Tried YouTube ads again, this time with more analysis

I've tried advertising on YouTube before, but have since gotten a little bit more sophisticated about my approach, and thought I'd share what I've learned. I'm tantalizingly close to break-even such that I could leave the campaigns running indefinitely, but not quite there yet.

The business I'm trying to gain customers for is Candy Japan, which is a service that sends you Japanese candy on a twice-monthly basis. Actually now there are multiple tiers, so you can also choose just once a month if you'd like. That was a big enough change that it warranted trying ads again to see if they might work better this time.

I'm going to call finding a new subscriber a "conversion", because that's the typical lingo when dealing with ads.

What is a conversion worth?

I had never really calculated this properly before, and only had a ballpark guess of $50. This means that if someone subscribes, I figured that they would stay on for long enough that eventually on average I could make $50 of profit from a new subscriber. However this is only "ballpark", because I didn't properly gather the data necessary to make a more accurate calculation.

Well, I did so a few days ago, and found that I wasn't too far off: the average conversion originating from a Google search is worth $50.06 if I take into account the first 24 months, or $41.33 if only taking into account the first 12 months.

I thought that people who go out of their way to Google for my site would not behave the same way as people who click on YouTube ads, but when I pulled the data and ran the numbers for anyone who originated from or mentioned the word "youtube" in the "where did you hear about this site" questionnaire, the results were very similar: $48.66 profit for the first 24 months, $41.05 for the first year.

Since I ran these campaigns only recently, I didn't have 24 months of data for YouTube customers, but I did the best extrapolation for how long people stick around for that I could, based on the just the first 4 months (shown below).

What is a conversion worth?

What does it cost to convert a customer from a YouTube ad?

This was surprisingly tricky to figure out, because it's hard to account for everyone who might have seen the ad and then converted a bit later on, possibly even on another device. If I did the same trick of looking at the referer, utm tags and "where did you hear" field, I could account for 32 people that subscribed thanks to the ad campaign.

This matches very well just eyeballing how my new subscriber count changed during the campaign vs. how it would have typically changed without the campaign, so I don't think this undercounts by much, so I decided to use that as the value to base my profitability calculation on.

With 32 conversions and the campaign costing $1,566.71, I came up with a cost-per-conversion of $48.96.

Where to improve

I hope you'll agree that the cost-per-conversion of $48.96 is tantalizingly close to the $48.66 lifetime value of these customers. But "lifetime" here is two years, meaning that if I spend that money to get more of these customers, then I'll need to keep running the service for a full two years — just to get my money back!

Where to improve

I think the biggest improvement that could be done would be to make the ad itself better. At the moment it's cringy enough that I hesitate to even link to it, but take a look if you'd like. It's the first ad I've ever produced, cost money and took months to get done, so even if I'm not entirely happy with it, I'd rather find something else to improve first before changing it.

I don't think I can improve my profitability by much. If anything, I'd like to spend more on each customer, and staying on budget always takes some self-control. I want people to get as many amazing sweets as they can, but I have to keep reminding myself that without staying profitable, I can't keep doing this at all. I don't really have any extraneous costs to cut either, just the box, shipping, handling and the candy itself. All of those are already reasonably priced.

If improving profitability isn't possible, then I must either reduce the cost of showing the ad, or up the proportion of people who buy after seeing it.

Annoyingly I am already bidding the lowest amount possible: $0.01 per ad view, so I cannot go any lower there, except if I started bidding in a different currency. Bidding in Japanese yen would let me actually bid 1 yen = $0.0092 per view, but that's still not low enough.

Where to improve

Another place to improve would be the targeting, but I have already spent hours and hours optimizing it, and while I'm sure I could have missed something, this wasn't just a naive "show ads to random people" approach. I've already looked at which countries, demographics, targeting options etc. converted the best and winnowed down the campaign to only the best groups.

Perhaps the landing page could be improved, although that is also already the result of A/B testing two different designs against each other, and is problematic to do as I don't have enough data to rapidly try many different designs.


I will not be continuing this campaign for the time being, until I come up with some trick to get cost-per-conversion down by about 30%. My best options for doing so would be to either improve the ad, the landing page or to somehow target the ads better.

Thanks for reading, and if you'd like to get candy surprises from Japan on a recurring basis, please do subscribe.