Andrew Chen Archives

Buying ads make sense for direct monetizing products
When it comes to products that directly monetize their audience using subscription, ecommerce, virtual items, etc., it can make a lot of sense to rely on advertising as a distribution channel. The reason for although there are 100s of millions of internet users, only a small fraction (usually <1%) will be in-market for your services at any given time. As a result, you are looking to acquire these users, and only these users, and everyone else is considered wasted energy.

Note that the freemium model is a variation on this concept – where you acquire a large base of “casual users” that stick around, and you slowly convert some % of them to subscribers. You can think of it as vertically integrating your distribution, and instead of spending money to buy ads, instead you are spending money to support this large base of free users.

Run an ad-supported consumer internet app? You better understand ad-buying too
These days, it’s simple to think that advertising is about placing javascript code on a page, and seeing what numbers AdSense gives back to you. Instead of thinking about it that way, publishers would do their customers (their advertisers) a great service by understanding what it means for ad inventory to perform or not perform. They should really understand how advertisers analyze their cost-per-acquisition so that the publisher can better service them – that’s at the heart of indirect monetization. If you’re not selling directly yourself, you’re helping someone else sell.

CPA, the common currency of user acquisition
The first step is to understand your user acquisition funnel, from start to end. Although there are many ways to price things, be it CPM, CPC, or CPA, the key is that it all rolls back to how much it costs you to have a registered user. You need this cost-per-acquisition number to be lower than the lifetime value number, and what you have left is profit (before cost of infrastructure, etc).

So you want to build something that looks like this:

The above is an example of two traffic sources, Google and Advertising.com (the latter being an ad network), as well as clickthrough rates, signup %s, and the cost per acquisition.

A couple important notes on the above:

• the SOURCE of your traffic is the most important segmentation – make sure you track acquisition and LTV numbers, since you often get vastly different numbers depending on where you are buying ads
• you want to break down your funnel into as small of steps that make sense, from the clicks into the signup page into any intermediate profile forms and then the final registered numbers. Your funnel may be larger or smaller
• Google might charge you CPC and Ad.com might charge you CPM, but you have to normalize that back into how much it costs you to acquire a registered account. In a CPC model, you don’t care about the CTR much since you don’t pay for impressions that don’t result in clicks, whereas you do care about CPMs
• the only difference between a good CPA and a bad CPA is whether it’s above or below your customer LTV
• In addition to tracking source of traffic, you may also want to track important factors like what campaign it was in, what creative it corresponded to, the banner ad size, and other things that might affect CPA. The last thing you want is a variation that is very unprofitable, but is obscured by being grouped together
• You may also want to group all your marketing channels into the above, including email, partnerships, blog traffic, viral invites, etc. Obviously for stuff that’s free traffic, the CPA is infinity, but it’s good to know what kinds of funnel %s the other traffic throws off, for comparison’s sake

Got all that? Good :)

What factors influence ad performance?
The second thing worth considering is what factors actually influence GOOD numbers for CPAs versus what numbers are generally bad.

I made a quick, anecdotal table below to enumerate some of the factors:

A couple additional notes:

• As mentioned previously, the source of traffic is very important – you should dedicate a significant amount of time buying lots of different kinds of traffic to see what works
• Cost model is something you should be able to normalize into CPA and mostly ignore, except for cashflow and risk reasons
• User requirements can be a huge issue – if you are forcing users to download, that will kill your CPA. Similarly, asking a demographic that doesn’t have credit cards for their credit number can kill you. Make sure that you understand your audience and that your funnel is optimized for them
• Audience Theme revolves around the concept that strongly themed products are often quite vertical in nature, which causes a large % of users to reject the product. For example, a site for teens obsessed with vampires is much narrower than a web email site. The narrower the theme is, the harder it is to find appropriate ad inventory to buy
• Funnel design and A/B testing is key – definitely worth investing in
• Similarly, for those who can find a viral angle in their product, that can be a huge benefit as well. It has the capability to create order-of-magnitude decreases in the CPA, which can be the difference between profitability and bankruptcy! But this also has issues if your product is not widely appealing enough, since virality depends on a horizontal offering to work

If you have questions or comments, feel free to leave a message!

I write a high-quality, weekly newsletter covering what's happening in Silicon Valley, focused on startups, marketing, and mobile.