How to measure video marketing ROI without lying to yourself
Views are a budget report in a performance costume. Here are the four numbers that tell you whether a video worked, and how to value the ones that never produce a click.
Video is the marketing asset most often reported on and least often evaluated. Every dashboard has views, watch time and engagement rate. Almost none of them answer the only question a finance team is actually asking, which is whether the money produced anything.
The problem is that the easy metrics measure distribution rather than effect, and the metrics that measure effect require deciding in advance what the video was for.
The metric that means nothing
View count is a function of budget. Spend more, get more views. Reporting it as a result is reporting that you spent the money you were given.
It is not that views are false, they are just not evidence about the video. Two creatives with wildly different quality, run at the same budget against the same audience, will report similar view counts and completely different outcomes.
The four numbers that do mean something
- Three-second hold rate. The share of impressions that survived the first three seconds. This is the hook's own score, and it is the first thing to fix because everything else is downstream of it. Why the hook decides the rest.
- Completion at 25 and 50 percent. Whether the argument held once you had attention. Full completion is mostly a measure of length, so the quarter and half marks are more diagnostic.
- Action rate. Clicks, demo bookings, add-to-carts, signups, whatever the specific next step was. Measured against impressions, not against views, so a video that failed to hold anyone is properly penalised.
- Cost per outcome. Total production and media spend divided by the actions produced. The only figure that is directly comparable to every other marketing activity you run.
Attributing a video that never gets a click
A landing page hero, a sales-deck film, a trade stand loop, none of them produce a click, and all of them can be measured. Three methods, in increasing order of rigour.
| Method | How | Good for |
|---|---|---|
| Page-level split test | Same page with and without the video, traffic split evenly, compare conversion rate. | Landing page and product page video. The cleanest evidence available. |
| Cohort comparison | Compare deal progression for prospects who were sent the video against those who were not. | Sales enablement film, where sample sizes are small and slow. |
| Assisted-conversion view | Look at whether a video interaction appears in converting paths at a higher rate than in non-converting ones. | Awareness work that influences without closing. |
The split test is worth the setup effort. A single well-run page test on a product page hero settles an argument that otherwise recurs every budget cycle.
Building the ROI figure
Video ROI is unusual in one respect: production is a one-off cost against a benefit that keeps accruing, so a first-month calculation almost always looks bad and is almost always wrong.
- Amortise production across the asset's real life. A film used for a year on the site, in sales and in email is not a cost against month one.
- Count every placement. The same asset in a deck, an email, a trade stand and a landing page is four uses of one production cost.
- Include what it replaced. If the video removed a recurring explanation from sales calls or a repeated onboarding session, that is a saved cost with a real number attached.
- Keep media and production separate. A brilliant video with poor targeting and a poor video with excellent targeting both underperform. Averaging them hides which one you have.
A benchmark worth having, with a caveat
Benchmarks vary enormously by platform, audience and category, and anyone quoting a single industry number is selling something. As a rough orientation for paid social in 2026:
| Metric | Weak | Working | Strong |
|---|---|---|---|
| 3-second hold rate | under 20% | 25 – 40% | over 45% |
| 25% completion | under 10% | 15 – 25% | over 30% |
| Click-through rate | under 0.5% | 0.8 – 1.5% | over 2% |
The genuinely useful comparison is against your own last video to the same audience. It controls for everything an industry average cannot.
Reporting it to people who did not commission it
Four lines. Anything longer invites a discussion about the music.
- What the video was supposed to make someone do.
- How many people did it, and at what cost each.
- How that compares to the last thing we ran for the same purpose.
- What we are changing next as a result.
That fourth line is what turns a report into a programme. Video gets better through iteration, and the cheapest iteration available is testing the opening against a body that already works.
The measurement mistake that costs the most
Judging a brand film by performance metrics, or a performance ad by brand metrics. They are different products with different jobs, and applying the wrong scorecard reliably kills the more valuable of the two, usually the brand film, because its returns arrive later and are harder to attribute. Decide which one you commissioned before you decide how to score it.
Common questions
- What is a good ROI for video marketing?
- There is no single figure, because it depends on what the video replaced and how long it stays in use. The meaningful calculation is cost per outcome, production plus media divided by the actions produced, amortised across the asset's real life rather than charged entirely to its first month.
- Which video metrics actually matter?
- Four: three-second hold rate, completion at 25 and 50 percent, action rate measured against impressions, and cost per outcome. Fix them in that order, because optimising a call to action on a video most viewers never reached achieves nothing.
- How do I measure a video that does not get clicks?
- Split-test the page it lives on with and without the video and compare conversion rates. For sales material, compare deal progression between prospects who received it and those who did not. Both give cleaner evidence than any engagement metric.
- Are video views a useful metric?
- No. View count is a function of budget, spend more, get more views. Two creatives of very different quality run at the same budget will report similar views and produce completely different results.
