The Peel and Stick Method | Building YouTube Audiences

Who is Simon and how did he get into YouTube?

I used to be an accountant, and I escaped the accountancy to now do the numbers on YouTube. I have an agency in the UK, and about seven years ago I was doing a lot of work in the Google Display Network. About three years ago we moved over to YouTube, and we have worked with companies like BMW, Rolex and Red Bull, as well as a lot of smaller advertisers.

How does Google choose where your YouTube ads appear?

Google chooses your placements for you unless you are direct targeting. Everybody is coming into YouTube now, Google is making it easy to build audiences, and there are a lot of off-the-shelf audiences available. They are moving from cookie-based tracking to federated cohort audiences, which Google is making available within the platform.

A few years ago people were still building campaigns around placements: searching a keyword, getting a list of channels, and sticking ads into those channels. The issue is that it is difficult to scale that way because there are not enough placements or impressions available. If you use standard or custom audiences instead, once you launch the campaign Google chooses those placements for you, overlaying your keywords, placements and topics, and depending on your budget it could choose five thousand or ten thousand placements. At that point you are trusting Google to spend your budget, and if you give Google the keys to the castle and open the wallet, it will take as much as it can. That said, the results from standard YouTube campaigns are still fantastic, people get mega results, we just think there is a slightly better way to do it.

Why do YouTube campaigns often stall after a strong start?

They stall because the AI does not know what to do at launch, so it makes a best guess and then learns from wherever the early conversions come from. When your campaign launches you want a conversion action to take place, be that a web visit, a phone call, a form fill or a sale, and often you see good results straight out of the gate before the campaign stalls, runs brilliantly for a week or two, and then drops off a cliff.

What is probably happening is that either your budget is not big enough for the test, or Google is trying other placements and needs to see that you are serious about continuing. There is always a test point where it decides whether you are going to carry on. Once you start getting conversions, ideally you want to move to a target cost of acquisition model so you can scale against that and grow the campaign more reliably.

What is the peel and stick method?

The peel and stick method is a process that lets you manipulate the placements yourself, on top of the AI doing its own job. In traditional PPC you would build a long list of negative keywords, and in display campaigns you would weed out the placements that were not working for you. YouTube serves up thousands of different videos and channels and tests your budget looking for conversions, so this process does the same thing for placements.

You need to let your campaign learn for probably seven to ten days before you do anything, whatever else you have got going on. At that point, hopefully you are getting some conversions, and that is when you jump in and look at where Google is serving the ads. I would be looking at the placements after a week regardless, and if a particular placement is generating a high view rate but no click-throughs and no conversions, it has to go, because your ads are being watched but nobody is taking any action, and that is spending your budget for nothing beyond brand awareness.

You would love to set an exclusion rule such as "if I get a hundred views on a particular channel and no click-throughs, exclude it", but Google will not let you set up rules like that against a channel. What you can do instead is go through the report yourself, decide what to exclude, produce a list of the channels that are not performing, and add them to an exclusion list. If you reduce the number of placements in a campaign by roughly fifty per cent, which is where we start in the first week, Google will substitute in a whole load of new placements for you.

One of the analogies we use is sticking a poster up in ten thousand retail outlets. Are people coming through the door? If the first five thousand outlets you have put posters up in have nobody coming in, take those down and choose another batch.

Why does removing bad placements also bring your costs down?

Removing bad placements brings costs down because it raises your click-through rate against a more qualified audience, and Google prices ads partly on click-through rate. Just as you could once increase the performance of Google Ads by building a long list of negative keywords, you benefit two ways: your ads are only shown to people who are qualified, and because of that you get a higher click-through rate against the total audience. Google is trying to maximise the amount of money it makes per pixel on the page, and that works out as conversion rate times click-through rate times bid price, so if your click-through rate is higher because your ads are not being shown to unqualified people, your price goes down. It was the same when the Display Network first came out. Algorithms have got better since, so there is less need to do it, but there is still a benefit, and the benefit is reducing your cost per acquisition by about twenty per cent, and over time, if you manage it correctly, I have seen it go down as low as fifty per cent less.

How do you actually build the exclusion list?

You run your ad, let it run for seven to ten days, and then pull the initial set of data, which will probably be somewhere between three and five thousand placements depending on your budget. Once the campaign has run for a week, go into the ad group, go to placements, and click "see where your ads appeared" to get the full list of channels or placements Google has served you.

As an example, in one of our internal campaigns, which drives fifty views an hour to one of our own videos about YouTube SEO, there were 2,741 placements tested by Google's AI, with 87,000 views, 177,000 impressions and a 49 per cent view rate. Download that report into a Google Sheet, and keep the placements that are producing conversions. What you are looking to do is remove a chunk of the rest, not all of it, because if you take it all out it will absolutely butcher the campaign and it will have to go away and learn again. You are just giving Google more of a signal about how you want the campaign optimised.

One thing I do not like to see in these reports is a hundred per cent view rate on a placement with only ten impressions and ten views, so those go straight onto the exclusion list. Then I look for high impressions and high views with no conversions, for example over a hundred impressions and no conversions, or tightened down to thirty impressions and seventeen views with no conversion. I build those out into a separate sheet I call "for exclusion", keeping placements that have not been tested yet, the ones showing no impressions or views at all, because Google has probably just substituted those in.

In this example the target was around 1,300 exclusions, roughly half of the 2,700 placements the campaign started with, and it ended up being 1,356. You only need the channel URLs. Back in Google Ads, go to the exclusion section, click "exclude placements", then "exclude multiple placements", and paste your list in. In this case that took the campaign to 2,010 total exclusions, because some had already been added before.

How often should you repeat the process?

You repeat it weekly: roughly fifty per cent of placements excluded in the first week, twenty five per cent in the second week, and fifteen per cent in the third week, and you should see a dramatic improvement in your campaigns if you keep this up. Monitor the campaign to make sure your stats are improving, and do not butcher it completely by taking out too much at once.

A lot of it comes down to feel, learning which placements are working, and whether your audience is responding to certain channels, genres or types of content. When you find particular themes within the data of where your good conversions are coming from, that is when it gets really exciting, because you can expand and create new campaigns around what is actually working, rather than just telling Google "this is a conversion" and hoping.

What results should you expect?

In the short term we have seen this save people fifteen to twenty per cent of their ad budget, which in the long term means a lower cost per view. You also get a lot more customer insight, and the videos themselves gain massively improved retention rates, which is great for SEO, so you get better SEO results, the channel gains more subscribers, and the videos start popping up in suggested feeds. It is all down to getting more targeted, more relevant traffic and better audiences, rather than relying entirely on Google to do it for you. In the long term, done right, this should save you around fifty per cent of your budget, and even if you only remove some of the clearly non-performing placements, it is still saving your budget.