You can, as the saying goes, lead a horse to water, but you cannot make it drink. The same saying also seems to be true of analytics and particularly of platforms: You can provide a platform, but how can you make sure that it will be used, providing both value and a return on investment?
There is no question that acquiring the platform is only the beginning of the story. Developing its use until it becomes part of how we do business is a whole new challenge in its own right.
The purpose of platforms
To answer the question of developing use, I think you have to look at why customers turn to platforms. In my experience, there are two main types of approach. The first is the customers who have a definite use case in mind. They buy, build or develop a platform with a view to solving that problem. In many ways, that works really well: There is a problem, and here is a purpose-built solution. These customers tend to get high short-term value from the platform as they use it to address the particular problem. The platform will have high levels of use, but chiefly among the team that first identified the problem.
In the longer term, however, the value may not be so high. The initiating team may continue to use it, but others may not be able to jump on the bandwagon, especially if the purpose-built platform turns out not to be so good for other problems.
Other organisations are therefore taking a different approach. They see that a platform may be helpful to solve a number of problems, buy an off-the-shelf model, and then try to encourage users to come and play. They show people how to explore data in the hope of finding a few insights during the process. This is feasible with the advent of Hadoop and other storage options, and the rise of application programming interfaces, or APIs. It is also easier with visualisation tools, as this opens up data exploration to more users, and particularly business users rather than pure data scientists.
Platforms in practice
If these are the main reasons that customers turn to platforms, how can organisations ensure the platform gets used? In the first context, there are clearly issues about looking into the longer term, and being confident that the platform has wider applicability for the organisation. But perhaps there are also questions about scalability: making sure that the platform has options for multiple teams to use it, with small incremental costs charged to each new team. It also means continuing to encourage teams to use the platform to address problems, and showing them how where necessary.
The scalability and multiuser nature of platforms does, however, bring its own challenges, notably security. A multitenant environment, where data is handled, managed and manipulated, is something of a security concern. It is, however, worth getting over this. Some organisations, for example, have found it possible to restrict access to certain subgroup users.
Encouraging use is also a matter of having the right tools available. The platform must be open to a range of users with quite different skills – for example, in Python, Java, Lua, SAS® or R – and also different levels of skills, from experienced programmers to new business users. For widest use, the platform must be open and accessible, as well as easy to use.
It also needs to have the right data, which means that this must already be available to the organisation. It is important to ask whether there are enough useful data sets available, because without data, the platform will not get used. Finally, the platform must perform. The algorithm must respond within the required time, and it must also deliver useful and reliable results. It is no use encouraging people to use a platform that delivers unreliable answers.
Looking to the future
There has unmistakably been something of a step change in thinking about platforms. This may be a result of the amount of data now available. It may also be because of the rise in accessible tools that means business users can get involved in analysis. It may, however, not least be a result of the increased use of machine learning and artificial intelligence. All of which means that an analytics platform is now moving into the realm of “essential” rather than “nice to have.”