Do YouTube ads work for ecommerce?

Somebody watching a video was doing something else. There is no query, no basket, no stated intent. What a good video does is make the product memorable enough that the person searches for it later, which means the order arrives through brand search, direct traffic or a remarketing click days or weeks after the view.

That has two consequences. The first is that the video campaign's own reported return will look weak next to search, because most of the value it creates is credited elsewhere. The second is that judging it on that figure leads stores to switch off the campaign that was feeding the rest of the account.

It also means video needs a longer horizon. A search campaign can be assessed over a fortnight. A video campaign creating awareness needs a run long enough for the demand it created to show up as searches, which is usually a matter of months rather than weeks.

So budget it as an investment in future demand, run it steadily rather than in bursts, and set the expectation with whoever approves the spend before it starts.

Products that need showing are the strongest fit. Anything with a mechanism, an assembly, a before and after, a texture or a use case that is hard to picture from a photograph gets more from thirty seconds of video than from any amount of copy.

New categories are the second strong fit. If shoppers do not yet search for what you sell because they do not know it exists, search advertising has nothing to harvest and video is one of the few ways to create the demand in the first place.

Higher order value helps, because the cost of reaching enough people to produce a sale has to be recovered from the margin on that sale. A product with a small basket and thin margin usually cannot carry it unless repeat purchase is reliable, in which case the lifetime value changes the arithmetic.

Weaker fits are commodity products bought on price, replacement parts, and anything where the buyer's decision is entirely specification driven. For those, the demand already exists and is best met at the moment of search.

Targeting settings matter far less than what plays. The first seconds decide whether anyone stays, so lead with the product in use rather than with a logo, and make the situation recognisable to the person you want.

Build for how it will be watched. Sound is often off, so the message has to survive with captions. Vertical placements need vertical framing rather than a cropped landscape edit. And the product should be on screen for most of the runtime, because a video that only reveals what is being sold at the end has lost most of its audience before the reveal.

Make several. One video is a guess, and the difference between the best and worst performing cut of the same footage is usually larger than the difference between targeting setups. Vary the opening, the framing of the problem and the call to action, and let the campaign find which combination holds attention.

Attach the product feed where the format supports it, so viewers who are ready can move straight to the product. That does not turn the campaign into a direct response channel, and it does capture the share of viewers who were already close to buying.

The reliable method is to withhold the campaign somewhere and compare. Run it in one set of regions and not in a comparable set, then compare total store revenue, branded search volume and new customer counts between the two groups. That measures what the campaign added, which is the only number that matters for a channel this indirect.

If a regional split is impractical, use time. Run several weeks on, several weeks off, repeated, and look at total store revenue and branded search demand across the cycles rather than at any single week.

Alongside that, watch marketing efficiency ratio: total revenue against total marketing spend. A video campaign that is working shows up as the store's overall efficiency holding steady or improving while spend rises, even though the video campaign's own reported return is modest.

What not to do is judge it on view-through conversions. Those tell you who saw an ad before buying, not who bought because of it, and a channel measured that way will always look either brilliant or useless depending on the attribution setting rather than on reality.