Why has Standalone Cloud BI been such a Tough Slog?

I remember when I left Business Objects back in 2004 that it was early days in the cloud.  We were using Salesforce internally (and one of their larger customers at the time) so I was familiar with and a proponent of cloud-based applications, but never felt great about BI in the cloud.  Despite that, Business Objects and others were aggressively ramping on-demand offerings all of which amounted to pretty much nothing a few years later.

Startups were launched, too.  Specifically, I remember:

  • Birst, née Success Metrics, and founded in 2004 by Siebel BI veterans Brad Peters and Paul Staelin, which was originally supposed to be vertical industry analytic applications.
  • LucidEra, founded in 2005 by Salesforce and Siebel veteran Ken Rudin (et alia) whose original mission was to be to BI what Salesforce was to CRM.
  • PivotLink, which did their series A in 2007 (but was founded in 1998), positioned as on-demand BI and later moved into more vertically focused apps in retail.
  • GoodData, founded in 2007 by serial entrepreneur Roman Stanek, which early on focused on SaaS embedded BI and later moved to more of a high-end enterprise positioning.

These were great people — Brad, Ken, Roman, and others were brilliant, well educated veterans who knew the software business and their market space.

These were great investors — names like Andreessen Horowitz, Benchmark, Emergence, Matrix, Sequoia, StarVest, and Tenaya invested over $300M in those four companies alone.

This was theoretically a great, straightforward cloud-transformation play of a $10B+ market, a la Siebel to Salesforce.

But of the four companies named above only GoodData is doing well and still in the fight (with a high-end enterprise platform strategy that bears little resemblance to a straight cloud transformation play) and the three others all came to uneventful exits:

So, what the hell happened?

Meantime, recall that Tableau, founded in 2003, and armed in its early years with a measly $15M in venture capital, and with an exclusively on-premises business model, literally blew by all the cloud BI vendors, going public in May 2013 and despite the stock being cut by more than half since its July 2015 peak is still worth $4.2B today.

I can’t claim to have the definitive answer to the question I’ve posed in the title.  In the early days I thought it was related to technical issues like trust/security, trust/scale, and the complexities of cloud-based data integration.  But those aren’t issues today.  For a while back in the day I thought maybe the cloud was great for applications, but perhaps not for platforms or infrastructure.  While SaaS was the first cloud category to take off, we’ve obviously seen enormous success with both platforms (PaaS) and infrastructure (IaaS) in the cloud, so that can’t be it.

While some analysts lump EPM under BI, cloud-based EPM has not had similar troubles.  At Host, and our top competitors, we have never struggled with focus or positioning and we are all basically running slightly different variations on the standard cloud transformation play.  I’ve always believed that lumping EPM under BI is a mistake because while they use similar technologies, they are sold to different buyers (IT vs. finance) and the value proposition is totally different (tool vs. application).  While there’s plenty of technology in EPM, it is an applications play — you can’t sell it or implement it without domain knowledge in finance, sales, marketing or whatever domain for which you’re building the planning system.  So I’m not troubled to explain why cloud EPM hasn’t been a slog while cloud BI absolutely has been.

My latest belief is that the business model wasn’t the problem in BI.  The technology was.  Cloud transformation plays are all about business model transformation.  On-premises applications business models were badly broken:  the software cost $10s of millions to buy and $10s of millions more to implement (for large customers).  SMBs were often locked out of the market because they couldn’t afford the ante.  ERP and CRM were exposed because of this and the market wanted and needed a business model transformation.

With BI, I believe, the business model just wasn’t the problem.  By comparison to ERP and CRM, it was fraction of the cost to buy and implement.  A modest BusinessObjects license might have cost $150K and less than that to implement.  That problem was not that BI business model was broken, it was that the technology never delivered on the democratization promise that it made.  Despite shouting “BI for the masses” in 1995, BI never really made it beyond the analyst’s desk.

Just as RDBMS themselves failed to deliver information democracy with SQL (which, believe it or not, was part of the original pitch — end users could write SQL to answer their own queries!), BI tools — while they helped enable analysts — largely failed to help Joe User.  They weren’t easy enough to use.  They lacked information discovery.  They lacked, importantly, easy-yet-powerful visualization.

That’s why Tableau, and to a lesser extent Qlik, prospered while the cloud BI vendors struggled.  (It’s also why I find it profoundly ironic that Tableau is now in a massive rush to “go cloud” today.)  It’s also one reason why the world now needs companies like Alation — the information democracy brought by Tableau has turned into information anarchy and companies like Alation help rein that back in (see disclaimers).

So, I think that cloud BI proved to be such a slog because the cloud BI vendors solved the wrong problem. They fixed a business model that wasn’t fundamentally broken, all while missing the ease of use, data discovery, and visualization power that both required the horsepower of on-premises software and solved the real problems the users faced.

I suspect it’s simply another great, if simple, lesson is solving your customer’s problem.

Feel free to weigh in on this one as I know we have a lot of BI experts in the readership.

15 responses to “Why has Standalone Cloud BI been such a Tough Slog?

  1. Dave, interesting post. A couple of my beliefs:

    1. Data has inertia. It hates to move, especially at volume.
    2. Cloud BI isn’t just a BI Platform, it is an entire analytics stack of ETL, EDW and BIP.
    3. Ergo, firms that have tons of app data in the cloud, and have a full stack analytics solution, are likely going to be the gorillas of the cloud market [e.g., AMZN, MSFT, CRM and maybe GOOG].

    AMZN of course has nearly infinite data from ISVs / IT apps, so consuming that data in Quicksight after storage in Redshift augmented with their ML fed via pipelines in Glue makes for a pretty complete solution. Their leader, Anurag, is impressive. The same can be said for Azure, with PowerBI, SQL Azure and ADF, with their tons of apps built on the same platform. 3000+ ISV apps on Salesforce makes them another huge steward of data and a rich analytics stack. For GOOG, I say maybe because we need to see if their enterprise efforts can thrive, and if they can attract enterprise app developers en masse.

  2. My guess is that:
    1) historically a lot of on-premise BI was on operational systems. At Business Objects, we worked with in-house operation systems, then PeopleSoft and Siebel. When those systems went to the cloud, it wasn’t easy to connect a BI system on top, because the cloud systems didn’t have good APIs or a SQL interface. Also the operational reporting got good enough quick enough that people weren’t going to buy a BI system to report on their Zendesk tickets or Workday pay.

    2) once the lower-hanging fruit of operation reporting is gone, the rest is selling cross-business unit BI to the CEO and CFO. Here, there aren’t a lot of users, the setup cost is high, and there are a bunch of integration issues. For the on-premise world, this meant Informatica, a new data warehouse and the costs of that were way above $150k, with a team full of DBAs, ETL specialists, and business people arguing ontology – what “revenue” meant as they integrated SAP, Siebel and Remedy.

    Talking to the vendors you listed above, they never sold on top of a single system. You had to have multiple systems feeding them, and they brought the data together, and then could present Marketo+Salesforce or Eloqua + Pipedrive + Zendesk. The only exception is GoodData who were embedded as the operational analytics for a set of cloud vendors.
    The people using BI on top of multiple systems aren’t usually operational, because they already have some integration in the operational systems if viewing the data from two places was that important – either the vendor doing point to point, or the cloud ETL/ELT/ESB vendors who push data between operational systems. So they are Ops, Analysts or the CxO offices.

    So I think there wasn’t really a good market for BI in the cloud, and it was expensive to implement each one, and to make true on the “ease of use” promise, the cloud vendor had to own the complexity, and couldn’t get paid much to do it.
    Then:
    1) For most people, point-to-point integrations allowed users to succeed using the reporting in the operational systems, and for ops and analysts to use Tableau, Qlik and Excel to stitch them together when they had to.
    2) For the larger companies that had complex needs, they also came with their legacy warehouses, and couldn’t get the same behavior in the cloud, and never adopted cloud.

  3. Great post, Dave! I believe that the main issue with the mainstream, self-service/discovery BI – is the need for flexibility. While the cloud offers scale, availability, and performance, desktop products are inherently more flexible, and this is why Excel is the best BI product out there.

    I also agree with Keith that Cloud BI is not a product but a value chain. This is why Tableau and SFDC are losing to Microsoft and AWS who control their respective data value chains better.

    So where does this leave GoodData? We believe that self-service/discovery BI is only one use-case and there is actually more cloud-friendly and potentially bigger problem at every enterprise: the production and delivery of insights to people who can’t or don’t want to play with their data. We believe that the notion that hundreds (and maybe thousands) of employees have the time and inclination to slice and dice data on a regular basis is fundamentally flawed. And we are now using AI and ML not only to produce these insights, but we also augment and automate millions of mundane decisions (such as approval of facility requests at large retailers). That’s where cloud scale, availability, and performance truly pay off.

    My initial vision for GoodData was to use the cloud to turn everyone into data analyst. Now I am using the cloud to make sure that no one needs to be a data analyst. How ironic… :)

    • Roman — thanks for commenting (and reading). I loved your quote at the end about the irony and think you’ve been very smart about your strategy and the adaptations to it over time at GoodData.

  4. The cycle of innovation in BI tools seems to always start at the analyst desktop like Cognos and BOBJ before Tableau and Qlik after them. With each cycle, the vendors attempt to move from the analyst to the entire organization, where larger licensing revenue will support their growth. Unfortunately, the access analysts have to data are not as readily available to the other users in the organization. IT holds the keys to data access which starts the magic project to solve everybody’s reporting needs which also is the project that never ends. The cycle repeats itself as the anointed BI tool of choice for the company becomes the hated tool and the new desktop darling starts to look more appealing.

  5. A lot of “cloud” is focused on getting rid of traditional IT, but BI as a platform is one of the few horizontal use cases that is IT-based. I think BI either needed to shift to LOB-specific functions, an executive sell, or be an IT-owned product for the new IT, but these are three separate things that BI needs to choose from.

    IT-based solutions are better when they’re owned, customizable, and increasingly composable at an API level. The big cloud BI companies have struggled to know who they are, other than GoodData and Looker.

    I think Birst-Infor puts ﹰBirst in the right direction by making it part of a larger platform and putting Birst squarely into embedded BI owned by analytic app developers while getting a huge LOB resource increase. Birst has pivoted its go to market a couple of times and I hope this acquisition leads to the right direction. The initial press is all saying Birst will remain a standalone subsidiary, but Infor as Birst’s top ISV should help as an “anchor client” of sorts.

    • Thanks for commenting Hyoun. Most of the time, imho, these things end up getting integrated within a year — even if at deal time they say they are going to run it standalone. We will see.

  6. Hey Dave,

    First time commenter, long time reader here…

    Great post. Totally agree with Keith’s point above. I worked at GoodData and the sheer extent of product they had to build and support was (and I am sure still is staggering). Especially since each pieces of the stack are going through tremendous innovation…making it very difficult for a single vendor to keep up. I also think the struggle with Cloud BI vendors is that they took a very traditional, tops-down model approach to BI that relied on a lot of intensive modeling and schema definition. This also impacts the GTM/selling process. Compared to Tableau where any user can just pick up data and start visualizing it without having to conform to a larger data model. Users also bought individual seats or departmental licenses vs. a whole BI implementation. I think this is a faster and more scalable GTM and more closely aligned with how people buy software nowadays.

    My 2 cents… Great post as usual.

  7. Hi Dave — BI may be one of those rare (really, really rare) markets where returns from first generation (call it on-prem) top returns from the second generation (aka cloud). Thank you for helping to shine some light — from experience, rather than just postulation — on why that might be the case.
    Brenon

  8. Seann Gardiner

    Great post Dave. It certainly captures a few of the reasons Alteryx has grown so quickly – most importantly addressing the needs that the users face – quickly while giving enterprises the flexibility as to the deployment model (cloud/on-prem/hybrid). There is no doubt Cloud BI is hard, but putting Alteryx in the middle to prep/blend/analyze data before you pipeline it to the cloud or other consumption layer will make that transition much easier.

  9. Pingback: Tableau subscription pricing - a proxy for software acquisition

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s