DBA 8060 Module 8 Final Project: A White Paper on Implementing Big Data at DAKOKOO Distribution Example

Seeing the Supply Chain: A White Paper on Big Data for DAKOKOO Distribution Corporation

Student Name

Doctor of Business Administration, Thomas Edison State University

DBA-8060: Empirical Entrepreneurship: Data, Technology, and Business Intelligence

Mentor Name

Month Day, Year

Seeing the Supply Chain: A White Paper on Big Data for DAKOKOO Distribution Corporation

Executive Summary

DAKOKOO is losing ground with its three retail customers because it cannot see its own supply chain. Global product stockouts cost an estimated $9.6 million in contract penalties and lost orders last year, expedited shipping charges reached $4.1 million and 59 percent of import orders have no recorded receipt date. This white paper recommends an eighteen-month program that builds a cloud data platform and four analytics capabilities in sequence: recovering lead-time data from emails, predicting stockouts, optimizing safety stock and sharing shelf visibility with customers. The program would cost about $3.2 million over two years and, on conservative estimates, return $5.0 million a year in reduced expedites, penalties and lost sales from the second year. The main risks are poor data quality, weak adoption by buyers and dependence on a single vendor, and each has a specific mitigation below.

The Problem

DAKOKOO's contracts with its three retail customers set fill rate and on-time targets that the firm missed in seven of the last twelve months. The problems reinforce one another. Without reliable lead-time data, buyers cannot set safety stock by supplier, so they apply one rule to every item. That rule holds too little stock for items from unpredictable suppliers, which run out, and too much for stable ones. When stockouts loom, buyers pay to expedite. Expedite charges have more than doubled in two years.

The cost extends beyond DAKOKOO. Corsten and Gruen (2003) found in a large multi-country study that out-of-stock rates in retail averaged about 8 percent and that shoppers facing an empty shelf often bought elsewhere, substituted or did not buy at all, so retailers lose both the sale and some loyalty. Our customers therefore judge us on availability, and their scorecards show it.

The firm's information problem also has a supply chain dimension. Lee et al. (1997) showed that distorted demand information amplifies order swings as it moves upstream, the bullwhip effect, and that sharing point-of-sale data and lead-time information reduces it. DAKOKOO receives retail sales data each night but does not use it in ordering, and it shares nothing back with suppliers or customers.

Why Big Data, and Why Now

Waller and Fawcett (2013) argued that data science, predictive analytics and big data would transform supply chain design and management, and they called for managers who combine supply chain knowledge with analytical skill. Wang et al. (2016) reviewed big data analytics in logistics and supply chain management and found applications in demand forecasting, inventory, sourcing and transport, while noting that many firms lack the data quality and skills to realize them. Fosso Wamba et al. (2015) found in a systematic review and a longitudinal case study that big data creates value through transparency, better decisions and new business models, but only when organizations change processes to act on it.

DAKOKOO meets the conditions these studies describe: large and varied data, decisions that repeat weekly and costs that respond to better information. What it lacks is the platform, the data quality and the analytical capacity.

Current State Assessment

The assessment shows that most of the data needed already exist; they are scattered, unjoined and partly unstructured. Hazen et al. (2014) argue that data quality is a basic condition for supply chain analytics and recommend measuring it on dimensions such as accuracy, timeliness, consistency and completeness. Completeness of receipt dates is our largest gap.

Proposed Architecture

The platform has four layers, built on a managed cloud service so the firm does not need to run its own servers.

Ingestion collects data from each source each night: order and receipt records, point-of-sale feeds, carrier portal extracts and the purchasing mailbox. Carrier arrival events are the one exception; they stream in as they occur, because rerouting a short container within hours can avoid an expedite.

Storage keeps raw data in low-cost cloud storage in open formats and loads cleaned data into a warehouse organized as a star schema around order lines, with supplier, item, store, date and carrier dimensions.

Processing uses Spark on the managed service for the text pipeline, model training and simulation, scaled up only when jobs run.

Delivery provides a morning scorecard for managers, a weekly watch list for buyers and, in the last phase, a shelf-availability report for each retail customer.

Analytics Use Cases

The program builds four capabilities, each depending on the last.

Lead-time capture comes first. A text classifier and entity extractor read supplier, forwarder and carrier emails, identify shipment, delay and arrival notices and attach dates to purchase orders. Pilot tests on a hand-labeled sample suggest the pipeline can fill most missing receipt dates. Everything else depends on this step.

Stockout prediction follows. A stockout model, starting with rules a buyer can read, warns three weeks ahead so there is time to reorder, move stock between centers or expedite.

Safety stock optimization sets buffer stock by item, using each supplier's lead-time spread and each item's demand variability, to minimize holding and expedite costs while meeting a 98 percent fill rate. Simulation suggests this could cut expedited shipments by about half.

Shared visibility is the last step and the most entrepreneurial. A monthly report to each retailer would show availability by store and item, the causes of shortages and the actions taken. Sharing information in both directions is the remedy Lee et al. (1997) proposed for the bullwhip effect, and it could become a service that differentiates DAKOKOO at contract renewal.

Implementation Roadmap

Each phase ends with a review by the operations and finance leaders before the next begins, so the program can stop or change course if results fall short. The reviews use the same scorecard the program builds, which keeps the judgment of progress tied to evidence rather than enthusiasm.

Governance, Data Quality and Privacy

A data governance council of the chief operating officer, the purchasing director, the finance director and the senior data scientist would own definitions, quality targets and access rules. Each data source would have a named owner responsible for its completeness and accuracy, with quality reported monthly on the dimensions Hazen et al. (2014) recommend.

Privacy matters even in a business-to-business firm. Emails contain names and phone numbers of supplier staff, and retail sales feeds are governed by contracts. Martin and Murphy (2017) show that privacy concerns shape trust and that firms gain by being transparent about data practices and limiting collection to what they need. The pipeline would therefore drop personal details at extraction, retail data would stay within contract terms and any social media signals would be used only in aggregate.

Costs and Benefits

Benefits are estimated conservatively from the simulation and from an assumption that only a third of current contract penalties are avoided by the second year. Even if benefits reach only half of these estimates, the program would recover its two-year cost early in the third year.

Risks and Mitigation

Data quality could limit every capability; the program addresses it first and measures it monthly. Buyers may not trust or use model outputs; the watch list starts with readable rules, buyers help design it and their feedback shapes revisions. Vendor dependence could raise costs later; raw data stays in open formats so it can move. Skills are scarce; the firm should hire two data engineers and two analysts and train its buyers in reading the new reports. Finally, a supply disruption outside past experience would break any model built on history, so the scorecard should flag when conditions move outside the range the models were trained on.

Future Trends and Managerial Considerations

The program should be built with the next five years in mind. Three trends matter most. First, machine learning services offered by cloud providers are making forecasting and classification cheaper, so the firm's advantage will come less from algorithms than from clean, joined data that competitors lack. Second, retailers are moving toward sharing point-of-sale and shelf data with suppliers and distributors in near real time, which rewards partners who can use it; a distributor that cannot will be passed over. Third, rules on data protection and on the use of automated decisions are tightening, so practices that are acceptable today, such as keeping supplier staff details in raw email archives, may not be tomorrow.

Managers will need to change as well. Sharda et al. (2018) stress that analytics succeeds when managers frame the right questions and act on results, not when a technical team works alone. Buyers will move from building spreadsheets to judging model output and handling exceptions, which changes their jobs and should be discussed openly with them. Executives will need to decide which decisions remain human, for example whether to expedite a shipment for a strategic customer, and which can follow the model. A short training program for buyers and managers, run in phase one, should cover how to read the watch list, when to override it and how overrides are recorded so the models can learn from them.

Conclusion

DAKOKOO's three problems share one cause: the firm cannot see the supply chain it manages for its customers. The records it needs mostly exist, scattered across systems, spreadsheets and inboxes. A phased program that first repairs lead-time data, then predicts and prevents stockouts, then shares visibility with customers would turn those records into fewer empty shelves, lower expedite charges and a stronger position at contract renewal. The research is consistent that such value comes only when organizations change how they decide, and the roadmap is built to make that change one tested step at a time.

References

Corsten, D., & Gruen, T. (2003). Desperately seeking shelf availability: An examination of the extent, the causes, and the efforts to address retail out-of-stocks. International Journal of Retail & Distribution Management, 31(12), 605-617. https://doi.org/10.1108/09590550310507731

Fosso Wamba, S., Akter, S., Edwards, A., Chopin, G., & Gnanzou, D. (2015). How "big data" can make big impact: Findings from a systematic review and a longitudinal case study. International Journal of Production Economics, 165, 234-246. https://doi.org/10.1016/j.ijpe.2014.12.031

Hazen, B. T., Boone, C. A., Ezell, J. D., & Jones-Farmer, L. A. (2014). Data quality for data science, predictive analytics, and big data in supply chain management: An introduction to the problem and suggestions for research and applications. International Journal of Production Economics, 154, 72-80. https://doi.org/10.1016/j.ijpe.2014.04.018

Lee, H. L., Padmanabhan, V., & Whang, S. (1997). Information distortion in a supply chain: The bullwhip effect. Management Science, 43(4), 546-558. https://doi.org/10.1287/mnsc.43.4.546

Martin, K. D., & Murphy, P. E. (2017). The role of data privacy in marketing. Journal of the Academy of Marketing Science, 45(2), 135-155. https://doi.org/10.1007/s11747-016-0495-4

Sharda, R., Delen, D., & Turban, E. (2018). Business intelligence, analytics, and data science: A managerial perspective (4th ed.). Pearson.

Waller, M. A., & Fawcett, S. E. (2013). Data science, predictive analytics, and big data: A revolution that will transform supply chain design and management. Journal of Business Logistics, 34(2), 77-84. https://doi.org/10.1111/jbl.12010

Wang, G., Gunasekaran, A., Ngai, E. W. T., & Papadopoulos, T. (2016). Big data analytics in logistics and supply chain management: Certain investigations for research and applications. International Journal of Production Economics, 176, 98-110. https://doi.org/10.1016/j.ijpe.2016.03.014