Open the analytics tab on almost any platform and you are handed thirty numbers. Reach, impressions, views, average view duration, engagement rate, saves, shares, link clicks, profile visits, new follows, and a dozen more, each with its own little sparkline. The dashboard treats them as equals. They are not.
Most of those numbers cannot change a decision you are about to make. A handful can. The skill isn't collecting more metrics. It's knowing which two or three answer the question you actually have, and ignoring the rest without guilt.
This is the map. It doesn't re-argue the deeper points made elsewhere in the library: how to tell a counted number from an asserted one is covered in measured vs reported, and how to write a goal your data can serve is covered in setting a content goal. This piece sits on top of those: given a goal, which metric tells you whether you are getting there.
Every metric answers one question
A metric is only useful as the answer to a question. If you can't name the question, the number is decoration. Here is what each of the common ones actually answers (and, just as important, what it doesn't).
Before the list, one caution that runs through all of it: the definitions differ by platform. A "view" on one network is three seconds; on another it's the moment the video renders on screen; on a third it counts a loop. "Reach" and "impressions" are split differently on each. So the words below describe what each metric is for, not its exact definition on your platform. For that, always check the platform's own glossary (YouTube's Help Center, Meta's Business Help Center, TikTok's Support Centre, LinkedIn's Help), because the counting rule decides what the number can honestly be compared against.
Reach answers how many distinct people saw this at all. It's a headcount of unique accounts. It's the honest denominator for awareness work, and it's the number most worth having when your goal is simply to be in front of a group of people who weren't before.
Impressions answer how many times it was served, including repeat views by the same person. Impressions are always equal to or larger than reach. On their own they mostly tell you about frequency, not audience size. Watch the gap between the two: impressions far above reach means the same people are seeing it repeatedly, which is either good (a campaign building memory) or bad (a feed showing it to a small pool over and over) depending on your goal.
Views answer how many people started watching. Useful, but the weakest of the video numbers, because "started" is a low bar and the counting threshold is short. A high view count with nothing behind it is the classic vanity result.
Watch-through (average view duration, or the percentage who reach a given point) answers did the thing actually hold attention. This is the one that separates a video people saw from a video people watched. For any video goal, watch-through is worth more than raw views, because it's much harder to inflate and it reflects the content rather than the packaging.
Engagement rate answers of the people who saw it, how many did something. The word "rate" matters: it's actions divided by a base, which is what makes it comparable across posts of different sizes. But engagement rate is a family, not a metric: likes, comments and shares are not the same signal, and lumping them together hides which one moved. It's also defined against different denominators on different platforms (over reach on one, over followers on another), so an engagement rate is only meaningful compared to itself on the same platform.
Saves and shares answer was this worth keeping or passing on. These are the quiet high-value actions. A like costs nothing; a save is someone saying "I want this later" and a share is someone spending their own credibility to send it to somebody else. When saves or shares run high relative to likes, you've made something useful rather than merely agreeable, and that's usually the more repeatable win.
Link clicks answer did anyone want to leave the platform for you. The first genuinely commercial signal in the list. Clicks are where interest becomes intent. They're also where the funnel narrows hard, so they'll always be a small number, which makes them easy to over-read on a single post (more on that trap in the companion piece below).
Follows answer did this earn a longer relationship. New follows attributable to a post tell you it did more than perform once: it made someone opt in to the next one. Follower count as a standing total is close to pure vanity; new follows from a specific piece are a real signal about what earns you an audience.
Vanity or decision: the test
The label "vanity metric" gets thrown at whole categories (followers bad, engagement good), and that's too crude. Any metric is a vanity metric when it can't change what you do next. The test is a single question:
Follower count usually fails this test: it drifts up regardless, and no realistic reading of it changes next week's plan. Watch-through usually passes it: a clear drop-off point tells you exactly where to cut. Impressions with no reach or click context usually fail. Saves-per-view usually passes. The metric isn't inherently vain or virtuous. Its status depends entirely on whether a plausible change in it would move your hand.
The map: which metrics for which goal
Pick the goal that is actually binding this quarter (not all of them, as setting a content goal argues at length), then watch the two or three metrics that serve it. Everything else becomes context you glance at, not a scoreboard you answer to.
| If the goal is… | The metric that answers it | A second, to catch a false positive | Safe to ignore for now |
|---|---|---|---|
| Awareness: be seen by new people | Reach | New follows (did any of it stick) | Impressions, likes |
| Attention: make what's seen land | Watch-through, or saves-per-view | Engagement rate | Raw views, reach |
| Consideration: earn a relationship | New follows from the piece | Saves and shares | Impressions |
| Action: get people to move | Link clicks | Click-to-reach ratio | Likes, follower count |
| Learning: find what works | Your own outlier multiple | Engagement per view | Absolute view counts |
Two things to notice. First, every row has a second metric whose job is to catch a false positive: reach that recruited nobody, clicks from a tiny reach, views nobody finished. A single number is almost always game-able; a pair is much harder to fool. Second, the last row (learning) isn't about a threshold at all. It's about comparison against your own baseline, which is a different discipline: outliers, not virality covers how to find the posts that beat your own norm and read what they had in common.
Compare like with like
A metric is only interpretable against the right comparison, and the right comparison is almost always you, on the same platform, over time (not a rival, and never a blend of platforms). Why a cross-platform average is a number that measures nothing is set out in same-platform benchmarking; the short version is that a view here and a view there are different events, and averaging them produces confidence without information.
So: fix the platform, fix the format, and read the metric as a trend against your own recent median. "Our watch-through on Reels went from the low forties to the mid-fifties over the quarter" is a sentence you can act on. "Our engagement rate is 4.2%" (across everything, against nobody) is not.
Where Acumin fits
The two or three numbers you settle on still have to come from somewhere honest. A Snapshot pulls your top performers by format from real public data, and Content Analysis ranks them against your own baseline rather than against the internet, which is the comparison this whole piece argues for. The cockpit is where the goal you set is scored against those numbers over time.
None of that is the only way to do it. A spreadsheet with your last thirty posts, one platform per tab, the median of each metric, and a column for the multiple, does the core job perfectly well, and building it once teaches you more about your own numbers than any dashboard will. Acumin earns its place by doing that continuously and refusing to invent the figures; it doesn't replace knowing what you're looking at.
How to use this today
Write down the one goal that's binding this quarter. Find its row in the table above. Then open your analytics and hide, mute or simply stop looking at every metric that isn't the two named in that row.
That's the whole exercise. Not more numbers: fewer, chosen on purpose. When you can state your two metrics and the question each one answers, you'll make faster decisions than the team staring at all thirty, because they can't tell which of the thirty is talking.
Then read reading your numbers without fooling yourself before you draw a conclusion from either of them, because picking the right metric is only half the job. The other half is not being tricked by it.