[c4mpute]

All these cool-looking dashboards are just too inflexible. You cannot add your own aggragates beyond trivialities. You cannot just "color that one value that bugs you". You cannot just generate a readable report plus some explanatory text.

Spreadsheet export + pivot table gives you all that. Doable for any moderately competent office drone without a round-trip through some endless backlog-spec-sprint-program-test-respec-sprint-... loop

[c4mpute]

To be somewhat constructive: What you rather should have done is not create more elaborate dashboards. What imho the world needs is an easy way to use a spreadsheet tool to generate and publish a dashboard. A "make web dashboard" button right next to the print button. With auto-updates when input data changes of course.

[uxp8u61q]

Have you... used Excel? It's very simple to create any kind of "dashboard" (AKA graphs on a page) and then you just share the web link to the page.

[c4mpute]

Yes, I have. What Excel is still lacking is an easy solution for the input side. You can bind tons of data sources, but all are weird, hard-to-use, manual. There is no easy "grab this from that website, get the current data of what I just pasted there, mash it together, publish it"

Hell, it cannot even do proper CSV import. You need to reformat your CSV to match the locale Excel is running under!

[uxp8u61q]

Uh? Are you sure you've actually used Excel? The CSV import is highly configurable and leads you immediately into Power Query where you can massage the data any way you want.

[c4mpute]

The LibreOffice CSV import is configurable. The Excel one isn't.

You can do things in PowerQuery, but that is far from obvious and still buggy. Not to mention all the woes after import, like date/time auto-interpretation and autocorrections that cannot be switched off.

I stand by what I said. Excel imports are a huge mess.

[uxp8u61q]

Would you like me to send you some online tutorials on how to import CSVs into excel? Because at this point it's just crazy. Are you using excel 2009? Do you not know about the "Data" tab in the ribbon? There's a whole dialog to complete with several options when you import a CSV file.

[maxerickson]

Are you evaluating import or open? Open does stuff to the data right away, import lets you push buttons first:

https://www.fcc.gov/general/opening-csv-file-excel

[Seanambers]

Powerquery oh god never again!

[airstrike]

Yup. That's what I want to build. Thank you for saying that -- I feel like it really validates my feelings hah

[fiddlerwoaroof]

The problem is always this project turns into “let’s build excel or tableau” and the customers that care usually already use one or the other.

[airstrike]

That's fair, but fortunately I'm not planning on doing either. (Well, I am still implementing ~all of Excel's formulas for compatibility, but not the the UI/UX...)

People don't really consume data, they read documents. I think that's (part of) the vision these projects lack.

[amcaskill]

I'm working on an OSS BI tool focused on a document form factor. Might be of interest to you. https://github.com/evidence-dev/evidence

[airstrike]

Thank you! Definitely interesting! I had actually starred that repo when I saw it being discussed on some HN thread a week or two ago

[fiddlerwoaroof]

Yeah, I misread the post I replied to: I’ve been on a bunch of internal dashboard projects that were in danger of losing focus and turning into full-fledged visualization platforms.

[jamesmaniscalco]

There is Smartsheet, which mostly works well for this, but its power-user features are pretty limited compared to Excel.

[Bishonen88]

That's a bit oversimplifying IMO.

There's a place for well-crafted analytics dashboards in today's business, too. They're mostly tailored to specific user requirements/use-cases and look nothing like the flashy stuff one sees on dribbble or elsewhere.

Tailored analytics dashboard can solve many pain points of Excel + Spreadsheets if done well. If ~1k people need to access the same data each day and 'analyze' it for similar things (patterns/outliers/seasonalities etc.) then a good dashboard will be quicker, better and cheaper than 1k office workers trying to create pivot tables. If that dashboard is tailored to the use case, then those 'color that one value that bugs you' can oftentimes be implemented within minutes after hearing a good use-case from a user. I say that from experience.

And from experience, I'd say that most Excel users know the basics of basics. I'd bet that 90%+

[c4mpute]

The problem here is that you usually do not have ~1k users with all the same requirements. You have 200 groups of average 5 users each, all with their own department-specific, country-specific or workflow-specific requirements. Of course a central solution will be better and cheaper. But it will never be quicker, because you will take ages to just gather requirements from all 200 distinct user groups. As soon as you have those requirements, they will have changed already, so you are working on yesteryear's problems.

And of course, given a working system, the users can drop you a quick email, explain their problem (yes, in an ideal world they could do that, and you would understand them right away...) and you implement a 5min change. In reality however, their problem will first have to be specified in a user story, with a ton of clarification requests until the story is really understood by the dev team, then you need goodwill, time and money for the implementation. And maybe their problem can only be solved by an ugly hack, a weird special case for the ternary currency and ages-old lunar-calendar-based tax-system of lampukistan. Would that really be quicker than just the lampukistan team throwing together a few formulas and be done faster than the initial email? Even when multiplied by the special requirements of the other 100 country sales teams?

Also, I've had similar change requests where is was explicitly asked to provide a spreadsheet prototype of what the statistics should look like. Well, thanks, why again do we need a dev team?

I know that spreadsheets suck. They are ugly, undebuggable hacks, always and without exception. You need tons of time to implement in hours what would be a quick one-liner SQL query. With terrible error behaviour, weird edge cases and hell knows how many hidden bugs when the locale uses the lampukistan-currency-separator instead of a decimal dot...

...Except that they provide those office drones with velocity, which, as the usual wisdom around here goes, is everything.

[airstrike]

This is the way

[hackandthink]

Agree.

And a tool like Superset enables users to customize their dashboards and charts.