Great Post! But i wonder ultimately how much willingness anyone will have to use such "versioned analytics" system? In my experience business owners don't really care plus their could be so many confounding factors that these versions on its own might become useless to use for anything. In the end it would be one poor data analyst logging all these decisions and sharing which would be similar to screaming into the void. It sounds useful theoretically, but I am skeptical about its practical use cases.
Eh, I depends on how it's done. Analysts are often answering these questions anyway; spending a little bit of extra time recording that could slow things down in the short term to speed them up in the long term.
I don't think it needs to be particularly onerous, either. If data teams answered questions in a tool like Stack Overflow, it'd do a better job of logging answers and making them discoverable, than answering everything in Slack or email. And if you have that history, people start to find answers there, you can build off of those things, and so on.
Agreed that it's somewhat theoretical, but I think it's doable.
This is awesome. You are so spot on with "We excuse our lack of organization because it’s defined by its disorganization". To combat this, I have been keeping a running google doc for each squad with each ad-hoc request I get. I include the question, the objective, the results and reccos, and links to code in Github. This certainly doesn't solve the problem of sharing/storing results via a central repository but it does allow you to keep records of what you have learned so far, so you can remind stakeholders when they ask you the same question again :). It comes in handy come review time so you can remind your manager of all the knowledge you have developed for your squads. Finally, It will also make inputting results to a hypothetical central repo easier if that time comes :)
Great post. It made me think of Ursula Le Guin and the power that comes from knowing the true name of things in her EarthSea stories. "Ad-hoc analysis" would be just a use name, and we are still searching for its true name.
Maybe it is a name that reminds us that what it finds is itself precious, and therefore should not be discarded? Ben Shneiderman once said that "the purpose of visualizations is insight, not pictures". And visualizations are a big part of exploratory data analysis, so the theme fits. "Insight analysis" doesn't rolls out of the tongue, but how about "insight discovery"?
The idea of naming it to "remind of why it's precious" is interesting. It's kind of shame that "insight" has been co-opted by marketing language though (like "platform"). It feels like the right word, but I can't hear it anymore without assuming it's just vague marketing fluff.
Great Post! But i wonder ultimately how much willingness anyone will have to use such "versioned analytics" system? In my experience business owners don't really care plus their could be so many confounding factors that these versions on its own might become useless to use for anything. In the end it would be one poor data analyst logging all these decisions and sharing which would be similar to screaming into the void. It sounds useful theoretically, but I am skeptical about its practical use cases.
Eh, I depends on how it's done. Analysts are often answering these questions anyway; spending a little bit of extra time recording that could slow things down in the short term to speed them up in the long term.
I don't think it needs to be particularly onerous, either. If data teams answered questions in a tool like Stack Overflow, it'd do a better job of logging answers and making them discoverable, than answering everything in Slack or email. And if you have that history, people start to find answers there, you can build off of those things, and so on.
Agreed that it's somewhat theoretical, but I think it's doable.
Saw a typo quoting a newyorker article from the future :) 2103! Great read and still so relevant.
Sticking with the Latin, revelatio (unveiling). Great post Benn. “Words mean things”. A phrase I tire of hearing myself say.
Thanks Tim!
This is awesome. You are so spot on with "We excuse our lack of organization because it’s defined by its disorganization". To combat this, I have been keeping a running google doc for each squad with each ad-hoc request I get. I include the question, the objective, the results and reccos, and links to code in Github. This certainly doesn't solve the problem of sharing/storing results via a central repository but it does allow you to keep records of what you have learned so far, so you can remind stakeholders when they ask you the same question again :). It comes in handy come review time so you can remind your manager of all the knowledge you have developed for your squads. Finally, It will also make inputting results to a hypothetical central repo easier if that time comes :)
I like “deep analysis”
When we do deep analysis, do we...deep learn?
(Jokes aside, I like this a good bit.)
Great post. It made me think of Ursula Le Guin and the power that comes from knowing the true name of things in her EarthSea stories. "Ad-hoc analysis" would be just a use name, and we are still searching for its true name.
https://notionparallax.co.uk/2020/the-true-name
Maybe it is a name that reminds us that what it finds is itself precious, and therefore should not be discarded? Ben Shneiderman once said that "the purpose of visualizations is insight, not pictures". And visualizations are a big part of exploratory data analysis, so the theme fits. "Insight analysis" doesn't rolls out of the tongue, but how about "insight discovery"?
https://interactions.acm.org/blog/view/the-purpose-of-visualization-is-insight-not-pictures-an-interview-with-ben
The idea of naming it to "remind of why it's precious" is interesting. It's kind of shame that "insight" has been co-opted by marketing language though (like "platform"). It feels like the right word, but I can't hear it anymore without assuming it's just vague marketing fluff.