The ICE Score and Powered ICE Score have a big problem - it doesn't consider potential revenue impact.
Two experiments with the same MDE aren't equal if one could generate substantially more revenue.
The solution is a formula that balances both revenue potential and how hard it is to detect wins (MDE).
Want to implement this? Grab my spreadsheet by dropping your email below.
In a past article, I explained that the ICE Score is lacking in a very fundamental way, especially for multi-million brands.
It does not consider power, the ability of an experiment to detect a change with confidence.
And introduced Powered ICE to solve it.
But actually, there's more to consider: Potential Revenue.
The potential revenue impact of a journey can be calculated using a simple formula:
In other words, this formula calculates the revenue uplift we’d see if the experiment reaches the MDE.
However, taking this number at face value would prioritize journeys with higher MDEs, since the uplift would be greater.
If two journeys have the same MDE but one has a higher Potential Revenue Impact, that one should be prioritized higher.
After all, we’re also interested in revenue impact, not just the ability to detect it.
The question is: how can we combine the two?
The best way I found to combine both Potential Revenue and MDE is a proxy that follows these criteria:
I played around with Claude to determine the best formula and arrived at the following:
Proxy Potential Revenue = Potential Revenue * e^(-MDE/0.05)
As you can see, the higher the MDE, the lower the Proxy Potential Revenue.
The 0.05 might look arbitrary, but it makes a lot of sense:
It's chosen because most A/B tests typically aim for MDEs between 2-10%
You could play around with the decay factor based on your situation:
The final step is to rank the journeys’ Proxies relative to each other, which can be done manually or automatically using an Excel or Google Sheets formula.
This approach allows us to prioritize journeys with the highest impact and the smallest MDE—while assigning a low pICE Score to experiments that would be less likely to achieve statistical significance or have a low revenue impact.
To make this process easier, I’ve created a Spreadsheet that automatically assigns the new Power Score to a journey based on the relative Proxy Revenue Impact.
All you need to do is:
And voilà—the spreadsheet calculates the journey’s Power Score.
On the second tab, you can reference the Journey and a Power Score will be automatically assigned.
If you want access to the spreadsheet, click on the button below: