When I worked for SAP back in 2007 (I was a fresh grad at the time), I was working in the business intelligence (reporting, analytics, and data warehousing) group and noticed how cumbersome it was for organizations to simply create and view reports (we're talking millions of dollars). I once said to my boss "you realize that in the future we'll simply just write 'show me a line graph for sales in the northeast'".
I'm currently evaluating BI vendors for my company and just about every major contender has this functionality or is developing it. Read the Gartner report from 2017 and they basically just come out and say that this will be the new standard in BI.
The much harder problem is data management and preparation. Anyone with half a brain and a decent visualization tool can create basic graphs - but that doesn't mean they should, especially if the organization doesn't have good data management processes in place.
Issues like data governance, data prep, and data modelling are the major pain points for me. And honestly, developing a useful BI solution is more about culture change than it is technology. If a company has poor data governance, it doesn't matter how whiz-bang their technology is, they're still not going to get useful insight from their numbers.
The problem is that this approach requires well structured data and most corporations have data all over the place, in legacy systems or some half-assed backed solution that kind of works. So even if you provide them with some state of the art platform that pulls graphs out of thin air they'd still need shitloads of consulting and custom made tinkering to make it work.
The other problem is that most executives don't even know what kind of graphs/reports they need. Sure, they can have the classic sales by region and stuff, but the really important things are hidden and require a lot of digging and asking to get to.
So many of our clients (I'm still working in the BI space) are still using Excel workbooks and sending them to each other over email. I think it will take quite some time before we see any major improvements in this area. But I agree with you. Some of the stuff I see makes me cringe.
Watching the gif on that page, it actually seems slower to use. Instead of just drag selecting the table, I have to write out all the header text now by hand.
How about we just 'say' it ? At one hackathon, I had the idea of making an Alexa skill that would translate simple voice commands like 'graph column A' into matplotlib functions and then show the graphs. Teammates weren't too excited about spreadsheets though.
I played around with this the other day. I have a spreadsheet with a bunch of columns. It wasn't immediately obvious how to use the explore feature intuitively. It graphed data but not really the ones I wanted. I was also hampered by it using only about 200 pixels on the right side of the screen.
I started typing in a question but it couldn't guess what I was interested in. YMMV. Perhaps with a fairly simple spreadsheet you can intuit things? Back 10 years ago I built a google spreadsheet competitor called Numbler (well, I didn't know if was a competitor, google sheets came out a couple of months later). But one of the things I learned is that people use spreadsheets for just about everything, and it can be in the wierdest format.
"people use spreadsheets for just about everything"
That is so true. I used to do consulting for small businesses and was amazed at how non-programmers used Excel to solve so many problems outside of finance and accounting. The CRM/HR solutions were very common and interesting (e.g. Lead/prospect management, sales, timesheets, vacation schedules, etc.).
Can we talk about getting data into Google Sheets? Is there a standard way to build a pipe from, say, a reporting database to dump aggregates into Google Sheets?
I built a private Add-on for my company that surfaces specific aggregates as Sheets functions (i.e. getSalesByDay(...)) and I have found so many bugs with that whole ecosystem. Deploys are completely manual and require copy-paste, you can't reliably tell what version is being invoked in a sheet, invisible cell-level caching that caches error state, concurrency limits that are too low and impossible to work around, and more. It all kinda sorta works but Google doesn't make it easy.
Yes, so much this. I've encountered this same issue.
For our non-technical employees I write an endpoint that gives them a CSV with the most current info (often with a super simple front-end that they can use to query for specific date ranges or with filters). They download this CSV and upload it into their sheet manually to update the underlying data when they need new info.
It would be so great if you could just say "here's a URL; keep my data fresh from this source" and it would automatically do it.
what kind of DB was the reporting database if you don't mind me asking? Seems like translating the data into Google Sheets is unnecessary if you're just trying to graph/chart it, although as I say that I'm realizing there may not be a tool that could just graph it directly that's easy enough to use/cheap enough.
Cool stuff - someone else mentioned Thoughtspot - that was my initial thought as well. Very similar idea.
I wish that Google would take the same sort of "embed" idea further in G-Suite. I find it amazing that I can't (as far I know) reference slides from another deck in Google Slides. The use case would be putting together a series of "core" slides that are updated across your organization as they change. Given the web nature of G-Suite, this, to me, would seem like a no brainer.
Also, inserting charts from Google Sheets into Google Presentations looks pretty terrible. I often revert to Excel because the charting is fair superior imho (though just as challenging to wrangle).
They are solving a problem that doesn't really exist, the challenge is not the last step of a data report, it's the steps involved in the beginning, getting good data in, formatting, joining multiple sources, automation, dealing with junk data, procedures,etc.
I don't understand the example, what's the difference between typing "Show me a line graph" and clicking a button in excel that does the same thing.
Did you try it? Yes, getting good data is a challenge but that's not one they can solve in Sheets. Making a graph is also not always one button, unless you have very simple data.
I found it immediately useful. They've solved some UX and discovery issues around creating charts. And it's not _just_ charts... they're answering questions and identifying trends within the data. e.g. I threw some pretty basic data at this and it told me “Flights” contains a yearly cycle: “Flights” increases until May 1, decreases until October 1, and increases until December 1.
I tried to rely on your first bullet point and randomly end up with BigQuery surpassing some API treshold towards Drive (paying customer). Have to carefully manage doing copy-queries over to BQ, which is a pain, but better than nothing I guess.
Oh, wow. I love where this is headed. Spreadsheets are one of the most abused products in a normal business--used for everything, and then some poor excel jockey ends up being forced to create a semblance of order from the chaos.
That's because spreadsheets are an excellent tool for prototyping, and developers have shitty tools for building robust products from a prototype.
If developers had a quick way to create properly engineered applications from a working workflow involving a spreadsheet, it would be easy to define most company processes on Excel or Google Docs and turn them into solid software.
But building an application that is functionally equal to the spreadsheet, just with robust engineering practices, typically involves several-months-long projects with many developers and managers, which is expensive.
Someone needs to create Spreadsheets Anonymous and share the crazy stuff people do in Excel, the best one I've seen
is a full GUI wizard (back, next etc.) in VBA.
I'm at a 250 EE company with very small ops budget and I built a stable and modular account planning and CRM system (when we had 150 EEs) that uses structured + validated + protected Google Sheets for user inputs and some viz, Alteryx as the ETL, and Tableau (Reader) as the visualization. Took half a junior person and half a senior person to maintain it for 100 clients and 200 users. It has served well as a prototype for our Salesforce installation this year, which has enabled us to decommission half of the modules in the Google Sheets system. If we had gone straight from nothing to Salesforce it would have induced shock in most of our employees, and our data and processes would not have been ready to just drop into Salesforce. In my mind, that is the best use case for a spreadsheet 'app' ... build it from the start with the goal of it being a transitional tool to a 3rd party app, nudging users along the way to professional processes.
Well charts is a good addon but just wanted to understand how they are able to do this ... i mean Machine Learning part , for example if somebody asks "Show me sales of X product in last year" , from machine learning perspective how this gets interpreted in actual SQL query ..
Presumably some ad engineers got sick of producing arbitrary customer reports, took the corpus of all customer requests, and tied it to the sheets that generated those reports. And a ton of glue code to make it all work.
I'm wondering how Microsoft is responding to this. Do they expect their current Excel dominance to continue despite competitors constantly catching up to feature parity and even extra goodies, like this one?
You don't shop for Excel, you shop for Office. The contracts around O365 are pretty onerous to get out of if you're an EA customer, and the process of going away is full of pitfalls and surprises.
Google used to say that everyone who did an RFP chose Google Apps/GSuite. The problem is that nobody does that!
I wonder if we will see more software including query based input like their charts, and what sort of speed improvement we could see? At first I was not excited to type something where I could click a couple buttons, but then I recognized the other enhancements such as applying a filter right away.
I'm not convinced it's better just because it has machine learning on the back end, but if excel would learn how I want my graphs made from how I manually adjust the graphs (adding axis labels and a title, color preferences, never a 3d bar or pie chart), that'd be a nice enhancement. I'm sure there's a setting, but I haven't searched for it.
There's a lot of basic stuff like column titles, moving columns about, filtering, search that I found had a quite a learning curve with sheets. I built and use this instead. Bell+Cat https://bellpluscat.com
does anyone know how this kind of stuff gets built ? I'm considering a spreadsheet-y internal admin dashboard for my startup. I was looking at https://github.com/JoshData/jot to be able to sync stuff on the client side to the server.
has anyone worked on something like this ? the big challenge is synchronization - between server and multiple clients - while being able to offload a lot of computations on to the client.
I wonder how is the security built ? if i maliciously change the formulas in my browser.. will the backend datastore still accept the data ?
This sort of thing it's easy to get to 80% but good luck getting that last 20% without formalisms. Might be useful for getting a quick feel for a data set to confirm some intuitions, but not really useful beyond that.
[+] [-] mbesto|8 years ago|reply
And so here we are now.
[+] [-] tryitnow|8 years ago|reply
The much harder problem is data management and preparation. Anyone with half a brain and a decent visualization tool can create basic graphs - but that doesn't mean they should, especially if the organization doesn't have good data management processes in place.
Issues like data governance, data prep, and data modelling are the major pain points for me. And honestly, developing a useful BI solution is more about culture change than it is technology. If a company has poor data governance, it doesn't matter how whiz-bang their technology is, they're still not going to get useful insight from their numbers.
[+] [-] elorant|8 years ago|reply
The other problem is that most executives don't even know what kind of graphs/reports they need. Sure, they can have the classic sales by region and stuff, but the really important things are hidden and require a lot of digging and asking to get to.
[+] [-] MitjaBezensek|8 years ago|reply
[+] [-] simo7|8 years ago|reply
NLP interfaces in reporting work for simple things, when you want to start doing more advanced things they become incredibly verbose and inconsistent.
That's where a well designed UI really shines...and will continue to do so.
[+] [-] hacker_9|8 years ago|reply
[+] [-] Nydhal|8 years ago|reply
How about we just 'say' it ? At one hackathon, I had the idea of making an Alexa skill that would translate simple voice commands like 'graph column A' into matplotlib functions and then show the graphs. Teammates weren't too excited about spreadsheets though.
[+] [-] spinlock|8 years ago|reply
[+] [-] harshaw|8 years ago|reply
I started typing in a question but it couldn't guess what I was interested in. YMMV. Perhaps with a fairly simple spreadsheet you can intuit things? Back 10 years ago I built a google spreadsheet competitor called Numbler (well, I didn't know if was a competitor, google sheets came out a couple of months later). But one of the things I learned is that people use spreadsheets for just about everything, and it can be in the wierdest format.
[+] [-] neovive|8 years ago|reply
That is so true. I used to do consulting for small businesses and was amazed at how non-programmers used Excel to solve so many problems outside of finance and accounting. The CRM/HR solutions were very common and interesting (e.g. Lead/prospect management, sales, timesheets, vacation schedules, etc.).
[+] [-] teej|8 years ago|reply
I built a private Add-on for my company that surfaces specific aggregates as Sheets functions (i.e. getSalesByDay(...)) and I have found so many bugs with that whole ecosystem. Deploys are completely manual and require copy-paste, you can't reliably tell what version is being invoked in a sheet, invisible cell-level caching that caches error state, concurrency limits that are too low and impossible to work around, and more. It all kinda sorta works but Google doesn't make it easy.
[+] [-] yeldarb|8 years ago|reply
For our non-technical employees I write an endpoint that gives them a CSV with the most current info (often with a super simple front-end that they can use to query for specific date ranges or with filters). They download this CSV and upload it into their sheet manually to update the underlying data when they need new info.
It would be so great if you could just say "here's a URL; keep my data fresh from this source" and it would automatically do it.
[+] [-] jboggan|8 years ago|reply
[+] [-] guiomie|8 years ago|reply
[+] [-] tommynicholas|8 years ago|reply
[+] [-] rrggrr|8 years ago|reply
[+] [-] inthewoods|8 years ago|reply
I wish that Google would take the same sort of "embed" idea further in G-Suite. I find it amazing that I can't (as far I know) reference slides from another deck in Google Slides. The use case would be putting together a series of "core" slides that are updated across your organization as they change. Given the web nature of G-Suite, this, to me, would seem like a no brainer.
Also, inserting charts from Google Sheets into Google Presentations looks pretty terrible. I often revert to Excel because the charting is fair superior imho (though just as challenging to wrangle).
[+] [-] wyck|8 years ago|reply
I don't understand the example, what's the difference between typing "Show me a line graph" and clicking a button in excel that does the same thing.
[+] [-] Nagyman|8 years ago|reply
I found it immediately useful. They've solved some UX and discovery issues around creating charts. And it's not _just_ charts... they're answering questions and identifying trends within the data. e.g. I threw some pretty basic data at this and it told me “Flights” contains a yearly cycle: “Flights” increases until May 1, decreases until October 1, and increases until December 1.
That's pretty handy. YMMV, but this is awesome.
[+] [-] SubiculumCode|8 years ago|reply
[+] [-] vgt|8 years ago|reply
You can:
- Query data in Google Sheets from BigQuery
- Create virtual views in BigQuery that are powered by Google Sheets
- One-click export data from BigQuery to Google Sheets (< 20k rows or so)
- Using AppScript, build dashboards and reports in Sheets that query BigQuery for results.
(work at G)
[+] [-] dagss|8 years ago|reply
[+] [-] taylorwc|8 years ago|reply
[+] [-] TuringTest|8 years ago|reply
If developers had a quick way to create properly engineered applications from a working workflow involving a spreadsheet, it would be easy to define most company processes on Excel or Google Docs and turn them into solid software.
But building an application that is functionally equal to the spreadsheet, just with robust engineering practices, typically involves several-months-long projects with many developers and managers, which is expensive.
[+] [-] cube00|8 years ago|reply
[+] [-] cepetersen|8 years ago|reply
[+] [-] uberneo|8 years ago|reply
[+] [-] killjoywashere|8 years ago|reply
[+] [-] spinlock|8 years ago|reply
The definition of machine learning that I use is: an algorithm which improves its performance through experience.
So, if charts doesn't get better the more you use it, it's not machine learning.
[+] [-] Raphael|8 years ago|reply
[+] [-] andrea_s|8 years ago|reply
[+] [-] fudged71|8 years ago|reply
[+] [-] amelius|8 years ago|reply
Does it perform NLP on company documents?
[+] [-] pbreit|8 years ago|reply
But the best, still-mostly-hidden feature I've found recently is App Scripting and especially the ability to do a UrlFetch.
I use it as an "API Runner" to run various batch jobs against APIs.
https://developers.google.com/apps-script/reference/url-fetc...
[+] [-] darwhy|8 years ago|reply
[+] [-] tryitnow|8 years ago|reply
It's fairly trivial to get this to work with well-formatted data.
I'm currently evaluating BI solutions for my company and just about every single one has something like this.
[+] [-] MitjaBezensek|8 years ago|reply
[+] [-] Spooky23|8 years ago|reply
Google used to say that everyone who did an RFP chose Google Apps/GSuite. The problem is that nobody does that!
[+] [-] intrasight|8 years ago|reply
https://powerbi.microsoft.com/en-us/documentation/powerbi-se...
[+] [-] froindt|8 years ago|reply
I'm not convinced it's better just because it has machine learning on the back end, but if excel would learn how I want my graphs made from how I manually adjust the graphs (adding axis labels and a title, color preferences, never a 3d bar or pie chart), that'd be a nice enhancement. I'm sure there's a setting, but I haven't searched for it.
[+] [-] zitterbewegung|8 years ago|reply
[+] [-] fakename|8 years ago|reply
[+] [-] f00_|8 years ago|reply
You would message the bot something like "sessions for this month", and it would send back a graph.
wonder if you could make a similar bot with google sheets if they provide an api
https://statsbot.co/slack https://medium.com/slack-developer-blog/bots-you-can-count-o...
[+] [-] gsvclass|8 years ago|reply
[+] [-] aeorgnoieang|8 years ago|reply
[+] [-] sandGorgon|8 years ago|reply
has anyone worked on something like this ? the big challenge is synchronization - between server and multiple clients - while being able to offload a lot of computations on to the client.
I wonder how is the security built ? if i maliciously change the formulas in my browser.. will the backend datastore still accept the data ?
[+] [-] blazespin|8 years ago|reply
[+] [-] shostack|8 years ago|reply
[+] [-] synaesthesisx|8 years ago|reply
[+] [-] KaoruAoiShiho|8 years ago|reply