a16z Podcast: Making the Most of the Data That Matters

Episode Summary

Executive Summary: A16Z panel on modern data strategy argues that “big data” is less about raw volume than speed, context, and decision-making. The founders of Kazina, SnapLogic, and GoodData debate cloud vs. on-prem, predictive analytics, and the role of business users versus data scientists, converging on a pragmatic view: data must be routed to the right place, for the right purpose, with tools and workflows that match business change.

Main Topics: Big data as speed and context, not just volume (Priority: 5/5): Speakers redefine big data as the ability to rapidly turn multiple data sources into business outcomes, rather than simply storing petabytes. Context and agility matter more than size alone. Cloud, on-prem, and the data pipeline (Priority: 5/5): The discussion emphasizes a hybrid future where data flows through pipelines across on-prem and cloud systems, with different technologies serving different jobs rather than one platform replacing all others. The last mile of analytics (Priority: 4/5): GoodData frames its role as delivering analytics to business users outside the data warehouse/Hadoop layer, especially via white-labeled distribution to partners and field users who need understandable outputs. Business outcomes vs. exploratory data hoarding (Priority: 5/5): One view argues failed data projects start by collecting everything and hoping to discover questions later; better projects begin with a defined business problem and then determine what data is needed. Predictive analytics and machine learning (Priority: 5/5): The panel agrees predictive analytics is essential, but disagrees on who benefits most. One view stresses large-scale data science; another says many companies lack enough local data and instead need cloud-scale cross-company learning. Excel, Tableau, and human adoption (Priority: 4/5): The speakers repeatedly note that many users still rely on Excel because it fits familiar workflows and bridges the gap between complex back-end systems and simple row/column analysis. Security, regulation, and data gravity (Priority: 4/5): Adoption depends on where data physically resides and on regulatory constraints. Data should often stay where it is, with analytics brought to it, especially in regulated industries and fragmented regions like Europe.

Key Arguments: Big data is a mindset focused on using data quickly to make decisions, not merely a storage-size threshold. Successful analytics projects start with a business problem and then identify the data needed, rather than collecting everything and searching for patterns. Many companies are still “data bankrupt” because back-end systems do not deliver usable insights to frontline users. The cloud enables new analytics distribution models, especially for data shared across organizational boundaries and external networks. Predictive analytics is now table stakes; companies without it will fall behind. No single technology subsumes others: Spark, Hadoop, data warehousing, streaming, Excel, and Tableau each solve different problems. Data gravity matters: analytics should often be deployed near the data’s location due to compliance, latency, and practicality. The real transformation depends as much on people and leadership (CIOs, CMOs, CTOs, CDOs) as on technology. Excel remains dominant because it matches how people think and work, even when it is not the ideal long-term architecture. White-labeled analytics and embedded distribution can make data useful for external partners and field users who cannot interpret raw warehouse outputs.

Data Points: Data warehouse appliance customer size: 3-4 petabytes - Prat Moghe notes that this was once considered a huge customer size at Netezza. Data warehousing industry size: $10 million industry - Gaurav Dhillon references the historical scale of the data warehousing industry while contrasting it with modern big data use cases. GoodData users: about half a million users - Roman Stanek says most of these users do not know they are using GoodData because it is white-labeled. Google/Amazon-style scale: massive data sets across tens of thousands of companies - Stanek argues cloud analytics can learn from cross-company data where single-company datasets are too small for useful machine learning. Customer example scale: one of the large credit card issuers - GoodData describes a customer whose audience sits mostly outside the firewall, including merchants, issuing banks, and acquirers. Regional deployment challenge: multiple data centers - Stanek explains that European regulations can require fragmentation/balkanization of data across jurisdictions.

Pivotal Quotes: "I sort of define big data as it's a mindset. It's about being really fast about using data to make decisions." — Prat Moghe: Defines big data beyond volume and frames Casina’s view of agile decision-making. "The goal is to be that kind of last mile of analytics." — Roman Stanek: Describes GoodData’s role in delivering analytics to business users and external audiences. "The rising tide of the data lake, we think, will drown out the data warehouse in the fullness of time." — Gaurav Dhillon: Argues that modern cloud pipelines and data lake economics will increasingly overtake legacy warehouses.

Implications: Organizations should adopt hybrid, business-led data strategies: keep data near where it lives, prioritize predictive use cases, and design analytics for actual users. Winning will depend on leadership, not just tools.