How to prioritize with the ICE framework

How do you decide what to work on as a growth team?

The ICE framework is a quantitative way to answer that question.

It was invented by Sean Ellis to help growth marketing teams evaluate and prioritize their ideas and experiments.

What does ICE stand for?

ICE is an acronym for Impact, Confidence, and Effort (this one is interchangeable with Ease).

Impact: How big is the impact? It’s measured in terms of growth marketing KPIs like number of users and revenue or more granular metrics that ladder up to them, like trial-to-paid conversion rate or retention rate.
Confidence: How likely is it that this will succeed? This factor takes into account available data, results from previous experiments, and team expertise.
Effort (or Ease): How much effort will it take to implement? Factor in time, budget, and involvement from other people and departments.

How do you calculate ICE scores?

To calculate the ICE score of each experiment, rate them on a scale of 1-10 for each of Impact, Confidence, and Ease. Then add them up together for a final score between 3 and 30.

The experiments with the highest score are the ones you should prioritize and work on first.

While this sounds elegant in theory, I find that there’s nuance to scoring ideas and experiments.

Unless you have clear definitions for each of Impact, Confidence, and Ease, people on your team can end up assigning wildly different scores to the same experiment. And the whole purpose of using a (somewhat) quantitative framework gets defeated.

So here’s a rough guide to help you assign ICE scores.

Impact

Judge impact based on three factors.

Scale: If this works, can we scale it? An example of low scale is leveraging the founder’s network for customer referrals. High scale ideas could be running Google Ads or running cold outreach.

Repeatability: Is it a one time thing or can we do this again and again? Launching on Product Hunt or sponsoring an event are examples of projects that have a shelf life. On the other hand, building a template for competitor pages is repeatable.

Flywheel: Does this make a flywheel (or growth loop) spin faster? For example, if your app grows when other people refer their friends, experiments that make that process easier feed a core growth loop.

Here’s a scoring guide for Impact.

8-10: Scalable + repeatable + spins flywheel
5-7: Fits one or two factors
1-4: Fits one factor or none

Confidence

You can be more confident in assigning scores for this factor if you’re running experiments for which you have good data or experience. But if it’s a completely new channel, knock the confidence score down by a few points.

You’re free to flesh out the scores a bit more but I usually start off with these.

9-10: High confidence.
5-6: Semi-confident.
3: Not confident.
1: Zero confidence.

Effort

Think both in terms of resources as well as time.

If an idea requires engineering and design resources, it’s going to take far longer to ship. But if it’s something you can knock out on your own, it’ll score highly.

For example, if you need to make a new landing page but you can’t make changes on the site without the help of an engineer, the idea’s going to get a poor score. But if your site is hosted on Webflow and you’re somewhat savvy, the same idea would get a high score.

8-10: Something you can do quickly without involving others.
5-7: Need support from other teams.
3-4: Need a ton of support from other teams.
1-2: Require buy-in and involvement from across the company

A word of caution against using the ICE framework

The most important thing I’ve learned after over a decade in marketing is that what you do matters far more than how you do it.

So while the ICE framework is a great tool, the caveat is that it’s only as good as your backlog of experiments. For example, if you are haemorrhaging users, running acquisition experiments isn’t going to help even if they’re successful.

That’s why it’s good to take a step back and build a quarterly backlog of experiments that are aligned with your business goals.