A retail chain’s IT director recently described a situation that many executives recognize: three years ago, significant budget was allocated to big data infrastructure. Server racks filled up, Hadoop clusters were configured, data engineers were hired. Today the company sits on terabytes of raw data but has nothing concrete to show the board. This pattern is common among Turkish enterprises that have embarked on big data initiatives. The problem rarely lies in the technology choice; it lies in failing to ask, from the outset, how the investment will translate into business value.
Big data, defined by the dimensions of volume, velocity, and variety, describes data sets that exceed the capacity of traditional data warehousing tools. But that technical definition does not provide a sufficient framework for investment decisions. Decision-makers need to answer a prior question: which specific business problem are we trying to solve? Reducing excess inventory? Predicting customer churn? Lowering operational costs? Without a clear answer, the infrastructure built becomes little more than an expensive storage system. Every project that has successfully generated returns shares one characteristic: the technical team and the business unit sat at the same table before the project started and agreed on a measurable objective.
An honest assessment of value creation requires looking at three categories: revenue growth, cost reduction, and risk mitigation. In retail, demand forecasting models produce visible reductions in inventory carrying costs — a directly measurable financial gain. In financial services, transaction anomaly detection reduces fraud losses — again, a concrete figure. In manufacturing, analysis of machine maintenance data shortens unplanned downtime, and the monetary value of that reduction can be calculated. Projects built around vague objectives such as ‘better decision-making’ or ‘building a data culture,’ however, make ROI calculation nearly impossible. Managers can position these as strategic investments, but they should know in advance that they will face pressure to produce hard numbers when budget reviews come around.
Total cost of ownership, or TCO, is consistently underestimated in big data projects. License and infrastructure costs are visible; what gets left out is the annual cost of data engineers and data scientists, the time required for data quality improvement work, and the training investment needed for business units to actually interpret analytical outputs. A company that builds its infrastructure on open-source tools and considers it ‘free’ is simply ignoring these hidden costs. When a realistic TCO calculation is performed, some projects turn out to be incapable of recovering their investment within the first two years. This is not a reason to abandon the project — it is critical information for managing expectations correctly.
A second common trait of value-generating projects is that they start with small, fast wins. Large-scale initiatives that attempt to integrate all enterprise data in a single effort typically lose momentum over 18 to 24-month implementation cycles, and often stall due to management changes or budget constraints. By contrast, projects focused on a specific business problem with a measurable outcome target within 90 days tend to build organizational support and lay the groundwork for the next investment cycle. This approach has proven effective in Turkish logistics and retail companies: analytical work that begins with a single warehouse or a single product category, once it demonstrates proven value, gets scaled across the organization.
The most realistic obstacle facing big data projects is not technology — it is data quality. Meaningful analysis is not possible when data arriving from multiple systems is inconsistent, incomplete, or incorrectly labeled. The mandatory adoption of e-Invoice and e-Ledger requirements has standardized financial data quality to a degree; but production, logistics, and customer data remains scattered across different formats and different systems. Data cleansing and harmonization work consumes a significant portion of project budgets and timelines. Companies that fail to plan for this find themselves stuck in data preparation before they ever reach the analysis phase.
Before committing to a big data investment, managers need to have clear answers to three questions. Is the business problem we want to solve well-defined, and is its financial impact measurable? Is our current data quality sufficient to support this analysis, or do we need to invest in data governance first? And over what time horizon, using which metrics, will we measure the return on this investment? If these three questions can be answered clearly, the project can move forward. If they cannot, every lira spent on infrastructure will continue to generate cost rather than value.
This article was originally written in Turkish by Gökhan MERCANOĞLU on February 2, 2015 and has been automatically translated into English and other languages using machine translation.