17 Sep · 2 min read
Cost per install is the cost associated when a user downloads your application and opens it for the very first time. Cost per install only counts when an app is opened once after downloading it. This is the first time the user’s statistical information can be captured by whatever software development kits or (SDKs) you have within your application. The CPI metric is just one option in a long line of cost-associated metrics: CPI, CPR, CPATC, CPP, CPA. Moving forward, CPI is a purely volume-based metric and is not a volume that defines much. At the top of the pathway, installs do not always correlate to value at the bottom.
For example, let us say you have an eCommerce application, and you are driving people to purchase products. If you achieve ten thousand downloads, note that these downloads are not providing you with any value at all except for numbers on a screen that you can report to stakeholders and investors.
The mindset shift is easy because if you suddenly change it to wanting 10,000 purchasers, you can then assign a lifetime value (LTV) and work out your maximum cost per purchase metric.
If the average user gives you fifty pounds, then you would want five hundred thousand pounds in revenue. Then let us say you have a forty percent blanket profit margin, which means you can afford to pay up to twenty pounds per user to generate that revenue. However, it is important to note that paying twenty pounds will be breaking even. Anything below that cost per purchase/acquisition is going to be money in your pocket.
Suddenly, now have the numbers to work out issues like estimated media spend and how long it will take to get that five hundred thousand pounds in revenue. The math would suggest you have two hundred thousand pounds in media spend to break even or one hundred thousand pounds at ten pounds CPA to increase your profit margin. The following steps ensure you have the right growth strategy to be able to achieve this.
Now you can see, this is why CPI is considered a no-mans-land metric. It does not hold anywhere near the same weight as the occasion at the end of your funnel. You cannot draw specific metrics from it; however, you could apply a conversion rate model and go from installing to registration to purchase and work it out based on that initial cost. However, a better recommendation is a more valuable metric.