Notes for owners · Digital marketing
How to make a YouTube ad
The Skip button arrives after five seconds and a skip costs you nothing. That single fact inverts everything television taught you about how an ad should open.
The GullySales team · Updated 21 Sept 2026 · 7 min read
A YouTube ad has five seconds before a Skip button appears, and a viewer who skips costs you nothing. That is the whole craft problem, and it runs against everything television teaches. On television you pay for every household in the break, so the film tries to hold everyone. On YouTube you pay for the people who stayed. So the opening should send the wrong viewer away as fast as possible, and give the right one a reason to sit still.
The first five seconds are a filter, not a hook
Say what this is and who it is for, immediately. "Modular kitchens, Bengaluru" in the first line loses everyone in Coimbatore who was never going to buy, and it costs you nothing to lose them. An ad that spends five seconds being intriguing keeps the wrong people just long enough to be charged for them.
Three things belong in that window: the category, the brand name, and something a buyer recognises as their own problem. A clinic can say the treatment. A machine tool dealer can say the machine.
Do not open on a logo animation. It is the most common waste in the medium, and it is paid for in the only seconds you are guaranteed.
Match the script to how you bought the audience
The targeting decides where the script starts, and most advertisers write one film and then buy three different audiences with it.
If you are buying against search behaviour, against particular YouTube channels, or as a follow-up to people who already visited your website, the viewer already has the problem. Start inside it. No setup, no explanation of the category.
If you are buying a broad demographic in a city, the viewer does not know they have a problem. Now you need the setup, which costs seconds you do not have. So the film has to earn its first five on something else: a face, a place they recognise, a price.
That is two films, not one. It is cheaper to plan both than to run the wrong one.
The formats want different things
| Format | What it is | What the creative has to do |
|---|---|---|
| Skippable in-stream | Runs before or inside a video, skippable after five seconds | Filter in the first five seconds, then be worth staying for at any length |
| Bumper | Six seconds, cannot be skipped | One idea and the name. No story arc fits |
| Non-skippable | Fifteen or twenty seconds, cannot be skipped | Closest to a television spot, and the audience is trapped rather than willing |
| In-feed and Shorts | Chosen from a list, or scrolled past vertically | Works like a thumbnail and a first frame, and needs the vertical cut |
A bumper is not a shortened film. It is a poster with sound. Take the end frame of your main film, add one line, and stop.
Sound off, captions on, phone in one hand
More YouTube viewing has sound than a social feed does. But enough of it is muted, on a bus or in an office, that a film which only speaks its message loses viewers it paid for. Burn captions into the file rather than relying on the platform.
Keep the bottom left of the frame clear. On a phone the platform puts its own button and text there, and it will sit on top of whatever you placed in that corner. The same goes for the top right on some placements. Ask for a screenshot of the ad running on a real phone before you sign off, not the file playing in a browser.
Repetition is a setting, and it can be turned against you
Television repetition is bought. Here it is capped, and if nobody caps it a small audience sees the same film twenty times in a week. That is how a decent ad turns into the thing people complain about in the comments under a Kannada news channel's video.
Set a frequency cap before the flight starts, then use the room it leaves for a sequence. A thirty-second film first, a six-second version to the people who watched it, and nothing at all to the people who skipped. The second film can assume everything the first one said, which is the only place in advertising where that is safe.
Change the film before the numbers flatten, not after. Four weeks is a long time for one execution against a narrow audience, and the stay rate falling is a lagging signal.
When this is the wrong medium
Say the uncomfortable part. If you sell to forty companies in Karnataka and you know their names, video advertising is an expensive way to reach a handful of them. Two people calling from a list will do more in a week. LinkedIn copy is the better spend if you want an ad at all.
YouTube earns its place when the buyer is a person rather than a procurement department, and when the purchase involves seeing the thing. Plus enough of those buyers in your city to be worth a film.
A worked example
A furniture manufacturer in Peenya that sells to interior designers runs two YouTube buys in one month.
The first targets people who watched its website pages in the last month. That film opens with a shot of a delivery arriving late at a site and a voice saying "three weeks, not three months". No introduction, because these viewers already know who it is.
The second targets interior designers in Bengaluru by interest, broadly. That film opens with a designer standing in a half-finished flat in Whitefield with a client on the phone. It takes eight seconds to arrive at the point, so it accepts a lower stay rate and is only two weeks long.
Both end on the same frame, and both send traffic to a campaign landing page built for designers rather than to the home page.
What to do next
Open your last video ad and watch only the first five seconds. Write down what a stranger would know at the end of them: what is being sold, by whom, and whether it is for them. If any of the three is missing, recut the opening. Do that before you spend another rupee on the buy. If your film was made for television and is being run here unchanged, start with the cutdowns instead. To have the media, the film and the page that receives the click looked at together, book the free audit.