5 ms·

How Statwing (YC S12) Makes It Easier To Ask Questions About Data

67 points by glaugh 14y ago

bryanh 14y ago
The little time I got to play with Statwing my impression was it is really intuitive and powerful, but IMO the biggest hurdle is getting your data into the app (or in this case, into a CSV first). Data sits all over the place, if you guys can reach out and snag it for me after a quick OAuth dialog, that would be a killer feature. I'd use it a lot, lot more.
- DrJokepu 14y ago
  Or have the ability to connect to the local corporate Oracle database and get the data from there directly?
- ernestipark 14y ago
  You mean like http://filepicker.io http://filepicker.io? :)
- glaugh 14y ago
  Appreciate that feedback. Sounds like getting data in is the theme of this whole page of discussion, too. We agree. Hooking into existing sources of data will probably always need to be a big priority for us. Certainly not an easy challenge, but definitely solvable, bit by bit.
  - kiyoto 14y ago
    Does Statwing have an API? If so, I am willing to write a plugin for Fluentd, a versatile log collector (Disclaimer: I am one of the active contributors to Fluentd)
jgamman 14y ago
my pain point over the last few years has been pulling data out of pdf documents (ie govt departments think this counts as open data). I'd pay through the nose for a tool that i could just dump pdf files into and have them find tables and convert them to csv. even if it was only 95+% that would save hours of tedious data entry...
- _delirium 14y ago
  You might already know this, and it's not as automated as just dropping in a file, but one option for extracting tables from PDFs (other than retyping them) is the pro version of Adobe Acrobat. If you drag-select around a table, you can right-click to export it as Excel or CSV, which works most of the time.
pyoung 14y ago
I am currently working with some govt data. Not fun. If it ever got to the point that I could log in and pull some summary stats from a variety of clean, public data sources, this could be very useful. For me the issue isn't the complexity of SPSS (or in my case SAS and R), it is getting the data ready for analysis. Obviously I wouldn't expect Statwing to be able to do that to my own data, but for public data sets, which I occasionally use, it would save me a ton of time
- thronemonkey 14y ago
  This is something that a favorite instructor of mine has stressed to me—the most difficult part of data analysis (and the part in which mistakes are most often made) is not choosing the correct analysis to do, nor actually performing it, but rather manipulating the data into the necessary shape/format. My favorite utilities for this are Hadley Wickham's plyr and reshape (http://had.co.nz/ http://had.co.nz/) (not sure if you already use these, figured I'd drop you a hint if not)
  - kylemaxwell 14y ago
    Take a look at Google Refine; for some use cases, it's just fantastic.
jmitcheson 14y ago
The Statwing homepage looks really nice. I think they did a good job of something quite hard: making a data analysis tool look interesting!
jmduke 14y ago
Statwing is cool, but IMHO it comes from people who have never struggled with the problem it ostensibly solves. Ask anyone who uses SAS/Stata/SPSS what their pain point is, and I guarantee its not 'analyzing data nicely so I can look at charts instead of SPSS outtables'. The hardest part is collecting and organizing data (which I'm sure is a market Statwing is planning on entering.) Point-and-click stats software is nice, but the people who are using SPSS and Stata weren't hired because they could use SPSS and Stata. (Also, I think its pretty disingenuous to compare SPSS output with Statwing output, because it implies SPSS can't output a similar histogram -- which, of course, it can.)
- kylemaxwell 14y ago
  I don't think it's for people who need SPSS. Statwing is great for much simpler use cases. When you need statistical heavy lifting, go to SPSS or R or something else that can do all that, but if you're not a statistician / data person and have simpler needs, things like Statwing (and, presumably, Wolfram Alpha Pro) can do a decent job.
- glaugh 14y ago
  Cofounder here. Totally agree that the hardest thing about data analysis is getting the data together. But it's still true that once you have data, working with SPSS, R, etc. is a nightmare unless you're trained or somewhat technical. I actually did work with SPSS professionally for a couple years. And I was a lot like the roughly half of SPSS users who are nontechnical, and only need the basic analyses. The two biggest painpoints for these users are knowing what test to run when and the difficulty of interpreting and working with the output. There's a ton of people, particularly more technical folks, for whom those are not big pain points, and that's fine too. They're just not in our target audience. Also, RE the output comparison. Valid point if it looked like the implication was that SPSS can't output a histogram. Our point, though, is that we do it by default in the three clicks that it takes to output text, and it takes a lot more work to output even T-Test results in SPSS. Practically speaking, day-to-day nontechnical users don't visualize much in tools like SPSS because it's painful to do so. Thanks for the comment, really appreciate it.

67 points by glaugh 14y ago

The little time I got to play with Statwing my impression was it is really intuitive and powerful, but IMO the biggest hurdle is getting your data into the app (or in this case, into a CSV first). Data sits all over the place, if you guys can reach out and snag it for me after a quick OAuth dialog, that would be a killer feature. I'd use it a lot, lot more.

DrJokepu 14y ago
Or have the ability to connect to the local corporate Oracle database and get the data from there directly?
ernestipark 14y ago
You mean like http://filepicker.io http://filepicker.io? :)
glaugh 14y ago
Appreciate that feedback. Sounds like getting data in is the theme of this whole page of discussion, too. We agree. Hooking into existing sources of data will probably always need to be a big priority for us. Certainly not an easy challenge, but definitely solvable, bit by bit.
- kiyoto 14y ago
  Does Statwing have an API? If so, I am willing to write a plugin for Fluentd, a versatile log collector (Disclaimer: I am one of the active contributors to Fluentd)

jgamman 14y ago

my pain point over the last few years has been pulling data out of pdf documents (ie govt departments think this counts as open data). I'd pay through the nose for a tool that i could just dump pdf files into and have them find tables and convert them to csv. even if it was only 95+% that would save hours of tedious data entry...

_delirium 14y ago
You might already know this, and it's not as automated as just dropping in a file, but one option for extracting tables from PDFs (other than retyping them) is the pro version of Adobe Acrobat. If you drag-select around a table, you can right-click to export it as Excel or CSV, which works most of the time.

pyoung 14y ago

I am currently working with some govt data. Not fun. If it ever got to the point that I could log in and pull some summary stats from a variety of clean, public data sources, this could be very useful. For me the issue isn't the complexity of SPSS (or in my case SAS and R), it is getting the data ready for analysis. Obviously I wouldn't expect Statwing to be able to do that to my own data, but for public data sets, which I occasionally use, it would save me a ton of time

thronemonkey 14y ago
This is something that a favorite instructor of mine has stressed to me—the most difficult part of data analysis (and the part in which mistakes are most often made) is not choosing the correct analysis to do, nor actually performing it, but rather manipulating the data into the necessary shape/format. My favorite utilities for this are Hadley Wickham's plyr and reshape (http://had.co.nz/ http://had.co.nz/) (not sure if you already use these, figured I'd drop you a hint if not)
- kylemaxwell 14y ago
  Take a look at Google Refine; for some use cases, it's just fantastic.

jmitcheson 14y ago

The Statwing homepage looks really nice. I think they did a good job of something quite hard: making a data analysis tool look interesting!

jmduke 14y ago

Statwing is cool, but IMHO it comes from people who have never struggled with the problem it ostensibly solves. Ask anyone who uses SAS/Stata/SPSS what their pain point is, and I guarantee its not 'analyzing data nicely so I can look at charts instead of SPSS outtables'. The hardest part is collecting and organizing data (which I'm sure is a market Statwing is planning on entering.) Point-and-click stats software is nice, but the people who are using SPSS and Stata weren't hired because they could use SPSS and Stata. (Also, I think its pretty disingenuous to compare SPSS output with Statwing output, because it implies SPSS can't output a similar histogram -- which, of course, it can.)

kylemaxwell 14y ago
I don't think it's for people who need SPSS. Statwing is great for much simpler use cases. When you need statistical heavy lifting, go to SPSS or R or something else that can do all that, but if you're not a statistician / data person and have simpler needs, things like Statwing (and, presumably, Wolfram Alpha Pro) can do a decent job.
glaugh 14y ago
Cofounder here. Totally agree that the hardest thing about data analysis is getting the data together. But it's still true that once you have data, working with SPSS, R, etc. is a nightmare unless you're trained or somewhat technical. I actually did work with SPSS professionally for a couple years. And I was a lot like the roughly half of SPSS users who are nontechnical, and only need the basic analyses. The two biggest painpoints for these users are knowing what test to run when and the difficulty of interpreting and working with the output. There's a ton of people, particularly more technical folks, for whom those are not big pain points, and that's fine too. They're just not in our target audience. Also, RE the output comparison. Valid point if it looked like the implication was that SPSS can't output a histogram. Our point, though, is that we do it by default in the three clicks that it takes to output text, and it takes a lot more work to output even T-Test results in SPSS. Practically speaking, day-to-day nontechnical users don't visualize much in tools like SPSS because it's painful to do so. Thanks for the comment, really appreciate it.