IoT Analytics: Buy vs. Build – The Key Questions to Ask Before You Decide

Advanced analytics is making it possible to create new IoT applications that drive better business outcomes. Use cases and applications for IoT include predictive maintenance, asset optimization, fraud detection, predictive 1-1 marketing, supply chain optimization, and more.

But determining the best approach to implementing IoT analytics is not simple. There are no clear guidelines available for IoT project decision-makers.

There are a wide array of technologies to create IoT analytics applications that can deliver sustainable business value:

  • Traditional Data Warehouses
  • Hadoop databases of structured and unstructured data
  • Predictive analytics tools
  • Historical analytical tools and data sources
  • Open source statistical tools like R
  • Visualization tools and technologies of various types

While narrow solutions that only use a subset of the above tools and assets may have merit at small scale, significant innovation and return on investment require integration of all of the above in some form.

Your organization may use one or more of the technologies on the above list, and are now faced with the decision of what is the most economical and effective approach to developing analytics applications for your organization.

Using existing internal development resources is one possible approach, but that approach comes with risks and limitations. Acquiring a new platform or technology that is designed for IoT analytics is an interesting alternative, but the technology learning curve and the need to integrate with in-house systems also raises significant challenges and questions.

The dilemma organizations face when considering building their own solutions

While it is maturing quickly, IoT Analytics application development is still early in its adoption life cycle.

This is similar to earlier phases of computing, and with much of this seedbed for IoT coming into production, organizations are now asking the next set of questions. They want to know how to gain value from the high volume of data that is inherent in IoT.

In this phase, operations managers need to be open to new kinds of analytics application development opportunities. But they have to ensure that existing investments in technology and human resources are leveraged in the process.

In addition, they must manage the significant risks inherent in any new development project. These projects are usually complex and can consume significant time to get to production. The dilemma boils down to:

How do I create innovative new IoT analytics applications leveraging existing investments, but keep risks and costs manageable, and under tight control?

Critical questions to ask when evaluating the choices

As has been seen in past technology cycles and is also true with IoT analytics, the choices to resolve this dilemma come down to internal development (build scenario) vs. acquiring a platform/application development environment (buy scenario).

Sorting through this choice requires asking and answering some critical questions.

1) Is building out all the pieces needed your team’s core competency?

Most organizations that are evaluating projects for IoT analytics include talented individuals in areas ranging from database integration to predictive analytics, and other disciplines. They may be capable of creating the entire stack of a working application.

However, the critical question to ask is not IF your team can work on these projects, but instead ask what is the best use of their skills within the overall stack of systems and application technology needed to create a production application.

Specific tools or other system level functionality of the solution may be available commercially or in other forms (open source, etc.) and are often better choices. It is useful to carefully examine your team’s skills and match them against the overall capabilities needed in any application.

Internal teams are often best used at the application layer where there is unique knowledge about business logic, workflow, and other application-specific requirements.

2) Is this the only IoT Analytics project your team expects to work on?

If you expect to do more than one project over time, it is then best to ask yourself if the core technology you are using or building can be used across multiple projects? If not, you may have a problem with the economics of your development methodology and platform.

3) Are you considering all the key components and tools needed for successful IoT Analytics? 

Predictive modeling tools are only one of the pieces needed for these types of projects. To meet the challenges of IoT analytics applications, your team will need tools or technology for data ingestion, integration with data warehouses, building descriptive analytics, predictive analytics, prescriptive analytics, and software for triggering actions within your application.

4) Do your business analysts and non-technical staff  have the tools they need to create and operationalize analytical models quickly – in minutes – in order to deliver value before it diminishes?

You want the application and domain knowledge of your team to become mainstream, you do not want their efforts and time spent on learning low level technical tools.

5) Are your data warehouses, predictive models and tools integrated with a larger platform that includes data ingestion, streaming analytics, descriptive analytics, predictive analytics, and an integrated action framework?

There are a variety of excellent predictive and other analytics tools on the market that work effectively for some use cases. But even the best tool may not lead to overall success unless it is integrated with a larger data management and analytics environment.

6) Does your methodology enable relatively quick movement to the most valuable part of the analytics value chain? 

IoT requires timely action to capitalize on the window of opportunities. But a Hadoop database and predictive tools in isolation are not sufficient.

A critical requirement for success in IoT is the ability to execute analytics in real-time across the analytics value chain (as shown below in Figure 1) with relevant contextual and situational data that addresses the critical “last mile” for timely outcomes.

IoT Analytics Value Chain

Figure 1: IoT Analytics Value Chain

 

This ability, when combined with the subsequent ability to take the next best action in any particular scenario, is what creates the greatest value.

The increasing value chain depicted below shows how each step in the process refines the data and adds more value and context. This progression of value is where you can get real ROI in IoT analytics.

7) Are you concerned about how your existing tools, technologies, and staff  will be leveraged? Concerned about having to start from scratch with new technology to learn? 

There are a variety of analytics tools on the market that work effectively for a variety of use cases. Your teams are likely using some of these tools and have developed expertise and skills in one or more of them.

The best outcome is to find new solutions and platforms that leverage these investments and skills but also enable accelerated development and more valuable applications.

8) IoT Analytics is a young technology and solution domain and so it is inherently unpredictable at this stage. Results will be tough to predict at the beginning of any particular project. How do you plan to mitigate these risks? 

There is no magic formula to addressing this important question. Early stage markets require agile, iterative development with a rapid prototyping approach that enables fast learning and innovation. Small iterations in rapid succession are the way to achieve results in these types of markets.

What is the goal?

As you consider the buy/build questions above, it is useful to ponder what is the overall goal for your IoT analytics efforts.

As is often the case with platform decisions of this nature, the general rule of thumb is that the more applications you choose to build will yield greater return on investment. This is true in both the buy and build cases, but as is shown in Figure 2 below, the learning benefit that accrues to the buy case is larger than the build case.

Time-to-Market for Applications

Figure 2. Time-to-Market for Applications

The reason for this is straightforward. It is difficult for internal teams to truly build out an application-agnostic platform, even if they intend to do so from the outset. Any one individual company is rarely able to gain a broad enough perspective to maintain the application neutrality that is central to a platform methodology.

Platform that are designed from scratch to be application neutral are based on a rapid learning principle so that once you learn how a tool or technology works for one application, the incremental work and learning is far less for each succeeding application.

Internal teams that build platform level capabilities will obtain some of these benefits, but unless they are focused on building a platform first and their application second, it is extremely rare for those benefits to accrue at the same pace.

Finally, there is also far less integration cost when buying and using platforms that are designed from scratch to integrate multiple tools and system components.

Final Thoughts

Building IoT Analytics applications that can deliver significant business value is not a simple proposition. Among the many decisions is to make a choice on internal development vs. buying a commercial platform.

The first step is to weigh the benefits and challenges of a going with a platform-based approach vs. building all the components internally. And understanding your long run goals is critical to the buy vs. build decision.

If your plans include more than a single application, then the weight of evidence and experience argues for a commercial platform based strategy for IoT analytics application development.

To learn more about the commercial platform based strategy, get the Vitria IoT Analytics white paper to discover Vitria’s unique approach to IoT analytics.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>