+45.1% conversion lift — video arm vs control
One-month window · visitor-level grain · orders attributed on/after first video-page visit
Visitors in test
50,476
Randomized 50/50 split
Statistical significance
p<0.001
z = 9.3 across full dataset
Video-enabled pages
135
Product pages in the test
Two arms, same pages, same month
Traffic was split 50/50 on the same 135 product pages over one month. One group saw a floating video card, the other saw nothing else different. Because the only systematic change is the widget, the gap between the arms is a fair read on what the video actually caused — not a before-and-after that could be explained by seasonality or ad spend.
| Arm | What the visitor saw | Visitors | Converters | CVR |
|---|---|---|---|---|
| Control | No widget | 24,777 | 982 | 3.96% |
| Treatment | PDP floating video card | 25,699 | 1,478 | 5.75% |
| Absolute lift | +1.79 pp | |||
| Relative lift | +45.1% | |||
Conversion rate, randomized arms
The video arm converted at 5.75% against the control's 3.96%, a 45% relative gain. This is the number to quote: it compares two randomized populations, so it is not skewed by which shoppers chose to engage.
CVR by arm
Like-for-like: same catalog, season, pricing and promotions. Only difference is the widget.
How the video arm interacted
The widget reached 88% of the arm, but only 8.9% actually watched a video and 3.9% watched at least half. Engagement is the lever that drives the lift, so this funnel is where most of the upside still sits: a higher watch rate would compound the gain.
Engagement funnel — video arm
n = 25,699 visitors in treatment arm
Engagement rate (watched any) of 8.9% sits just below the 10–15% shoppable-video median band.
Watchers vs non-watchers is not the lift
It is tempting to report that watchers convert 2.8× higher. Don't. Shoppers who choose to watch were already more likely to buy, so this comparison mixes the video's effect with the visitor's intent. It is the most common way shoppable video lift gets overstated.
| Group (within video arm) | Visitors | CVR |
|---|---|---|
| Watched a video | 2,289 | 13.85% |
| Did not watch | 23,410 | 4.96% |
| Apparent ratio | 2.8× |
⚠ Selection Bias
The 2.8× is real but not causal. Shoppers who choose to watch are already more interested, so part of that gap is who they are, not what the video did. The honest causal number is the +45.1% from the randomized arms above, not this comparison.
Conversion rises with watch depth
Conversion climbs from 5% for non-watchers to 17% for those who watch most of a video. An effect that scales with exposure is hard to dismiss as chance, so the dose-response pattern backs up the randomized result rather than replacing it.
CVR by watch depth — dose-response
Conversion climbs with how much of the video is watched. Supporting evidence alongside the randomized result.
Five checks the result passed
A lift is only as trustworthy as the checks behind it. These five are where most vendor numbers wobble. This result cleared all of them, which is why the +45% is reported with confidence rather than as a hopeful headline.
| Check | Why it matters | Result |
|---|---|---|
| Randomized control group | Not a before-and-after; confounders removed by design | Pass |
| On/after attribution | Only orders on/after first video-page visit counted | Lift held |
| Visitor-level grain | One row per person; removes session double-counting | Pass |
| Statistical significance | z = 9.3 across 50,476 visitors | p<0.001 |
| Dose-response present | CVR rises with watch depth | Pass |
What +45% was worth, this brand, this month
Translated into orders, the lift produced roughly 460 incremental purchases in a single month on these pages, with a modestly higher average order value on top. The revenue figure is directional, but the direction is clear and repeatable.
| Metric | Control | Video arm | Delta |
|---|---|---|---|
| Conversion rate | 3.96% | 5.75% | +45.1% |
| Average order value | ₹2,336 | ₹2,475 | +5.9% |
| Incremental orders (vs control rate) | — | ~460 | 1 month |
| Incremental revenue (directional) | — | ~₹1.14M | 1 month |
Incremental orders = test converters above the control conversion rate, applied to test traffic. Revenue uses video-arm AOV; directional.
Want a lift you can defend in a board meeting?
Book a free demo →
