Funnel Optimisation and CPM vs CPI to Reduce The Cost of App Installs via Facebook

We actively acquire mobile users via Facebook, among other channels and I often hear the question about the difference of paying per impression vs paying per download, so I thought I’d write a post on how we look at this and the general acquisition funnel.

Lets start with the basics, Facebook advertising comes down to the following elements — Audience, Creative and Bidding, further contributing elements are app store landing page and app. We are constantly iterating in order to optimise how these work together to maximise the output of the budget.

So, on a very high level, our funnel looks something like this:

In our optimisation efforts, we aim to optimise every step of the funnel. We iterate on the audience and creative combination to increase the click through rate (CTR), tweak the app store landing pages and screenshots to increase the download rate, keep an eye on the app size and icon to increase the rate of the first open and so on, until the final purchase of products in our apps.

Let’s look at a quick example, for an advert, which has generated 100,000 impressions, with a CTR of 0.5% and a download rate: 20%.

With the above performance metrics 100k impressions will result in 500 clicks and 100 downloads. If paying $5 per download the cost would have been $500 and in the case of $5 CPM the cost would have also been $500 — the same either way.

As we optimise the advert and test variations we may degrade or improve the performance metrics and that’s where things start to change.

If, for example, our CTR or download rates were to drop, then the funnel to the right, starts to narrow:

So lets assume that an optimisation attempt results in the CTR falling to 0.25%, then for the same 100,000 impressions, the number of clicks becomes 250 generating 50 downloads (at the above 20% download rate). So the cost at $5 per download would be $250, while at $5 CPM it becomes $500.

Effectively, as the number of impressions, is the same, when paying per impression, the cost remains the same, while the per download cost will reduce, in-line with the reduction in downloads caused by the lower CTR.

Obviously in this case it would have been more advantageous to be paying per download.

If, however, in our optimisation efforts, we increase the CTR or the download rate, the funnel to the right starts to to widen:

So lets say that the click through rate becomes 1% — for the same 100,000 impressions, the number of downloads becomes 200 (at the same 20% download rate), so the cost at $5 per download would be $1,000, while at $5 CPM it becomes $500 (an effective price per download of $2.5)

If, with further optimisation, we manage to double the click through rate to 2% and also managed to increase the download rate to 25%, the number of downloads at 100,000 impressions would become 500 and the cost at $5 per download would be $2,500, while at $5 CPM it becomes $500 (an effective price per download of $1).

In summary, when paying per impression, the cost will remain the same, regardless of how much we widen the funnel to the right, so downloads effectively become cheaper and cheaper with successful optimisation, so for well performing adverts, this will be much more cost effective.

In contrast, negative factors affecting the right side of your funnel, will ultimately result in less downloads at the same cost, effectively making each download more expensive.

It is important to actively monitor the performance of each advert to ensure that things don’t suddenly change, which is not unusual, especially when there is an algorithm change.