Using cost-per-click data to debunk advert design myths

http://www.sketchdeck.com/blog/we-tested-48-facebook-ads-to-bust-6-marketing-myths/

by edmack • 10 years ago

40 10 10 years ago

ssharp2• 10 years ago

It is important to note that we only measured who clicked through our ads. That means the irrelevant images could have lower conversion rates on the site. The people who ended up on the SketchDeck site may not need or care about design.

I was really looking forward to this until I saw they were using CPC as their barometer. Bad traffic at $0.87 / click is far costlier than good traffic at multiples of that.

nieksand• 10 years ago

Exactly.

The article is an exercise in optimizing for the wrong metric.

adamredwoods• 10 years ago

Their methodology is bad. The most effective CPCs are the ones that do not look like ads, therefore, are not representing accurately what a person is clicking on. A better approach would be to compare the different designs that LOOK LIKE ADS, and test which one is most effective.

The small logo set also included the most popular ad from this entire experiment. Which is one of my favorite pieces of stock photography and can be seen below...

The above statement confirms that choice in creative can have a tremendous impact. Each of their "myths" can be adjusted using effective creative.

DiabloD3• 10 years ago

This entire study makes no sense to me.

I've never heard of Sketchdeck, but CPC is meaningless. If they are selling a product, it should be cost per sale or cost per dollar or something in this vein. This is the only metric that means a damn, I could care less who clicks my ad if I can't convert them.

woolinsilver• 10 years ago

could not care less

20years• 10 years ago

Focusing solely on CPC is a mistake. Conversions are far more important IMO.

As an example for using irrelevant images - sure cute cat pics are going to get clicked way more resulting in a lower CPC but those are not going to convert.

Same thing with the text on images. Without text on the image that explains what it is about, users might click more out of curiosity. Those are not going to convert well though.

I would much rather have a lower CPL or lower cost per conversion over a lower CPC.

To sum it up, the ending summary of "To have a successful Facebook ad:" should really be "Focus on what drives in the lowest cost per conversion and highest ROI"

marcus_holmes• 10 years ago

So they discovered that people click slightly more on unbranded images of happy women with no relevance to a product. Good work guys!

But the total difference from best to worst was < 5%. If you're doing a vast ad campaign then sure that totals up. But what's the opportunity cost of the brand awareness? Spending ~5% more per click to use a relevant, branded, image seems like a good deal.

I think it's great that they published this, but it doesn't really bear out the conclusion. Of course, given the nature of the author the conclusion was always going to be: "design is really really important and you should totally spend more money on it".

ericclemmons• 10 years ago

I think most of the commenters are missing the point of the tests.

Having sat through a few meetings about "the best way to design FB ads", I've actually heard a few of these "myths" as converting best.

And they meant CPC. Because the goal for lots of ads is to get the foot in the door, then optimize the rest of the funnel.

The initial click is often the hardest, which is why this article is testing the myths against the CPC, since the overall conversation differs greatly once past that.

ssharp2• 10 years ago

All feet in the door are not the same. When you're iterating on ad concepts, maybe CPC is an okay metric, or at least an early indicator. But these ads weren't really interations. Without a baseline CPA and running ads related to the ad that generated that baseline, CPC has no context.

Also, if all traffic is going through that to-be-optimized funnel, how does that justify ignoring the CPA of any individual lead source?

The only explanation that makes sense is that conversions aren't frequent enough to find significance. That makes more sense given how many variations were run. In that case, it doesn't make sense to run so many variations. That, or maybe find an in between metric that at least indicates some greater level of lead quality.

cuchoi• 10 years ago

From the confidence intervals you can tell that there was no statistically significative difference between the different alternatives.

Using cost-per-click data to debunk advert design myths

Related Stories