The term Data culture is used loosely and often hyped up. What is important is a fundamental recognition across the organization that data is important... not just that it is important to make decisions based on data, but also that it important to ensure the entire data lifecycle is engineered right…recognition that data governance is the key to ensuring adequate data quality.

Example: Consider the chart of accounts (COA). This is a structured list of financial accounts used by an organization to categorize and record its financial transactions. It serves as the foundation for an organization's accounting system and helps in organizing financial data for reporting and analysis.

As you would know a typical transnational would have a subsidiary company in each country, while different LOBs (Lines of Business) operate across the countries. (Take for example: Unilever India & Unilever Singapore are subsidiaries of Unilever PLC, UK… while Food Products and Cosmetics are separate LOBs that operate across different countries… or each brand could be an LOB).

In the name of laisse-faire data governance, many transnationals made the mistake of letting each country define its own chart-of-accounts, either because it is required for local regulatory compliance, or sometimes just to satisfy the whims of the local CxOs. As each country fully flouted its freedom to define its own COA, soon, there is absolute chaos and zero similarity between COAs of different countries.

Unfortunately, a Transnational still needs to consolidate accounts at the Global Headquarters both for statutory reporting and for global management control. They also may need to look at consolidated global P&L for each LOB, and each brand.

The result?

Accounts reconciliations and the global consolidation of the accounts each quarter-end become an absolute nightmare, involving long hours of work, a battery of CAs and brute force. Releasing audited financial results take anywhere between 6-8 weeks after the quarter-end. Given the data quality is suspect, and in the absence of drill-down visibility external auditors have no choice but to close their eyes, or include enough caveats and audit-notes in audit statements.

Assuming it is implemented well, and not defunct, a MDM implementation helps in ensuring consistency in master data across different application instances within the organization.

So, it is important to fix MDM first. MDM helps you ensure consistency in data definitions and data standards; it helps you define all master data in one place, before replicating it to different applications. Each customer, or each material should have only one code, across different applications. E.g.: India Hotels is a big customer of Maruti. Then different units of India Hotels if they are buying cars from Maruti should not have different customer codes.

We should also ask what kind of data is being captured and what is not. Also, how is it captured?

Too many Excel sheets floating around in an organization is a cause for alarm. Reconciliation of data between different Excel sheets, data validation, ensuring authentication and data consistency, etc. is a very difficult exercise. Too many Excel sheets floating around indicate there are serious information supply chain bottlenecks. Excel data ideally must be moved to a web application as quickly as possible. Data lying in spreadsheets is an auditor’s nightmare. With an ERP, queries/drill-downs and validation can be done instantaneously.

If for some reason, Excel sheets cannot be avoided, they should be standardised. The formats must be locked and data validation rules should be built for each of the columns. Each Excel sheet should have a unique owner with authorization to change the format if need be. But ideally, one should move away from Excel sheets to MDM and web applications.

Note: MDM is used to define and manage the critical data of an organization. The data may include customers, products, employees, suppliers, and other types of business-critical information.

MDM aims to create a single, authoritative, and consistent source of master data that can be shared across an organization. This ensures that across the organization, people are using the same accurate and up-to-date data.

MDM establishes data governance policies, data standards, and data quality rules. This helps prevent data inconsistencies, errors, and redundancies.

MDM often involves integrating data from various source systems and applications across the organization. This integration helps eliminate data silos and provides a unified view of data.

MDM involves establishing processes, roles, and responsibilities for managing and maintaining master data. Data stewards are typically assigned to ensure data quality and compliance with policies.

MDM ensures that data is typically organized into hierarchies or taxonomies. For example, in customer master data, we might have parent companies, subsidiaries, and individual customers, all organized hierarchically.

MDM encompasses the entire data lifecycle, from data creation and entry to archival and deletion. It ensures that data is accurate and relevant throughout its lifecycle.

Implementing MDM can lead to several benefits, including improved data quality, enhanced decision-making, reduced operational costs, increased agility, and better customer and partner relationships.

MDM is particularly important in large organizations with diverse data sources and a need for accurate and consistent data across various departments and systems. It helps organizations make more informed decisions, reduce errors, improve operational efficiency, and respond more effectively to changing business needs.

One should first make a broad choice: to be a techie or a consultant.

MBAs are better off being consultants. They can focus less on coding and more on the holistic understanding of the problem being solved and the business value generation.

Techies can be more specialized. They should be very good in one or two technologies. Then they can easily pick up other technologies. In the case of Gen AI / LLMs for example, they can try learning Falcon or Llama, besides Open AI.

When we are trying to skill ourselves, we must choose areas that cannot be easily taken over by robots.

Use of IT and data analytics in the pharma industry is not new. The industry has documentation running into 100,000 pages or more for the development of new drugs. Trials should be monitored carefully and all the data allected analysed for insights. One important use case is Adverse Events Reporting. AER requires careful analysis. Huge amounts of data are involved. Earlier the number crunching would take time. Today, thanks to tools like Hadoop, the time to process data has been reduced.

Note: Adverse Event Reports (AERs), also known as Adverse Drug Reaction (ADR) reports contain information about unexpected or undesirable effects or incidents related to the use of a drug, medical device, vaccine, or any other healthcare intervention. These reports play a crucial role in monitoring and ensuring the safety of medical products and interventions across geographies.

The primary purpose of AERs is to collect, document, and report any adverse events or adverse reactions that occur in patients or users of a particular medical product. These reports help regulatory agencies, healthcare providers, and manufacturers monitor the safety profile of the product.

An AER typically includes details about the adverse event or reaction, such as the patient's or user's information, the date of the event, the healthcare product involved, the severity of the event, any relevant medical history, and a description of the adverse event itself.

AERs can come from various sources, including healthcare professionals (doctors, nurses, pharmacists), patients, consumers, and manufacturers. Healthcare providers are often required to report adverse events that they become aware of during patient care.

Regulatory agencies, such as the U.S. Food and Drug Administration (FDA) in the United States or the European Medicines Agency (EMA) in Europe, maintain databases of AERs. They use these reports to monitor the safety of medical products and, if necessary, take regulatory actions, such as issuing warnings, recalls, or label changes.

The science and activities related to the detection, assessment, understanding, and prevention of adverse effects or any other drug-related problems are collectively known as pharmacovigilance. AERs are a fundamental component of pharmacovigilance efforts.

Depending on the country’s regulatory framework, there may be legal requirements to report adverse events associated with medical products. Failing to report serious adverse events can have legal consequences.

Recommends No Code applications for someone with limited or no experience with coding.

Says his experience with these tools is dated. These tools depend on componentisation of code and componentising code is not easy. But they might have improved over time.

Imagine a set of managers are to collectively make a decision: their intuitions may prompt them to take diagonal and opposing views. Complete consistency is unlikely, and decisions are usually negotiated between competing perspectives. In contrast, decisions based on data are more likely to exhibit consistency, regardless of who is making those decisions.

So, depending solely on intuition is not a wise approach. However, if you could resolve 80% of the ambiguity with the help of data that is available, the rest can be based on ones’ intuition. Then, it is legitimate.

Ultimately, any decision one takes should be able to withstand the scrutiny of an audit. If decisions are taken solely based on instinct, they will not pass the test.

When we are entering a new market for existing products, or launching a new product, there may be no internal data available. And one needs data for creating a blue-ocean strategy.

In such cases, it will be ideal to source the ‘data required’ through a survey of the target market segment; and use a random sampling technique to simulate the population data.

While you may or may not get it right the first time, both the quality of the data and the model will improve over time.