My company shares a lot of customers with Looker, and from my point of view there’s a couple reasons why they’re so successful.
First, Looker has two layers. The dashboard layer gives you tables, charts and filters. The explore layer allows you to go in and build your own pivot tables off the same underlying models. No other BI tool has built an explore layer that’s as successful in practice as Looker’s.
Second, the underlying model of both layers is written in code and under version control. This probably makes sense to most HN readers but in the BI world this is very contrarian. If you ask most business users who make BI purchasing decisions, version control and pull requests are the last thing they’ll ask for. But it turns out it’s exactly what you need to maintain a complex BI deployment.
The version control and pull request design is exactly why I've selected Looker. LookML enables the power of the explore functionality as well.
Looker's visualizations, however, are very weak when compared to the competition. It's been their weakness for years now, and they've really failed to address it.
One aspect of Looker that I've also generally enjoyed is their on-boarding and support processes. They do a superb job at this and I feel like they should be a model for startups selling into B2B.
I agree that their viz is weak. I comfort myself by saying that this is the easy part and they'll surely fix it eventually...still waiting after two years.
Tableau is the best at getting non-technical users to be instantly productive, and is completely unmatched in pure analytical abilities and visualizations. They aren't standing still here either.
However Tableau has absolutely screwed up their evolution into a cloud-based service. The current single-user desktop software with client/server design is painful and archaic, and the Tableau cloud is just a hosted server run on undersized hardware. They are losing a lot of business renewals because of this, even if the consultancy and enterprise deals will keep going for awhile.
> Tableau is the best at getting non-technical users to be instantly productive
I keep hearing that this is Tableau's big selling point. And yet, as exactly the kind of "non-technical" user coming from the world of Excel, it has been astounding to me just how user-unfriendly Tableau is.
Want to create a line chart? No clear way of doing it other than messing around with Tableau's implementation of pivot tables, randomly dropping metrics and dimensions into boxes until it resembles a chart.
Want to not guess blindly and follow Tableau's tutorials? Prepare to sign in on every single video tutorial page. Opened multiple tabs? You'll have to sign in on each one. Refreshing the page won't do anything.
I have no idea how they compete with something like PowerBI.
There are 2 axis in the chart area or at the top of the window where you can just drag the fields you're interested in. It automatically makes a chart type that fits and you can change it with a click. Did that not work for your data?
PowerBI has no Mac app and supports less data sources, with all the same frustrations of client/server design.
Something I want to complain about here is the complete lack of due diligence that product-based, enterprise-focused companies fail at as they move to SaaS. If your primary customers are Federal/Finance/Health Care, you can't pull the plug on a product that is on prem and put it in the cloud until you get the compliance ducks in a row. I had hoped that GDPR would have fixed that but I am still seeing several companies making the same mistakes even today where this playbook should have been figured out.
Sure, but I don't think that's the problem here. Tableau's architecture is client/server using "workbook" files instead of a central server accessible via the web. They need to change that first, then they can make on-prem and hosted versions of it.
Technically speaking, what's a good way to connect to data sources for cloud apps? I'd be nervous to let cloud apps [or hackers thereof] to slurp arbitrary bits of data through JDBC. Quite a bit more comfortable having the connection happening from on-prem workstations.
Not really a threat. this article is mostly about Looker, which is a very different product with a significantly higher entry level price tag. Tableau on the other hand is part of the curriculum in many information science programs in colleges, and is significantly embedded in the enterprise sector. it will take significant stagnation on Tableua's part or a black swan type of event to dislodge them any time soon.
"it will take significant stagnation on Tableua's part"
As a former customer of Tableau, it feels like this has already happened right? I worked at a company where many teams used it, and there were zero defenders of Tableua when it came time to decide whether to renew...it's a mess, especially on Mac.
Any tool that has a purely web based design UI is not limited to an OS, so the alternatives would have to supply a native Mac design tool as Tableau does.
We are working on a cross-platform solution at Chartly. We are still early but can already create dashboards more quickly than what I've seen Tableau do. See https://chart.ly/ for details! :-)
my workplace uses both Cognos and Tableau. compared to cognos, tableau is a dream to use. There's some non overlapping functionality in cognos, but where tableau can be used it is greatly preferred. Coming from that "traditional" enterprise BI, tableau feels anything but stagnant. Spotfire seems like the closest competitor, but its barrier for entry went up sharply after acquisition. Power BI has a ways to go (but I think given the tech stack behind MS is the biggest threat to tableau as it matures long term)
We're going through the process of getting a BI solution. There are many players and Tableau doesn't fair well against them. I can see them holding on to existing customers because switching is hard, but they absolutely have a lot of competition for new sales.
Hello Tensor! Will you like to review what our team has built at Holistics (www.holistics.io)? We are listed in the Gartner FrontRunner Quadrant for Business Intelligence Oct 2018.
Love to get your feedback on how you find us relative to the tools you are currently evaluating! Send me a note at v(@@)holistics.io or our team at hello(@@)holistics.io]!
Would you share the ones you found worthwhile? Our business wants PowerBI, but the sys admin side sounds like a pain : it seems we need:
1) a consultant to setup the visualisations
2) either a vm running windows server or create a db on azure to update manually because powerbi doesn't allow aws databases
3) the api automation capabilities are unclear says the expert (eg: create a new dashboard for a new customer ? Update all the visualisations through a script? Etc)
PowerBI feels so Microsoftish, I'm really on the market for something else
> 2) either a vm running windows server or create a db on azure to update manually because powerbi doesn't allow aws databases
I believe this is a misunderstanding. You can totally connect to AWS databases with PowerBI, assuming your AWS database is routable. You can't set up the connection through the web interface, but you can use PowerBI Desktop to set up virtually any database that has an ODBC driver as a source[1]. And once you publish it to the Azure-hosted PowerBI instance, it'll use that connection string to connect directly to the data source and automatically update/refresh like any other data source.
As for a VM running Windows Server - that's the option if your AWS database isn't routable by the Azure-hosted PowerBI service. You install a gateway for PowerBI to hop through, and configure all of the data sources within that gateway. That way your internal AWS databases are not publicly routable, and you can lock down and centralize the connection between PowerBI and AWS via that gateway box[2]. Data Gateway is a Windows-only application at the moment, so that part is a pain. But it's a relatively minor pain in the grand scheme of BI implementations.
Yes exactly, so the only way to have a non public database is to run a Windows vm on AWS, I would rather avoid that, especially because we need one vm per customer.
Thanks for the detailed write up!
Is there a reason you need one VM per customer, and can't use a single Data Gateway to manage multiple data connections?
On that note, you don't have to use the official data gateway application. It certainly has a lot of ancillary benefits, but isn't critical. You can just as effectively create a bastion linux host and use ssh to remote forward the connection to AWS database[1]. I've done that plenty of times in the past, to access databases like Redshift and Dynamo that were within non-public VPCs. Easier to manage and cheaper to run per client, if that's a hard requirement.
You could also use something like QuickSight, if you're purely AWS. But I'd only recommend that if you expect to allocate a large amount of development resources for support. PowerBI is a lot more user friendly for business users to be self-sufficient, and sometimes it's less of a hassle to deal with the up-front complexity of that than to deal with the non-stop application support needs of business users.
I'm currently helping a large company with their transition from Tableau/SAP/Excel to Power BI.
The adoption has been extremely painless and business users are very comfortable creating reports and discovering new ways to represent their data.
In response to your points above...
1) A consultant is not required to set up visualisations. With minimal training (a few hours) most users understand the fundamentals and are ready to go.
2) We use AWS Redshift. You have the option to import data into Power BI or do a direct query from the cloud platform to Redshift with odbc.
3) what are you trying to do with the API?
To automate reports you normally set up a gateway connection to your data source and schedule the refresh in powerbi cloud. No API required normally.
It also can use office 365 or Azure AD for single sign on, making sys admin activities a breeze.
Send me a message if you want to discuss anything specific and I'd be happy to help you out.
Our business users already hired the consultant and I'm trying to make sense of his answers, but he does seem to be knowledgeable only on using a mouse.
We want to create a template dashboard of visualisations, which would be duplicated for each of our customers, each one having its own data source and workspace. New customer creation should be automated through API calls and we would prefer centralize our resources on AWS. That's it.
The consultant has been so fuzzy that it seemed utterly complex to set-up, but by reading all the answers things are starting to make sense. Redshift would be great, thanks for the pointer!
PowerBI is really an amazing for the price point (free-ish for O365 accounts) for most businesses out there (10~300 employees). It's not one size fits all and YMMV.
1) You have this problem regardless
2) Huh? Why not just the cloud version? Are you trying to report directly on a productive transactional database for a software app? See #3 about source of data.
3) Depends on your source data. I build a dashboard for sales in with data from SalesForce in like 2 hours.
1) nope, it's a 100% self service tool, no outside assistance needed.
2) PowerBI uses AWS just fine, you may need a gateway depending on how you expose your AWS DB for external access, but that'd be true of any tool connecting to it.
3) why an expert? Here's the API spec, it's not hiding ...
Many thanks for the API link, the expert was completely unclear on what may be possible on the automation side (full api configuration without using a mouse).
Bit of anecdata: I used Looker in a previous gig, where we used it mostly a development tool for constructing and optmising difficult queries. The business continued to use Tableau for BI.
Do you mean that as Tableau's cost? Not where I'm at. And maybe you'd have to pay half a mil to implement a data model if you're hiring consultants and starting from scratch, but if you're in such a state that's going to be a cost no matter the product. No product is going to model your data for you or build something like a semantic layer. Those are (mostly) completely separate tasks from the BI layer. Where I'm at we simply pointed Tableau at existing data sources built for performance reporting already.
This is another "Forbes Contributor" post. This means that it's just a blog post, though, admittedly, Cohan is a smart guy.
Does anyone else think Forbes should do more to separate volunteer content from editorially directed and reviewed journalistic content? Or is this just a moot point since folks don't trust "Forbes" as a brand?
It feels like, to me, someone saying "I went to Stanford" in a conversation because they walked on the campus. Just because Forbes hosts the content doesn't mean it's at the same quality as a true Forbes spawned story.
Since the Forbes.com front page appears to now be almost 3/4 contributors, I guess it's working for them. Still feels wrong.
I typically ignore things on Forbes for this reason. In this case, Mr. Cohan works in venture capital for SoFi (according to his Twitter bio) and it's not immediately clear if he has a stake in any of these companies or why he's taking a shot at Tableau other than for the clickbait. It would be nice to know whether his current employer has any material stake in either company before reading. I understand he is also a business school professor but I would expect a disclaimer for an article like this.
> Or is this just a moot point since folks don't trust "Forbes" as a brand?
That's the stage I've been in for a while now. Forbes.com now has the same value to me as Huffington Post or Inc.com. It's a content mill with no clear brand voice that anyone can write for.
There is a really big metric on their dashboard, "Pageviews", and this is a best of all worlds solution (cheaper, more content, diverse content, diverse authors promoting their content, forbes.com SEO juice, etc).
I really hate the 'This startup's revenue grew at x%!'
Like what does that actually mean? For all we know they went from $100k to $170k in revenue while raising $100 million dollars. It could either be very impressive or just lame click bait to inflate a company they invested way too much money in.
I'm not sure the analysis front-end is really where the money is at. There are lots of industries that have trouble getting their data to a place where it can be analyzed at all. I know some very successful companies here in flyover country that picked a sector where this is a problem and made a business of fixing it.
Not just the content. I went on the page and a huge 3/4 page banner scrolled my view, then another popped into the side, then Forbes had a pop over asking for my email.
Places like Forbes, Inc and Huffpost appear to exist solely for content writers to use their logo on their portfolio websites under the heading "Published In: ".
Well they're also "owned" by lobbyist-style groups. HuffPost is clearly (ludicrously) pro Democrat, Forbes anti-Apple, and so on. It's close to amusing.
Growing a startup '70%' is very different to running a publicly listed global company
'...Looker is trying to become cash flow positive -- and its revenues are growing faster than its headcount. "We have visibility to cash flow break even over the next couple of years. We plateaued in spending. We now have 600 employees and expect to add 200 in 2019. We are seeing our cash burn rate decline," he explained.'
Has anyone used Arcadia Data (the other company mentioned in the article) and have thoughts on it? I work in BI and had never heard of it before reading this.
Just exited a company the was competitive to Tableau and Looker. The BI space is insanely competitive, the lock in is very deep, and there are a thousand small players swimming in the wake of Tableau, Looker, PowerBI, and a few other "owners" of the category.
Looker doesn't threaten Tableau. It doesn't need to--at least for now. The space is growing at a crazy rate.
Unlike Looker or Tableau, it requires users to know SQL.
That said: if you're an org where the business people know SQL or are willing to learn--Mode is clean, simple, and super powerful.
My company had a great experience with this. Our COO was super proficient with Excel. I showed him basic SQL, he loves it and has learned quickly. The jump from Excel power user to SQL user is totally doable, and probably most people making custom reports are already at least in the former category!
Additionally, a few of us have substantial Pandas experience: Mode makes it trivial to go from SQL to a Jupyter like notebook w Pandas, Matplotlib, seaborn etc all already there--no need for Python version or dependency hell. Zero devops at any layer below the raw datastore (MySQL / Redshift / etc).
It's really nice. You write code purely against open standards--SQL, pandas etc. Looker uses a proprietary query language, LookML. I am not tempted to go learn it.
Not the same level, Tableau is super strong in visualization, Looker is about Data Modeling, the new language is not that hard, but the most valuable thing is data model and explore
Isn't widely known? It was the top rated BI solution by Gartner in 2017 and 2018. Full disclosure I used to work on it, and when I left we had over a million users so some people might have heard of it.
There is lots of opportunity still to innovate in this space. We are focusing on making it super easy to create great looking charts and dashboards really easily at Chartly [1].
Personally I am surprised by lots of the "state of the art" options out there where the result is EITHER ugly or HARD TO USE.
We still have a ways to go, but you can create good looking, interactive dashboards with our platform very easily.
It can mean anything the company wants. 70% more site visits per month, trial users, support requests, paid users etc etc. Unless explicitly stated, it's hard to determine what metric they are talking about.
First, Looker has two layers. The dashboard layer gives you tables, charts and filters. The explore layer allows you to go in and build your own pivot tables off the same underlying models. No other BI tool has built an explore layer that’s as successful in practice as Looker’s.
Second, the underlying model of both layers is written in code and under version control. This probably makes sense to most HN readers but in the BI world this is very contrarian. If you ask most business users who make BI purchasing decisions, version control and pull requests are the last thing they’ll ask for. But it turns out it’s exactly what you need to maintain a complex BI deployment.