Most brands measure social video with the number that is easiest to find rather than the one that means anything. Views are reported, screenshots are circulated, and nobody can say whether the money did any work. The measurement problem is real but it is not hard, and solving it changes what gets commissioned next. This article looks at how to tell whether your social video worked, which numbers are worth watching, and which are comfortable noise.
Decide the Job Before Measuring the Result
A video cannot be judged without knowing what it was for, and a surprising number are commissioned without that being written down.
The four honest jobs for social video are: making people aware you exist, making them understand what you do, making them feel something about you, and making them act. Each has a different success signal, and a video built for one judged by another’s metric will always look like a failure.
Write the job down before production. It takes a sentence, it improves the brief, and it makes the post campaign conversation about evidence rather than opinion.
Why View Count Tells You Almost Nothing
Views are the number everybody reports and the number that carries least information.
Platforms count views differently, and most count generously. A view may be registered after one second, or after three, or when a video occupies half the screen momentarily. A million such views can represent almost no attention at all.
Views are also the number most easily inflated by paid distribution, which means a well funded piece will outperform a better one that ran organically. Comparing view counts across pieces with different media budgets compares budgets rather than content.
Views are not useless. They tell you reach was achieved, which matters for an awareness objective. They simply cannot tell you whether anything happened as a result.
The Numbers That Actually Signal Attention
- Average watch time and completion rate. The single most useful pair. They tell you whether the content held people, independent of how many arrived.
- Retention curve shape. More useful than the average. Where viewers leave tells you which second failed, and that is directly actionable in the next edit.
- Three second and fifteen second retention. On short form vertical, the first three seconds decide everything. A piece losing half its audience there has a hook problem, not a content problem.
- Rewatches and loops. A strong signal on short form, indicating either genuine interest or that the piece was too dense to absorb once.
- Saves and shares. Considerably more meaningful than likes, because both cost the viewer something. A save says useful; a share says I want to be associated with this.
- Profile visits and follows from the piece. The clearest evidence that a viewer moved from content to brand.
Reading the Retention Curve
If you take one habit from this article, make it looking at the retention graph rather than the summary numbers.
A steep initial drop followed by a stable line means the hook was wrong for the audience it reached, but the content held those who stayed. Fix the opening, keep the body.
A gradual even decline is normal and healthy. It means the piece is losing people at the rate any content does, with no specific failure.
A sudden cliff partway through indicates a specific moment that lost people. Watch that second. It is usually a pace change, an unexpected shift in tone, a logo appearing too early, or the point where the piece starts selling rather than giving.
A rise anywhere is worth studying carefully, because it means people scrolled back. Whatever happened there is the strongest material you have, and it belongs earlier next time.
Measuring Against the Job
Match the metric to the objective rather than reporting everything.
For awareness, the relevant numbers are reach, unique viewers and follower growth. Watch time matters less because the goal is exposure.
For understanding, watch completion and any measurable reduction in confusion downstream: fewer basic questions in comments, fewer support enquiries about the thing explained, more informed sales conversations.
For feeling, the signals are qualitative and still real. Comment sentiment, unprompted shares, and whether people describe the brand differently afterwards. Do not force a number onto this; read the comments properly instead.
For action, everything else is secondary to clicks, enquiries and conversions attributable to the piece. A video with mediocre retention that drives enquiries has worked, whatever the engagement numbers say.
Attribution Without Pretending It Is Precise
Connecting video to business outcomes is genuinely difficult and worth attempting imperfectly rather than not at all.
Use distinct tracking links per piece and per platform. It costs nothing and separates traffic that would otherwise be indistinguishable.
Ask on the enquiry form. A simple how did you hear about us field, with the options kept short, produces surprisingly usable data over a quarter.
Watch the shape rather than the point. If enquiries rise in the fortnight after a piece runs and fall afterwards, the piece contributed even if nothing can be attributed line by line.
Accept that view through effects are real and unmeasurable. Somebody who saw a video, did nothing, and enquired three weeks later after a search will attribute to search. That does not mean the video did nothing.
Benchmarks Worth Using
Absolute numbers mean little without a comparison, and the useful comparison is almost always yourself.
Compare against your own previous pieces on the same platform with similar distribution. This controls for the variables that make cross brand comparison meaningless.
Where you do look outward, compare against similar businesses of similar size rather than against category leaders whose budgets and audiences are not yours.
Build a simple running record: piece, platform, objective, spend, watch time, completion, saves, enquiries. After six months this small table is more valuable than any platform dashboard, because it shows what works for you specifically.
Give It Enough Time and Enough Data
Two common errors distort measurement in opposite directions.
Judging too early is the more common. Social content has a longer tail than it used to, and short form in particular can find an audience days or weeks after posting. A piece assessed after twenty four hours is being assessed on distribution rather than quality.
Judging on too little is the other. A single piece is one data point, and social performance is noisy. Three pieces of the same type give a far more reliable read than one, which is an argument for producing in batches rather than one at a time.
Where a piece genuinely underperforms, check the obvious before concluding the content failed. Was it posted at a sensible time, with a working caption, at the right ratio, with the sound design intact? Distribution errors are more common than content failures.
Testing Rather Than Guessing
The most reliable way to learn what works is to produce variants deliberately and compare them.
Test one thing at a time. Two openings on the same body tells you something about hooks. Two entirely different pieces tells you nothing you can act on.
The highest value test is the first three seconds, because it has the largest effect on everything downstream. Producing three different openings for the same piece is inexpensive when planned into the edit.
Caption and thumbnail are close behind and cost almost nothing to vary.
Keep the results. A brand that has run twenty deliberate tests knows things about its audience that no agency can tell it.
What the Comments Are Telling You
Comment sections are treated as a moderation task and are one of the richest sources of information available.
Read them for confusion. Repeated questions about something the video was meant to explain is direct evidence the explanation failed, and it is more useful than any completion rate.
Read them for language. The words people use to describe your product are frequently better than the words in your brief, and they belong in the next script.
Read them for objections. Recurring scepticism about price, quality or claims tells you what the next piece needs to address.
A single comment saying finally someone explained this properly is weak evidence statistically and strong evidence practically, particularly early on when volumes are small.
Reporting It Honestly
How results are reported internally determines whether the next video is better or merely different.
Report against the stated objective, not against whichever number looks best. A piece built for understanding should not be defended with reach figures.
Report what failed and what you will change. A report showing a retention cliff at eleven seconds and a plan to restructure the middle is more valuable to the business than one showing a large view count.
Include cost per outcome rather than cost alone. A piece costing more per view that produced enquiries is cheaper than one that produced none.
Resist reporting cumulative lifetime numbers as though they were campaign results. They flatter and they mislead.
Organic and Paid Are Different Measurements
Mixing organic and paid performance in one report produces conclusions that are wrong in both directions.
Organic reach is a quality signal. The platform distributed the piece because people responded to it, so strong organic performance genuinely means the content worked.
Paid reach is a budget signal. A piece can reach a large audience purely because money was spent, and its view count says nothing about whether it deserved that audience.
Report them separately. The most useful comparison is a piece’s organic performance before any spend, since that is the cleanest read on the content itself. If a piece performs poorly organically, adding budget usually amplifies a weak message rather than fixing it.
Where a piece performs strongly organically, that is the one to put money behind, and this sequencing is the single most efficient use of a small media budget.
Small Audiences and the Statistics Problem
Most Malaysian brands are not working with the volumes that make analytics reliable, and pretending otherwise leads to bad decisions.
A piece seen by eight hundred people carries real noise. A completion rate that moved four points may be a genuine improvement or may be who happened to scroll past. Treating that as evidence and restructuring the next production around it is how teams chase randomness.
At small volumes, look for large differences rather than small ones. A piece that doubled watch time told you something. A piece that improved it by six percent did not.
Pool your data instead. Ten pieces of a similar type, assessed together, give a far more reliable read than each judged alone, and this is another argument for producing in batches.
And weigh qualitative evidence more heavily than you would at scale. At eight hundred viewers, five thoughtful comments are a meaningful sample.
How to Apply It
Write the video’s job in one sentence before production, and choose the metric that matches it. Ignore raw view count except as evidence of reach, and look at the retention curve rather than the summary figures.
Use distinct tracking links, add a how did you hear about us field, and keep a simple running record of piece, objective, spend and outcome. Judge after two weeks rather than one day, and across three pieces rather than one. Test openings deliberately, read the comments properly, and report against the objective including what you will change.
At Avanguardia, we produce social video for brands across Malaysia and plan variants into the edit so hooks can be tested rather than guessed. If you are reporting view counts and cannot tell whether anything is working, talk to our team.
References
Wyzowl. (2026). The state of video marketing 2026. Wyzowl. https://www.wyzowl.com/video-marketing-statistics/
Think with Google. (2026). Consumer insights. Google. https://www.thinkwithgoogle.com/consumer-insights/