converted into one step towards you | Interest in your offer specifically | 2 → 3 |
| Bio-link / post link clicks | Willingness to leave the platform for you | A lead, a sale, or a qualified prospect | 3 — Action |
| DMs and replies started | The closest organic social gets to an inbound lead | Qualification — most are not buyers | 3 — Action |
| Impressions | Little that reach does not say better | People reached; Meta deprecated it in its API | Retire it |
On that last row: Meta's Instagram Platform changelog deprecated impressions in its 21 January 2025 entry, across all API versions from 21 April 2025, and introduced views instead. A report still leading with impressions is older than the platform.
The metrics worth more than they get
Four metrics do real work and get a fraction of the attention likes get.
- Saves and shares over likes. A like costs a thumb twitch. A save is someone saying "I will need this". Instagram's own media insights reference counts total_interactions as likes, saves, comments and shares "minus the number of unlikes, unsaves, and deleted comments" — even the platform treats these as revocable actions of unequal weight.
- Non-follower reach. If a month of posting reached almost nobody new, you entertained your existing audience. A real outcome, but not growth.
- Watch-through. The best proxy for whether a video idea worked. Reach says the platform pushed it; watch-through says people stayed.
- Profile-visit-to-link-click rate. Two numbers you already have, divided. It isolates whether your bio converts curiosity, separately from whether your posts earn it.
The same reference exposes a profile_activity breakdown by action type including BIO_LINK_CLICKED, and a Story navigation breakdown including TAP_BACK — someone going back to re-read a slide. Both are more diagnostic than anything on the front page of most reports.
The metrics that mislead
Not wrong metrics. Metrics routinely asked to prove things they cannot.
| The metric | The claim it gets used for | Why it fails |
|---|---|---|
| Follower count | "The account is growing / healthy" | A running total of every follow ever, including people who stopped looking years ago |
| Views without reach | "We reached a lot of people" | Views count repeats, reach counts people. Quoting only views hides the ratio |
| Engagement rate on followers | "Our content resonates" | If most reach came from non-followers, the denominator does not describe who saw it |
| Best time to post | "Post at 7pm for more reach" | Derived from when existing followers are active, which does not govern non-follower distribution |
| Total engagements (raw count) | "Engagement is up 40%" | Posting more produces more engagements. Without a rate, volume says nothing |
Reach needs a caveat too. Instagram's Help Centre defines accounts reached as the "Number of unique accounts that have seen your content on screen at least once", then states plainly that reach and engagement metrics are "estimated and in-development". Treat it as a direction of travel, not a headcount.
Engagement rate is not one metric
The correction that changes most reports. "Engagement rate" names at least three calculations, and the denominator changes the answer completely.
| Variant | Formula | Honest use |
|---|---|---|
| ER by followers | Interactions ÷ followers | Comparing your own posts while follower count is stable. Most tools default to it |
| ER by reach | Interactions ÷ accounts reached | The fairest read of "did the people who saw this respond". Reach is estimated |
| ER by views | Interactions ÷ views | Platform-native on LinkedIn. Always the lowest of the three, because views ≥ reach |
Working through the maths. Take one hypothetical post: 100 interactions, 4,000 followers, 10,000 accounts reached, 14,000 views.
ER by followers = 100 / 4,000 = 2.50% ER by reach = 100 / 10,000 = 1.00% ER by views = 100 / 14,000 = 0.71%
Same post, same interactions, a 3.5x spread. Which is why comparing two brands' engagement rates is usually meaningless — you rarely know both used the same denominator, and the platforms do not agree on one.
LinkedIn's API documentation defines its engagement field as the "Number of organic clicks, likes, comments, and shares over impressions" — it counts clicks and divides by impressions. Instagram's total_interactions counts no clicks and nets out unlikes. Not one metric with two names. Two metrics sharing one name.
So fix the denominator once, write it on the report, and never change it mid-year. Our engagement rate calculator runs all three variants side by side on your own numbers. Platform differences sit in the Instagram and LinkedIn sections of the pillar guide.
Why we will not quote a "good" engagement rate
Published benchmarks are everywhere. We are not repeating them. Four reasons.
- Most are vendor blog posts sampling their own customer base — brands that bought social software, not a neutral sample of Indian businesses.
- They rarely state the denominator, which swings the number by multiples.
- They rarely state the follower band. Small accounts post higher rates almost mechanically.
- They go stale. A benchmark computed before Meta retired impressions measures something the platform no longer reports.
Compare against your own trailing trend instead. Your last 12 posts, same denominator, same platform, same format — valid, available and actionable. An external benchmark is none of the three.
What a monthly social report should actually contain
One page per platform, in layer order, with a reason against every number.
- Objective for the month — one line, stated first, so every metric below has something to be judged against.
- Layer 1: reach, non-follower share of reach, net follower change.
- Layer 2: engagement rate (one named denominator), saves, shares, watch-through.
- Layer 3: profile visits, link clicks, DMs started, and the profile-visit-to-click rate.
- Top three and bottom three posts — with a sentence on why.
- What we are changing next month, and what we got wrong.
Once the inputs are stable, build it in a tool rather than a slide — Looker Studio handles this well. Inheriting an account? Run a social media audit before setting any KPI at all.
Social and last-click: why attribution under-credits it
Social rarely shows up in last-click reporting. That is a measurement limit, not evidence it did nothing.
Google Analytics assigns traffic to Organic Social only when the source "matches a regex list of social sites" or the medium is one of a short list such as social or social-media, per Google's default channel group documentation. Visits arriving without that referrer data fall into Direct, or Unassigned when no rule matches.
Plenty of genuine social-driven visits arrive that way: a link copied into a messaging app and reopened later, a brand name typed into Google after someone saw a Reel, an in-app browser that strips the referrer. None of it is attributable. All of it is real.
So report Layer 3 as a directional contribution, not the full value. Pair it with branded search volume and with asking new enquiries where they first heard of you. The structural version of this sits in marketing attribution.
What this page does not cover
Narrow by design, so each page owns one job:
- All-channel KPIs across SEO, email and paid: marketing KPIs.
- Auditing an existing presence: the audit process.
- How Instagram decides what to distribute: the Instagram algorithm.
- Measuring conversation about you, not your posts: social listening.
- Building the report view: dashboards.
Two posts are in production on employee advocacy and podcast marketing, both with their own quirks — advocacy reach lands on individuals rather than the brand page, and podcast listens are not comparable to views at all.
Measuring less, and meaning it
A short report everyone in the room trusts beats a complete one nobody reads.
Pick the objective. Pick the layer it lives in. Report the two or three metrics belonging to that layer, on one fixed denominator, against your own trailing trend. Everything else is appendix.
Then the monthly review stops being a defence of numbers and becomes a decision about content — the only thing a KPI was ever for.
Frequently asked questions
What are the most important social media KPIs?
The ones that match your objective. For awareness, reach and non-follower share of reach. For consideration, engagement rate on a fixed denominator plus saves, shares and watch-through. For action, profile visits, link clicks and DMs started. Follower count is not on the list.
Is follower count ever a useful KPI?
Rarely as a KPI, occasionally as context. Net follower change over a month tells you whether content is still recruiting. The cumulative total tells you about acquisition that happened years ago and says nothing about whether anyone saw your work this month.
What is a good social media engagement rate?
There is no figure worth quoting. Published benchmarks usually come from vendor customer bases, rarely state the denominator and rarely state the follower band. Compare each post against your own last 12 posts on the same platform, format and denominator instead.
Should engagement rate be calculated on followers or reach?
Reach is the fairer denominator, because it describes the people who actually saw the post. Followers is the common tool default. Either is defensible. What is not defensible is changing the denominator between reports, or comparing your rate to a brand using a different one.
Why does social media show so little traffic in Google Analytics?
Because Analytics needs referrer data to classify a visit as Organic Social. Links shared through messaging apps, in-app browsers that strip the referrer, and people who search your brand name later all land under Direct or Unassigned instead. The visits are real; the attribution is missing.
Social media that reports on objectives, not vanity
We run organic social for brands across India with a reason attached to every number.
Make Digital Hangover a preferred source
One tap tells Google to show more of our SEO and marketing coverage in your Top Stories.