8 Key factors to a successful analytics proof of concept

[fa icon="calendar"] Apr 3, 2018 4:36:53 PM / by Ricky Thomas

Ricky Thomas

8 Key Factors to a Successful Analytics Proof of Concept

Investing in a comprehensive proof of concept can be an invaluable tool to understand the impact of a business intelligence (BI) platform before investment. 64 percent of CIOs at the top-performing organizations are very involved in analytics projects, according to Gartner. In a competitive environment where analytics and BI is a business advantage, understanding the real-world impact of a solution is key.

The Value of Analytics Proof of Concepts

Creating a test run of prospective analytical solutions can demonstrate whether business intelligence platforms are worth the investment, and secure executive buy-in for big data initiatives.
The proof should be realistic enough to simulate performance and value in production, and instill confidence that a solution can scale to future needs. The following factors can shape a successful long-term partnership with prospective vendors.
 

1. Use Real Data

The use of actual business data is a necessity for valuable simulation. Tests run on a Business Intelligence Software provider's data sets prove very little.
The challenges associated with even a structured data set's unique characteristics and qualities are among the most time-consuming aspects of any big data project. Data scientists spend up to 80 percent of their time preparing data sets for analysis, according to CrowdFlower.
Test runs on simulated data also offer nearly no real-world value if a firm decides to green-flag the investment. While using real enterprise datasets can increase the time and technical challenges of a proof of concept, it's worth the investment.

2. Support Real-Time Insights

Support for real-time events and streaming data is among the key themes defining the fast-changing market for analytics and business intelligence platforms, according to Gartner. Dashboards of yesterday's data are no longer a competitive advantage.
Speed of data ingestion should be considered when evaluating any possible proof of concept. Firms should also evaluate whether the solution can support mobile devices, sensors, and other fast-moving streams of big data.

3. Flexibility for Future Disruption

CIOs know the analytics landscape is evolving rapidly, which presents a need for flexibility in any business intelligence software investment. However, understanding future requirements in an industry that's subject to significant disruption is near-impossible.
Ideally, a proof of concept should have the flexibility needed to support machine learning and analytics embedded into the core of the business. For more insight into future disruption, consider how solutions stack up against key 2018 market dynamics identified by Gartner:
  • Business intelligence software at scale to support enterprise governance
  • Automated, simple data preparation for fast adoption of unstructured data sets
  • Extensibility, and the ability to embed analytics into apps and existing processes
  • Support for real-time events and streaming data
  • Cloud deployment for end-to-end analytics and scalability
  • Additional data sources and increased speed of updates

4. Rely on a Modern Platform

In the past, there may have been clear lines between business intelligence platforms and proofs of concept for sophisticated analytics cases such as anomaly detection. Increasingly, these lines are blurred and the capabilities of traditional BI platforms used by analysts, and data science tools used by data scientists, are converging into citizen scientist tools.
The modern platform, by many accounts, is defined by ease-of-use for individuals without knowledge of Java or Hadoop. Gartner predicts by 2020, up to 80 percent of enterprise analytics tools will be self-service. A proof of concept which fits this ideal can allow your organization to achieve more self-serve data access and data-driven culture, despite the data science skills shortage.

5. Machine Learning Capabilities

Algorithms are exceptional at anomaly detection. They can identify issues and opportunities in your organization, such as far higher-than-average payment processing times in finance, much faster and more effectively than a human.
In modern business intelligence software solutions, embedded machine learning capabilities not only identify these patterns without prompting, they bring the insights to the user.
Embedded machine learning capabilities, according to Digitalist Mag, can support "the long-term dream of true exception-driven management." The right features can support your use cases for risk management, natural language processing and AI, and better integration of unstructured insights.

6. Collaborate

 
Previously, proof of concept evaluation may have been "an IT project." Today, collaboration between business and technical users is crucial to understand how a vendor and model resonates within an enterprise.
At a minimum, a collaborative team for evaluation should include an executive sponsor, a project owner, and individuals with expertise on both business data and processes.

7. Time-Saving Potential

Evaluating a proof of concept should involve both qualitative and quantitative measures of value, including its potential for time-saving. Some of these time-saving capabilities are related to on-staff data team members; analysts are likely to spend much less time on ad hoc reporting requests with self-service capabilities.
Other ways a platform could save time include superior features for data identification, preparation, faster scaling to new data sets or business models, and extensibility.

8. ROI

The ultimate success of an analytics investment is determined by return-on-investment. While many organizations realize cost savings in the capital expenditure (CapEx) of a cloud BI tool instead of on-premises data silos, the ROI of a proof of concept can extend to better business decision-making. Use cases could include:
  • Recommended corrective action based on supplier activity
  • Real-time marketing campaign recommendations based on client and prospect behavior
  • Predictive maintenance and anomaly detection capabilities for assets

Conclusion: Proof of Analytical Impact

A proof of concept for a possible big data solution should be more than just an in-depth demo of BI software. It should involve close collaboration, real data, and a platform that's in-line with industry trends. With the right approach and partnership, a successful proof of concept investment can be applied to deliver impact in the enterprise.

Topics: Analytics, BI, Business Intelligence, Machine Learning, AVORA, POC, BetterBI

Ricky Thomas

Written by Ricky Thomas