YouTube Ads
The format decides the campaign, and the measurement decides the argument
Skippable in-stream, bumpers and in-feed video are three different products that happen to share an account. Choosing between them is most of the strategy. Reporting on the result without flattering yourself is the rest of it.
The first decision
Three formats doing three different jobs
Most disappointing video campaigns are format mismatches. The asset was made first and the placement was chosen to accommodate it, which is the wrong way round.
| Dimension | Skippable in-stream | Bumper | In-feed video |
|---|---|---|---|
| What it is buying | A decision to keep watching, made at the five-second mark | Six seconds of unavoidable, unarguable reminder | A click on a thumbnail from someone browsing or searching |
| What the opening has to do | Earn the next twenty seconds before the skip button appears | Land one idea completely, because there is no second chance | Make a still frame and a title worth choosing over the alternatives |
| Typical job in a plan | Carry the actual argument to a cold or lightly warm audience | Maintain presence between longer messages, or reinforce a launch | Reach people already looking for something adjacent to your subject |
| Where it goes wrong | A slow open, so the only people who see the message were never going to skip anyway | Trying to explain a proposition in six seconds instead of landing one thing | A thumbnail and title treated as an afterthought, so nothing is chosen |
| What you need to make | A film with a real hook and a structure that survives being cut short | A single clean idea, produced to be understood without context | A strong frame, a title, and a video that pays off the click |
What this channel is for
High attention, low intent, and why that is not a criticism
Search advertising works because the person told you what they wanted. Video advertising works despite the person having told you nothing. That single difference explains almost every disappointment we are asked to diagnose here.
What YouTube offers instead of intent is attention. People arrive to watch something and they stay, which means a message has room to be more than a headline. That makes it a strong place to build demand, explain something that needs explaining, and re-reach people who already know you. It makes it a weak place to harvest people who are ready to buy right now, because those people are typing rather than watching.
Run properly, video sits upstream. It creates the searches that your search campaigns then capture, which is exactly why last-click reporting makes it look worthless and why the channel is usually the first thing cut in a bad quarter.
Scope
What running video with us involves
The list is deliberately weighted towards decisions made before launch, because that is where video campaigns are won or lost.
- Objective definition and a written statement of what the campaign is expected to change
- Format selection made before production, with the creative requirements each format imposes
- Campaign type and bidding choices, including which controls are available in each
- Audience construction, including first-party segments and the exclusions that stop overlap
- Frequency capping set to a deliberate number for the message length and buying cycle
- Placement, channel and content exclusions reviewed rather than left to defaults
- Creative feedback against the five-second decision point rather than against internal taste
- A measurement plan agreed in advance, naming which numbers are evidence and which are context
- Coordination with search and remarketing so the same conversion is not claimed twice
- Reporting that distinguishes what happened from what we can reasonably say the campaign caused
The hard part
Why nobody agrees on what a video campaign did
- View-through numbers make the campaign look extraordinary.
- A view-through conversion records that someone saw the ad, did not interact with it, and converted later within the window you chose. That is a correlation, and lengthening the window will usually increase the count without anything real having changed. Google keeps them out of the main conversions column for most campaign types, which is a strong hint about how much weight they should carry.
- Last-click reporting makes the campaign look worthless.
- If video creates demand that is later captured by a branded search or a direct visit, last-click gives the credit to the capture channel. The video budget then gets cut, the search results quietly weaken over the following months, and the two events are never connected because nothing in the reporting joins them.
- The measurement quietly shrinks as browsers change.
- View-through conversions cannot be reported from browsers that do not allow cross-site cookies, so part of the effect is invisible for reasons that have nothing to do with the advertising. A year-on-year comparison of that metric is comparing two different measurement environments.
- Nobody agreed what success looked like before launch.
- Video is the channel most often judged by whichever number happens to look best or worst at review time. Deciding in advance what evidence would justify continuing, and what would justify stopping, is the only reliable defence against a metric being selected after the fact to support a decision already made.
Sequence
How a video engagement is ordered
Decide the job, then the format
Demand creation, launch support or re-reaching a known audience are three different briefs. The format follows from the job, and the creative requirement follows from the format.
You get: Objective and format decision, written down
Agree the measurement before spending
Name which numbers will be treated as evidence, which as context, and what the read window is. This conversation is much easier before anyone has an opinion about the result.
You get: Measurement plan with stated limitations
Build audiences and exclusions together
Every audience needs its matching exclusion, otherwise the campaign spends against people other campaigns are already reaching and the reporting becomes unreadable.
You get: Audience and exclusion map
Launch with frequency controlled
Set the cap deliberately at the outset rather than reacting to complaints later. Uncapped delivery is not a neutral default; it is a choice to let the system decide how often someone hears from you.
You get: Live campaigns with documented settings
Read across channels, not within one
Look at what happened to branded search, direct traffic and assisted paths during and after the flight, alongside the platform numbers. Neither view alone is enough.
You get: Cross-channel read at the agreed checkpoint
Questions
What advertisers ask about YouTube
Which YouTube format should we use?
It depends on the job. Skippable in-stream buys a decision at the five-second mark and suits a message that can earn the next twenty seconds. Bumpers are six seconds and non-skippable, so they are a reminder rather than an argument. In-feed video appears as a thumbnail people choose to click, so it behaves more like a discovery placement.
The formats also differ in how they are billed and which campaign types they belong to, which is why choosing the format after the film has been made is the wrong order.
Are view-through conversions real?
They are real measurements of a real thing: someone saw the ad without interacting with it and later converted. What they are not is proof the ad caused the conversion.
Google treats them carefully for that reason. They are reported separately from the standard conversions column for most campaign types, they exclude people who also interacted with your other ads, and they cannot be reported at all from browsers that block cross-site cookies. Widening the view-through window will usually raise the number without changing anything about the world.
Should we run YouTube if we cannot prove it drives sales?
Only if you can live with a probabilistic answer. Video is bought on the argument that reaching the right people with the right message shifts behaviour later, and that argument is supported by patterns rather than proved by a row in a report.
If your business needs every pound tied to a traceable sale, put the budget into capture channels first and come back to video when you have the tolerance for a slower and less certain read.
How often should the same person see our ad?
Fewer times than an uncapped campaign will show it. Frequency capping limits how many impressions or views an individual can receive over a day, week or month, and setting it is a deliberate act rather than a default.
The right number depends on the length of the message and the length of the buying cycle. It is not supported on every campaign type, which is another reason format and campaign choice come before creative.
We have one video. Is that enough?
It is enough to start a test and not enough to run a channel. One asset gives you no way to separate a weak message from a weak audience, and it burns out against a capped audience quickly.
A practical minimum is one longer piece that can carry an argument plus short cuts derived from it, so you have something to run in each format without commissioning three separate shoots.
Decide what your video budget is allowed to be judged on
On a strategy call we will agree what job the video is doing, which format that implies, and which numbers you are willing to accept as evidence. That conversation is worth having before the film is made.
If we don't deliver the work we agreed to deliver for reasons within our control, you don't pay for the undelivered work. Read our guarantee
Last updated · Reviewed by Zubair Afzal