As businesses have started to implement or think about implementing big data platforms and tools like Hadoop into their business strategy, many have run into obstacles figuring out how to measure the return on investment big data brings. Some businesses have tried treating it like a marketing tactic: Spend a certain amount on a TV campaign and see how many sales are made. However, sales revenue does not always accurately reflect other important factors, such as customer satisfaction or increased productivity. Likewise, big data is unpredictable in that managers may not always get the data back that they expect.
Rather than viewing big data as a tactical budget item, business owners should treat it as a line-item of their budget. A certain amount of money is put aside for data research, just like it is put aside for payroll, and just like employees are measured based on performance standards. The same should hold true for big data.
What this means for business managers is that before investing in a big data tool or personnel, they need to decide which questions or problems they would like big data to solve. This criteria will then determine which big data strategy businesses should incorporate and determine how the success of big data is measured.
The types of questions business leaders want answered generally fall into one of four categories.
Performance management uses transactional data, such as inventory levels and turnover, to answer pre-determined questions, such as: which customer segment is the most profitable? While businesses have been doing this kind of tracking for years, big data allows managers to get answers in real-time to make both short-term and long-term business decisions. The benefit of using this strategy is it can apply to multiple departments within a company including HR, marketing, sales and customer service.
How is this measured? If the tool is able to provide faster, more accurate data on how to improve performance and productivity -which in turn translates into cold hard cash - the system has met its goal.
Unlike performance management, data exploration seeks to discover problems or advantages managers haven't thought of yet. This strategy relies heavily on statistics to conduct experiments, in order to predict user behaviour based on previous business interactions and preferences. Based on the data derived, marketers may segment customers into new groups that they can then target differently with a customised message, service offer, or up-sell. This strategy is also useful for experimenting with marketing channels, such as determining between two website formats.
With this strategy, big data returns will depend partly on the types of questions being asked and how much the data is able to tell you. This means that your return will be measured by actionable insights that will improve your bottom line. If you aren't getting enough insights, you may be asking the wrong questions or the data may simply not be there.
Social analytics is another well known strategy that is generally a measurement tool for known objectives. Social measures non-transactional data that is generally divided into awareness, engagement and word-of-mouth reach. The problem with measuring these types of things has already been discussed heartily in the business world - awareness doesn't always translate to increased web traffic and sales.
Until big data platforms, it has been difficult for businesses to maintain an updated and integrated picture of what was happening on social media and which conversations ultimately mattered. With big data, social analytics can be taken a step further by correlating social engagement with business metrics to determine a more accurate business impact.
Here, the return can ultimately be measured by increased traffic and sales due to the traffic the data's insights can bring. This may be brought about by increased customer engagement, better customised offers to individual customers, customised landing pages on mobile apps and web pages, and so on. In either case, notice the switch from measuring vanity metrics, like re-tweets, to actually using social data to impact marketing and sales efforts.
Decision science is another way to conduct experiments through field research, but by collecting new data rather than relying on past data, and then using this to pose relevant questions to a community.
For example, marketers may want to find out how consumers feel about different price points before launching a new product. Marketers may go in after a product launch as well to see how public sentiment has changed and adjust the price accordingly.
Return on decision science should ask the question: Was information gleaned from the experiment that can positively impact sales or customer satisfaction? If the answer is yes, then big data did its job.
Michele Nemschoff is vice president of corporate marketing at big data platform solutions expert MapR Technologies.