How to calculate the Total Cost of Ownership (TCO) of an analytics solution and what kind of solutions will be the next generation winners in the analytics TCO per user race? Let’s first look at the other side of the coin a bit and discuss analytics projects Return on Investment (ROI). Related to this topic is also useful to understand why analytics projects have had such a high failure rate in the past.
According to a study last year by analyst firm Nucleus Research, organizations that invest in analytics solutions, such as Business Intelligence (BI) software, are experiencing average returns of $13.01 for every dollar spent in 2014. Companies that invest more in analytics are performing better financially than their peers based on several studies. Achieving a great ROI in an analytics deployment is however all about the execution according to plan and in avoiding the “re-inventing the wheel” syndrome.
While some research results claim over 1000% ROI on analytics projects, some other research results indicate that 50-70% of business intelligence (BI) projects fail. Failure rate of this level is also the gut feeling of experienced data warehousing and business intelligence consultants that I have discussed with about this topic over the years.
Typically the reasons for failure include:
- Poorly defined goals for the project and limited understanding of business goals by vendors implementing the project. Trying to “eat the whole elephant” in one bite…
- Unnecessary project work done by building everything from scratch and maintaining own dedicated infrastructure with own resources
- Data quality issues
- Lack of committed resources from customer organization
- Poor end user interface design and unnecessarily complex solutions causing low usage rates
- Traditional analytics software licensing models often prevent grass root deployment because of high costs involved
So, on the way to realizing those over 1000% ROI analytics deployments it would be natural to start by eliminating these typical failures above. The list below provides potential ways to address these pain points:
- Work with vendors who understand your industry domain and communicate your plans and goals clearly to the vendor. Define a limited scope for phase 1 solution and take it into use quickly to gain business value immediately.
- Investigate alternatives for traditional business intelligence software and local deployments from the new crop of industry domain specific analytics solutions and cloud deployment models. Domain specific best practice analytics and/or integrations out of the box in a software as a service (SaaS) or platform as a service (PaaS) model can lower costs and speed up deployment.
- Provide visibility of the data to the business users and help them improve data quality in the source systems. This typically makes more sense than starting to build complicated data transformation business logic in the data warehouse or in other parts of the analytics process.
- Without internal buy-in it is very difficult to reach meaningful results. As the saying goes, the difference between involvement and commitment is like ham and eggs. The chicken is involved; the pig is committed.
- Analytics solutions end user experience needs to be designed differently for different user roles. For example a store manager busy with daily operational work does not need a statistical analysis tool with features for selecting between different kind of linear regression models. However, actionable insights delivered automatically in a visual format directly to email every morning might be just what is needed to make better decisions and increase the performance of the store this user is responsible for.
- Investigate if new subscription based pricing models enable deployments to large amount of end users, unlike the traditional costly business intelligence “cost per user” licensing models.
Total cost of your analytics solution heavily depends on what analytics tools and partners you select. Most of the cost on a 3 year total cost of ownership (TCO) period regarding an analytics investment in a traditional business intelligence tool deployment comes from cost of work (internal or external costs for the implementation and maintenance work). Software license costs vary a lot between vendors. Cloud BI vendors are currently the most affordable ones concerning per user license costs.
Traditional business intelligence project
To better understand the work related costs in a traditional business intelligence project, let’s look at an example case where a retailer is investing in building a business intelligence solution from scratch. This retailer is doing business via 3 channels: chain of 30 own brick and mortar stores, B2B wholesale business and a webstore for consumers. The goal is to set up a data warehouse and related reporting and analytics tools for 50 end users. The scope is to integrate sales and inventory data from 3 separate systems: Point-of-Sale (PoS) system used in own stores, ERP system used for running the wholesale business and an ecommerce system used to run the webstore.
This retailer might easily spend 750,000 USD within a 3 year TCO period for work costs alone. Off the top of your head you would say “No Way!” but ask any seasoned business intelligence consultant and they will tell you integration work is the killer in all traditional business intelligence projects. Let’s look at where all that money is typically spent in a traditional BI project based deployment:
The 300 man days spent on the initial project translates to on average 2,5 external consultants working full time for 6 months on the project. Nothing unusual there but let’s hope this project is not among the 50-70% of business intelligence projects that fail, as that would mean throwing some serious money out of the window..
In this example on years 2 and 3 on average 1 external consultant works 10 months per year with 50% workload. This work would typically consist of adding some new data sources and reports and making changes to existing solution based on end user requests.
Modern SaaS subscription
The next big disruption in analytics solutions total cost of ownership is expected to come from cloud based deployments of industry domain specific analytics solutions where TCO per user is often only 1/10th of the TCO compared with traditional business intelligence vendors. In monetary terms this translates to typical 10 000 USD – 50 000 USD per user TCO on a 3 year term for traditional business intelligence software deployment compared with typical 1 000 USD – 5 000 USD per user TCO on a 3 year term for a cloud based industry domain specific analytics solution deployment.
Let’s look at where the costs typically come in the modern SaaS subscription based business model:
The SaaS models vary a lot concerning the driver for the pricing – there are fixed monthly fees, per user fees or fees based on customer’s volume of business. Let’s use an example pricing based on the volume of customer’s business. The exemplary monthly subscription fee of 2 600 USD in this case could come e.g. from:
- Chain of 30 own brick and mortar stores with 34 cash registers: 34 x 50 USD/month/cash register = 1 700 USD/month
- B2B wholesale business with under 25 000 000 USD yearly sales : 600 USD/month
- Webstore for consumers with under 1 000 000 USD yearly sales: 300 USD/month
Below is a comparison example in a 50 end user deployment scenario using the example case used already before in this article:
In real life the difference between these different deployment models of course varies a lot and 1/10th of cost scenario will not always be the case, sometimes the models might have even same TCO, sometimes the SaaS model might be only 1/20th of the cost of the traditional model. One important factor is how much a SaaS solution can provide out-of-the-box working integrations vs. the traditional model of building integrations from scratch. The point is that you should understand the different cost components of TCO in an analytics deployment – license fees are just a small part of the total cost. In the end the choice between the traditional project based model and the industry specific cloud based SaaS model should be based on understanding of the following:
- How important is speed of deployment: SaaS based models typically enable fast time-to-value
- How important is total customizability: Many SaaS offerings enable customer specific customizations (for larger customers even via dedicated development projects) but they will never offer the full customizability that you get when starting totally from scratch
- How important is low Total Cost of Ownership (TCO): Fully customized solutions inevitably increase TCO
- How important is reliability and performance: Local deployments by default cannot compete in performance with cloud based scalable infrastructures nor have as stable environments
- How important is ease of use and industry best practice based analytics out of the box: end user experience design is a difficult sport, building things from scratch in this area will have a heavy price tag. Finding experienced people to build advanced analytics solutions (predictive analytics, data mining) is hard and expensive as well.
- How important it is to deploy the solution also to the grass-root level end users: traditional licensing models cost often prevents that
One can always argue that comparing traditional BI deployments against industry specific cloud based SaaS model deployments is comparing apples with oranges. I don’t really agree with that argument but I certainly agree that both models can provide great business value when deployed properly! Just be aware of the risks and rewards of both models when making the decision on analytics implementation.