An Advanced Guide to Impact Sizing

One of the hardest parts of PM

It’s mind-blowing that people still estimate the impact of a feature using something like high/medium/low (for an established product). I used to do it, too, 15 years ago. 

Now the Baseline

But, nowadays, the gold standard is actually estimating the impact of a feature. Most high-performing product teams estimate the impact of a feature to their OKRs (eg, engagement metrics). 

And the advanced one’s also estimate the impact of a feature to their output metrics like revenue and profit. This helps you truly prioritize the features that are going to change the trajectory of your business.

Aside from the obvious benefits of helping your career due to higher impact, this also sets you up well to impress in product reviews. You can create an amazing presentation for your roadmap that looks something like this.

The trouble is - impact sizing is super hard. As one public company senior PM said:

It’s always rough. I’ve rarely seen someone do it in a really scientific way.

How can you do it better?

So, the question is: how can you size better? 

It’s one thing to do the hand-wavy PM thing and say, “I’ll estimate a 10% adoption of this feature.” But that’s only going to get you so far. Your impact will be much less predictable. And rigorous product leaders, engineering and analytics counterparts will question your decision-making.

It’s much better to have data-driven reasoning behind your product sizing:

• Based on the actual number of users estimated to see the feature

• With a high confidence estimate of adoption and engagement impact

• And appropriate assumptions to understand the top & bottom-line impact

So, you go to Google and GPT-4 for help.

There’s No Content

Google and GPT-4 fail on this type of content. 

If you Google it, there’s not a single article that actually walks through how to impact size different types of features and metrics. It’s quite a sad state of search results for “how to impact size product feature.”

If you BingGPT it, the results aren’t any better. It’s just a high-level explanation, with no tactical guidance. If you push BingGPT to impact size a feature, it will just give up altogether:

This type of impact sizing is something PMs just started doing in the last 10 years. As a result, there’s a dearth of good knowledge or best practices.

Enter Today’s Piece

Uncovering this content gap, for today’s piece, I’ve teamed up with one of my favorite Product Creators, and current Senior PM at GoodRx, Carl Vellotti. Together, we bring you: an advanced guide to impact sizing. 

We’ll cover:

• Templates to estimate the most common metrics

• How to set up experiments to identify and de-risk your riskiest assumptions

• Tips & tricks to get to the key data points

This is a ‘201’ level course for impact sizing. It’s a great fit for PMs new to the practice or product leaders looking to adapt it. It’s the type of post you’ll want to send to your colleagues.

Let’s get into it.

---