A mid-sized retail chain invested heavily in big data infrastructure last year. A Hadoop cluster was deployed, the data warehouse was expanded, and two data scientists were brought on board. Twelve months later, the purchasing manager still struggles to forecast which product will run out at which branch, and the marketing team still targets campaigns by intuition. The technical team, meanwhile, reports successfully storing and processing terabytes of data. Both sides are right — and that is precisely why the project has produced no value.
Big data has become one of the most discussed topics on the corporate technology agenda in Turkey over the past two years. Research firms like Gartner and IDC pushed the concept into boardrooms, while technology vendors placed Hadoop, NoSQL, and real-time analytics at the center of their sales pitches. The result: many companies bought the technology before defining what it would serve. The majority of big data failures begin at exactly this point — the platform is selected before the business question is formulated.
The problem needs to be framed correctly. Big data is not a technology stack; it is a decision-making approach. Hadoop, Hive, HBase, Cassandra — none of these create value on their own. Value is measured by which business question these tools help answer, which process outcome they improve, and whose decision they accelerate. The first question any manager should ask when evaluating a big data investment is: ‘Which decision are we currently making incorrectly — or failing to make at all — because we lack the right data?’ Projects that cannot answer this question clearly will not produce a defensible ROI, regardless of how sophisticated the underlying technology is.
The second recurring problem in Turkish implementations is the absence of organizational ownership. Big data projects typically originate as IT department initiatives and remain under IT stewardship. The business units that should be consuming the data — sales, supply chain, finance — are not at the table during the design phase. In this structure, data scientists may build technically sound models, but the outputs do not connect to decisions that business users can act on. Until ownership is shared across the organization, no analytics platform converts to business value.
The third and perhaps least visible problem is the absence of data quality and governance infrastructure. Big data platforms can process high-volume, varied, and fast-moving data — but they cannot guarantee that the data is consistent, current, or trustworthy. In many Turkish companies, ERP systems have been in use for years, yet customer records, inventory codes, and cost center definitions within those systems remain inconsistent. Feeding this raw data into a big data platform does not solve the underlying problem; it reproduces it at a larger scale. Data governance — defining which data is produced by whom, in what format, and to what standard — is a foundational operational maturity step that must precede any big data investment.
The most concrete practical challenge is the talent gap. The data scientist role is still an immature career path in Turkey; profiles that combine statistical depth, programming competence, and business judgment are both scarce and expensive. Some companies attempt to close this gap by hiring professionals trained abroad; others turn to consulting firms. Both paths carry a sustainability problem: knowledge that is not internalized leaves with the consultant when the engagement ends. When total cost of ownership (TCO) is calculated, license and hardware costs are visible — but the cost of building organizational capability and developing internal talent is routinely left out of the budget.
For managers evaluating a big data investment, the decision framework reduces to three questions. What is the specific business question we do not currently know how to answer, and what would it be worth to answer it? Who produces the relevant data, who will consume the output, and how will it be embedded in an actual decision process? Is our current data quality mature enough to support this analysis? When all three questions have clear answers, technology selection becomes a secondary step. When the answers are vague, no platform choice will save the project — and the next budget review will file it under ‘big data disappointment.’
This article was originally written in Turkish by Gökhan MERCANOĞLU on February 11, 2013 and has been automatically translated into English and other languages using machine translation.