Debunking the Cloud BI Myth

on June 25th, 2014
Debunking the Cloud BI MythDebunking the Cloud BI Myth


Separating Facts from Fiction

Many expect that choosing a cloud platform is the best way to get a BI tool with the maximum level of self-service. Here’s where the myth that Cloud BI equals self-service perpetuates. Unfortunately, this is not necessarily true since the basic nature of business analytics includes the constant need to add, tweak and manage data. Meaning a Cloud BI solution that uses traditional BI technology and architecture will still require: countless service hours, an expensive ROI model, and a rigid data-driven environment with little tolerance for change. Quite the opposite of the personal control Cloud BI implies with its self-service association.

While this may sound counter-intuitive as it stands in direct contrast to the success of self-service cloud applications like SalesForce or Google Analytics, there is one rule of thumb that will steer you clear of choosing a Cloud BI solution that cannot provide the self-service benefits you’re looking for: If a BI solution is not self-service on-premises, it sure as heck won’t be self-service in the cloud. That’s right. The trick to finding a solid Cloud BI tool is to choose a software that is self-service BI, not managed-service BI, whether or not you plan on deploying it in the cloud or on-premises, and even if you are seeking a fully-managed BI service.

What “Cloud BI” Means & Its Self-Service Challenge

Choosing to install BI software on a “virtual” computer on Amazon EC2 is different than selecting a fully-managed BI service in the cloud, which is what most self-proclaimed cloud vendors offer today. Both are often referred to as “Cloud BI” or “BI in the Cloud”.

The former is a tactical choice which merely pertains to where the computer will be located: on the floor in your basement, or on the floor in Amazon’s basement. The latter, going for a fully-managed Cloud BI service, is a more strategic decision and requires the same considerations any organization needs to take into account before outsourcing their BI solution to a 3rd-party, on-premises.

With the success of many operational cloud applications, businesses expect to experience fully-managed Cloud BI the same way they experience other fully cloud-based self-service applications like SalesForce for CRM, Google Analytics for traffic analysis or Zendesk for help-desk management–which all provide true self-service.

Unfortunately, this is not the case with business intelligence since traditional BI solutions require end users to either call the customer support or bother their IT department every time they want to add data, alter a field or change a data visualization on the interface of a report or dashboard. The same process of transferring and handling data that’s required for on-premise will also be required for a cloud platform. There are several reasons for this, most of them technical in nature:

Adding and Synchronizing Data Is Like Treading Water

Technically speaking, BI software is best deployed to be as “close” as possible to the data it feeds off of. If this data is on the Amazon cloud, it makes sense to place your data on the Amazon cloud because there will be minimal overhead in transferring of data from the sources to the BI software for analysis. But, If the source data is on Amazon and the BI software is on Rackspace, that data would need to be transferred (over the www) from Amazon to Rackspace. Similarly, if the data is on-premises and BI is installed in the cloud, the source data would need to be uploaded (over the www) to the cloud first.

Keeping this data synchronized on an on-going basis combined with the frequent introduction of new data sets or sources will make things more complicated and feel like you’re constantly treading water. To top it all off, data that is transferred over the open Internet needs to be encrypted and then decrypted–slowing things down by an additional 60%.

Users will need to punch through these challenges just to get to the point where more BI-specific challenges emerge. These well-known and on-going challenges are around data warehousing, data modeling, query formulation and data visualization – and if traditional BI technology is used, they require a specialized pro to tackle them.

Popular Cloud Apps Are Innately Different from Cloud BI

The fundamental difference between “Cloud BI” and operational cloud applications like SalesForce or Google Analytics, which are in fact self-service cloud applications, is that if a BI tool is not self-service on premise it cannot be self-service in the cloud either.

One of the biggest reasons that SalesForce and Google Analytics are true self-service applications is because the same application is used for data entry, administration and operation. This means SalesForce/Google control (and can therefore pre-engineer) the entire data architecture, from how data is stored to what the user can do with it. In SalesForce, data is entered manually by sales teams, and in Google Analytics the data is collected automatically by Google via scripts embedded in your website. The data is then sent and stored on SalesForce and Google servers respectively, and structured to best serve each application’s pre-defined purpose.

Business Intelligence softwares, on the other hand, do not generate new data but rather plug into existing data landscapes–which changes everything from an engineering perspective, making it all the more complicated. Since the data is almost always located in a different location from the actual BI software, it can be generated by many different applications, in countless different formats, and stored in a variety of locations.

While it is impossible to pre-engineer a solution that fits every possible data landscape, you can use a solution that can be easily adapted to your existing data landscape and work to make it fit whatever changes are ahead. This is just as true for on-premises as it is in the cloud though, especially if you’re not using a single cloud provider or location. So again, self-service flexibility and control in this area is not a reason a company should choose a cloud platform.

Deciding If Cloud BI Is What You Want

Now that all misconceptions have been set aside, there’s an easy way to decide if Cloud BI is what you are really looking for. First, it is important to establish if your company actually wants a self-service BI tool, because, as you see, the data management process is practically the same whether your BI tool is on premise or on the cloud and will only vary depending on if the BI solution itself is build for self-service. Next, if you do decide your company wants a self-service tool, then the way you can take advantage of the cloud is by installing a truly self-service BI tool on whatever cloud infrastructure you store your data on (if possible). Then, because the BI tool is built for self-service, it will be the closest you can get to what we call Self-Service cloud.


Elad Israeli

Elad is a serial entrepreneur with over 15 years of experience in Product Innovation. At Sisense, he splits his time between Technology Innovation and working directly with customers to help them succeed with Business Intelligence.

See more posts by Elad Israeli
  • Theodore Omtzigt


    Most BI environments can be pre-engineered for the common case. The problem is whether or not that common case covers enough market to justify the cost. The reason Salesforce and Google Analytics work is because they have a large enough market to apply the solution to, justifying the investment in time and resources. Most enterprises, and particularly government, insists that their process is the key differentiator, and that kills most BI applications as the cost is not commensurate with the value provided. Point in case: look at what is going on with FedRAMP. FedRAMP adds a layer of cost to ANY data initiative that prices the solution out of reach of all but the largest states. A typical US State BI solution (<2TB data, ~100 concurrent users) is a $200-300k/year problem. But if you want it on FedRAMP it becomes a $1M+ problem. The pre-engineering of the infrastructure (=FedRAMP) is creating budget problems for all but the largest of the US States.

  • Elad

    Thanks Theodore. :)

    You are absolutely right that it is very hard to justify 200/300K a year BI solution, unless you can directly tie it to revenue which for most organizations is impossible to accomplish, especially for government institutions as you describe.

    For millions of organizations, BI has to be a <$50K a year problem, otherwise cost alone is a barrier to even start. This context of this blog is for these organizations. When your budget is this modest, technology and infrastructure is the first and primary challenge from a cost perspective, so misunderstanding what the cloud can or cannot do for you can lead to a project which will inevitably fail exactly as you describe.


  • Theodore Omtzigt


    Thank you for that elaboration of the context. The reality of most enterprise IT programs is that most of the cost is in people. I think this is particularly acute for BI, where most, if not all, of the value-add comes from a domain expert interpreting information and trying to draw conclusions that aid some business process. And with that I would say that budgets for analytics needs to be a lot bigger than $50k.

    If I am a business leader, and I delegate some decision process, I want my best people to work on that to reduce risk. The BI tool cost will quickly dwarf the salary costs, so the ROI argument has to be:
    1- is the cost of failure justifying the expense, and
    2- do I get a quality answer that I can verify.
    When you look at BI tools in this way, you’ll quickly see why spreadsheets are still the most ubiquitous BI solution. When a business process affects any more than a couple of $M of revenue, the BI tooling quickly becomes OPEX noise.

    With warm regards,


  • Jamal Syed


    Great blog! I totally agree with everything you said. Key to an enterprise BI solution is the ability of the system to provide a single view of entire organization by bringing disparate data silos on single platform. As you mentioned in your blog, Cloud BI fails to do that. From my experience working with several large organizations, real-time
    or near real-time reporting – albeit only 2% to 5% of overall enterprise reporting – is very near and dear to operational and financial users. Cloud BI falls short in that as well. I am sure there will be more agile and secured ways to move data between the on premise and cloud in coming days, but right now it is an uphill battle to convince organization to move their proprietary data in cloud.