Why Only 1 in 5 Consumer Goods Companies Use Data Science

Despite all the buzz about AI and machine learning, the reality on the ground is sobering: only an estimated 10–20% of consumer goods companies have real data science capability.

This isn't surprising given the landscape. Yet what's interesting is how the conversation has shifted recently. 71% of CPG leaders reported adopting AI in at least one business function in 2024, up from 42% in 2023. So more companies are experimenting with AI. But actual data science adoption — the kind that transforms decision-making — remains stuck in the low teens.

Why the gap? Because adoption isn't a technology problem. It's a people problem.

The Adoption Reality: What the Data Shows

The numbers paint a complex picture. Yes, 71% of CPG leaders have adopted some form of AI or machine learning in at least one function, and 56% regularly use generative AI. But this tells only part of the story.

What's actually happening in most organisations:

• Machine learning pilots and experiments — Teams are testing ML models for demand forecasting, promotion optimisation, and price elasticity, but most aren't scaling beyond pilots. Among large retailers and CPG brands with over $500 million in revenue, 64% are using some form of AI or machine learning, but that usage is often confined to specific use cases, not embedded across the business.
• No true machine learning capability — Running a forecasting model is different from having machine learning capability that continuously improves. One is a one-time project. The other is a competitive advantage.
• Talent as the bottleneck — CPG companies struggle to hire and retain machine learning engineers and data scientists. Even when they do, those specialists often operate in isolation, delivering model outputs that the business doesn't know how to act on.

The result: most machine learning in CPG remains experimental, not embedded in how teams actually plan.

Why Teams Can't Understand the Insights

Here's the uncomfortable truth that vendors won't tell you: data scientists often sound like they're speaking an alien language.

A data scientist builds a model, validates it with rigorous statistical methods, and presents the findings using terminology like "p-values," "coefficients," "model variance," and "feature importance." To them, it's perfectly clear.

To a sales manager or finance director? It's noise.

The result: models become black boxes. Teams tune out. The insights get lost in translation. And the data science investment — which likely cost six figures or more — delivers minimal business impact.

This is a real phenomenon. Teams don't distrust data science because it's complicated. They distrust it because they don't understand it. And what you don't understand, you don't use. And what you don't use, you eventually stop funding.

The Real Barriers to Adoption

It's not that data science is hard. It's that integrating it into a business is harder.

The typical path looks like this:

• Data is scattered — Sales data lives in one system. Trade data in another. Financials in a third. Before you can do any analytics, you need to spend weeks or months just pulling data together and validating it.
• Causation is obscured — A demand shift could be caused by a promotion, a competitor move, a seasonal trend, or a distribution change. Without seeing the full picture, you can't separate signal from noise.
• Business context gets lost — A statistical relationship that's statistically significant isn't always business-significant. A model might find that a promotional discount correlates with higher sales — but does it drive incremental sales, or just pull demand forward?
• Spreadsheets remain the default — Most CPG companies still use Excel for planning. Getting insights from a data science model into Excel, and then operationalizing those insights, is a manual, error-prone process.
• Translation breaks down — Even when insights are delivered, the gap between "here's what the data shows" and "here's what you should do about it" often goes unbridged. Sales hears "your forecast is biased" and gets defensive rather than curious.

The companies with real data science capability have solved these problems. But most haven't.

Turning Complexity Into Clarity

This is where the design of your planning system matters.

Most data science is delivered as a report, a dashboard, or a separate module. It's advisory. It sits alongside your decision-making process, but it's not integrated into it.

What if data science was built into how you plan?

This is what CauSelf was designed for. We don't ask you to become fluent in data science. Instead, we:

• Connect the messy data — Sales, trade, financial, promotional data all flow into one integrated planning environment. No manual stitching. No data validation nightmares.
• Explain what really drives results — in plain language — When you run a scenario in CauSelf, you see not just the forecast, but the drivers: how much uplift comes from the promotion, how much from seasonality, how much from distribution changes. The "why" is transparent.
• Link forecasts and promotions back to financial impact — You see instantly how a demand shift affects margin, cash flow, and inventory investment. Data science insights flow directly into business impact.
• Handle new products with confidence — New product forecasting is where most data science fails. There's no historical data, so statistical models struggle. CauSelf handles this by combining available analogues, market research, and sales judgment in a systematic way that produces defensible forecasts even when data is sparse.

This is data science without the jargon. It's analytics built for business users, not for statisticians.

Why This Matters Right Now

The CPG companies winning in 2026 are those making data-driven decisions faster than their competitors. Not by hiring more data scientists or running bigger machine learning projects. But by embedding intelligence directly into how they plan.

Data science adoption may be stuck in the low teens across the industry. But it doesn't need to be low for you. You don't need a team of PhDs. You don't need a multi-year AI transformation project. You don't need black-box models that your team doesn't trust or understand.

You need planning software built on data science principles: connected data, transparent logic, business-relevant insights, and instant feedback on financial impact.

That's how you move from "we have an AI project" to "data science is how we plan."