Stop Measuring Everything | Hacker News Reader

Stop Measuring Everything

savagethoughts.com

44 pointssmalter16 comments

The correct response to bad measurement isn't to throw up your hands and stop trying to measure. It's to get better measures.

Be careful about your measure. Check that what you're measuring actually corresponds to what you think you're measuring. There are a lot of good examples here, but the answer isnt' to stop measuring, to resort to "art" - it's to get better at it.

Yep. It seems like they took the opposite approach to what they should have.

They killed their billboard campaign because their attempts to attribute the value of the campaign to growth failed.

They talked to a few people (got attribution) and learned it had a wider effect than their original attribution did. This is more data, not less.

It sounds like they should have taken a different approach to attribution in the first place when they moved from purely digital marketing to meatspace ads.

I'm curious if you or the parent have any suggestions for a problem I've been thinking about.

I've been building a search engine for standalone lectures (https://www.findlectures.com). For me, the ideal outcome of being introduced to a new concept is that an idea sticks in my head for a while and won't let go, and I'd like to help people replicate that experience.

Are there ways that you could track something that gets at whether this happens for people, or is this so qualitative that the best option is to interview users?

That's tricky and I don't have any great advice. If you could find some tangible piece of knowledge associated with each video, then have a quiz after some time, that might help. Or introduce some kind of spaced repetition.

Alternatively, if you make the video categorization about what the viewer is interested in learning (that is, make the interface drive people to see videos based on what they want to learn, rather than having a hierarchical organization), then you can ask them later "did you learn X?" with some other survey like questions.

Well, one correct response is absolutely to figure out if the end justifies the means. Proper measurement often isn't easy or casual, between staging a reasonably clean test and figuring out the actual significance of the results.

If you naively insist on measuring a large amount of things, especially different things, chances are pretty good you're doing it slapdash and relying more on tea leaves than science when you interpret the results. Plus you run the risk of downplaying anything that can't be given a number.

I took the "don't measure all the things" lead-in more as pick your battles. If you actually have spare bandwidth to measure all the things in a valid manner, great, but some things can't be easily quantified and maybe the bandwidth is better spent qualifying them instead.

The piece isn't terrible – the author has found the value of qualitative research. Yay! (I do a bit of that, so I'm biased).

The title however, is wrong. It's more about figuring out how to measure the right things and not missing signal because of bad measures.

Open-ended, qualitative research is good at that because it helps the signal stand out – you can find unknown unknowns and then figure out how to be more systematic in measuring them.

Money. You should measure money. everything else is just a proxy for that.

That's a very sad way of thinking.

Yes, it is. But let's also not forget time. It should be money/time.

The issue there is it's not actionable. My company did exactly that, but a dollar from a biz-dev deal is worth less than a dollar from a one-time purchaser is worth less than an ongoing subscriber.

Or money is a proxy for value when value can be measured in those terms.

Money is worthless on it's own. Brings to mind the famous Warren Buffet quote:

"Price is what you pay, value is what you get."

Could not disagree more. Money is a byproduct of the activities you should be focusing on (which are difficult to measure). Former GE CEO Jack Welch is famous for putting it this way: "On the face of it, shareholder value is the dumbest idea in the world. Shareholder value is a result, not a strategy...your main constituencies are your employees, your customers and your products"

I don't understand this whole dichotomy between art and science. Or why people are so insistent that they are mutually exclusive. Science is an art, taking part in this myth hurts both art and science as a whole.

I agree. And most art these days involves quite a bit of science. Photoshop's image processing algorithms aren't just sloppily thrown together, for example! It takes a lot of science to make good artistic tools, and artists often do a lot of science when creating their art. (Checking audio levels, figuring out when the light will be just right, etc.)

I think it helps to learn how to gamble or /play poker properly.

You'll learn what the poker players call "varience".

If you know varience (not patience) then over the long term, you aren't gambling. Or you are gambling but in a way if done properly that will make you money over the long term with 100% success.

Assuming you are better than the other players.

Reply on news.ycombinator.com