Come up with experiments that will have the biggest impact with the least amount of effort
Eric Ries proposes that you focus on a metric that could "explain" the success or failure of your project. This is super specific to your project and sometimes referred to as "north star". The main characteristic of the north star is that according to your value hypothesis if this metric trends in the right direction then you have a business!
Now that you have a single metric to optimize for, how will you choose what to work on? You come up with experiments that will have the biggest impact on your north star with the least amount of effort.
The best way I found to do this is to list everything you can think of that according to you should have an impact on your north star. Next, for each item try to think how much “effort” it would take to make the simplest version of this. Then try to imagine what the impact could be and try to quantify it. At this point, “impact” would be how much you think that it could impact your north star.
An example of such a list could be:
Item #1: 2 | 6
Item #2: 8 | 8
Item #3: 2 | 8
Item #4: 6 | 2
Now that you have this, you can visualize it with a very simple chart. (Pro Tip: I don’t even bother making the list, I just drop items in a canvas with both impact and effort axis.)
^ impact
| |
| #3 | #2
| |
| #1 |
| |
| |
| |
|------------------------------------------------------
| |
| |
| |
| | #4
| |
| |
| | effort
|------------------------------------------------------>
Next, you focus on the ones in the upper left quadrant. The higher and more to the left, the better.
Another framework for this is the "ICE framework" ICE stands for impact, confidence and ease of implementation. Instead of having quadrants, you use a scorecard.
Curious what you think of this approach vs. what you wrote in the post.
I've also seen "reach" as being another dimension. Any amount of data that will give a "normalized" score will work in the end. However, beware, effort, impact, confidence, etc... are all subjective, and adding variables may lead to even less accuracy in your scoring.
My experience is that we can't really ponder too many dimensions and also personally, nothing beats the visual aspect of 2 dimensions. As I mentioned, I don't even assign numbers, I move "cards" around on a canvas (Miro.com is good for this).
That said, if I were to add "confidence on the impact estimation", then I think I would still try to stick to 2 dimensions by defining "impact = estimated impact * confidence" or something like that.