Proving the concept – adopting PoCs in the analytics journey

Intel Corporation explores how proof of concept (PoC) approaches can deliver on their potential, through selection and validation.

We are of course happy when our customers want to invest considerable money in their analytics infrastructure. This should come as no surprise. We are even happier, however, when a customer generates clear business value from its technology assets: in today’s complex world there is little wisdom in making a large investment before its value has been proven.

Because of such dynamism in the market, forward-thinking enterprises across a broad range of industries are moving away from built-to-last technology models and towards models that deliver insights and results fast. This makes sense: today’s report of sales figures may be superseded by new kinds of reporting next month, if not next week.

In the automotive industry, for example, new data sources such as engine and direct customer service data now needs to be managed and interpreted alongside dealership sales and service. As vehicles become increasingly digital and the industry looks to business models such as car sharing and even autonomous transport, it is difficult to say what reports, or even what data, will be most important in the future.

Analytics is also getting smarter over time. Passive reports are being superseded by real-time dashboards; these need to be viewed alongside advanced capabilities such as machine learning and AI, which can automate simpler decisions (such as low-risk authorizations) and free resources to focus on bigger tasks. In many cases, it is not possible to know in advance what these might be: given new data, one retailer may decide that its focus should be on reducing in-store theft, while another may look to increase its online presence.

Against such a background, exercises that test an idea – proof of concept studies, or PoCs – become a really good idea. They enable you to test that you are on the right track; they fit with the agile development models that modern businesses are looking to follow; they build skills and experience; they ensure you do not waste time on less valuable ideas; they help catalyze trust between business and technical teams.

To be successful, however, they cannot afford to be shots in the dark, but need to be based on hard criteria, in terms of both selection and validation. Let’s consider selection first: a PoC should look to focus on a specific business problem, driven by lines of business rather than technology groups, with appropriate levels of buy-in and sponsorship. You should be prepared to rule out PoCs that do not fit these criteria, as they can be undermined from the start.

In terms of validation, the goals of a PoC should be clear and achievable. You can consider questions like:

  • Will insight arrive in a timely manner, within a threshold of usefulness? All insight has a sell-by date.
  • Are results consistent over time? Data points need to be repeatable, in that the same set of inputs should yield the same output.
  • Are outputs accessible and usable, for example, do they generate alerts sent straight to a device, or do they have to be hunted out?

These success measures should be agreed between all parties as worthwhile and of direct value: perhaps the most significant is to ensure the information generated by an analytics solution is seen as having business value. Given the huge potential it brings, organizations should look to delivering results that reflect this potential. In doing so, they can achieve the double benefit of growing confidence and trust that subsequent analytics efforts will be worthwhile.

Armed with a successful PoC, you may look to scale it more broadly, investing in greater levels of infrastructure once the business benefits have been proven. Even as you do so, you can see the PoC as one of a series: this is the same mindset that agile innovation and DevOps organizations adopt. As the starting point for an innovation-led organization, perhaps one of the most important concepts that you can prove to the business is the value of the PoC model itself.

Learn more about how advanced analytics can help you transform your business, and what you can do to make it happen, by reading this new white paper from Intel.

Find more information on data-driven insights and advanced analytics by visiting our Turn Data Into Insight website, where you can find it all in one convenient location.