from reviews to revenue: 

How ORRO used AI and text analytics to uncover consumer insights hiding in plain sight

The Challenge

Online reviews are one of the most abundant and underused sources of consumer intelligence available to Consumer Packaged Goods (CPG) brands today. Thousands of customers describe exactly what they love, what disappoints them, and why they bought a product in the first place — in their own words, unprompted, at scale.

The problem isn't access to the data. It's making sense of it.

A brand manager can't read 10,000 Amazon reviews. A traditional research firm would charge six figures to run the focus groups that might surface the same insights. And a simple star rating tells you nothing about why customers feel the way they do.

ORRO set out to demonstrate what was possible when you apply rigorous text analytics and AI to publicly available review data in the competitive electrolyte drinks category — a crowded market with brands including Gatorade, Bodyarmor, Prime, Powerade, and Electrolit.

What We Did

We built a five-stage analytics pipeline:

1. Gather — We collected thousands of reviews from major retail websites across multiple brands, translated non-English reviews using ML, and supplemented with ad transcripts and other brand text to build a rich corpus.

2. Discover — Using topic modeling, we identified the major themes consumers actually talk about: taste, flavor, nutrition, energy boost, thirst quenching, packaging, side effects, and storytelling/brand — among others. Critically, we let the data surface these categories rather than imposing them in advance.

3. Build Personas — We identified six distinct consumer behavioral segments from the review data: Nutrition Conscious (70%), Flavor Champion (17%), Productivity Seeker (4.5%), Fitness & Health (4%), Critical Reviewer (2.9%), and Social Consumer (1.4%). Each persona has a meaningfully different relationship with the product category.

4. Label with LLMs — We used large language models to label each review across our topic taxonomy, capturing not just whether a topic was mentioned but the sentiment associated with it at a granular level.

5. Find Insights — We ran logistic regression models connecting text-based topic scores to star ratings, revealing what actually drives high and low consumer satisfaction — by persona.

What We Found

The Taste Paradox

The single most counterintuitive finding: taste — specifically bad taste — was the dominant driver of low star ratings, more predictive than any functional benefit claim. Consumers evaluate energy and electrolyte drinks as beverages first and performance products second. No amount of caffeine, vitamin content, or electrolyte messaging can rescue a product that fails the basic drink test.

We called this the Taste Paradox: the category markets primarily on functional benefits, but consumers judge primarily on sensory experience.

The Sum is Greater Than the Parts

Overall topic sentiment was 25 to 67 times more predictive of star ratings than any individual ingredient mention. Consumers don't rate products based on specific features — they rate based on holistic impressions. This has direct implications for how brands should frame their marketing:

• Holistic perception outperforms granular feature claims

• Emotional response outperforms rational specification

• Simple messaging outperforms detailed ingredient lists

Personas Require Different Messages

The Nutrition Conscious consumer (70% of reviewers) is the dominant segment — and sweetness claims actually hurt satisfaction scores for this group. Their strongest satisfaction driver was overall positive nutrition perception, not any single ingredient.

The Flavor Champion (17%) is almost the mirror opposite: taste attributes dominated their satisfaction scores, and nutrition barely registered.

A single marketing message cannot serve both groups effectively.

Context of Use Matters More Than Expected

Products associated with multiple usage contexts received significantly higher ratings. Products purchased to address a specific medical or remediation need (gastro issues, illness recovery) consistently underperformed — not because the products failed, but because the expectations were misaligned with the product's actual profile.

What This Means for Brands

For product development: taste is non-negotiable table stakes. Functional benefits are differentiators only after the beverage experience clears a baseline threshold.

For marketing: know which persona you're speaking to. The message that resonates with a Nutrition Conscious buyer will fall flat — or actively alienate — a Flavor Champion.

For consumer insights: the data to answer these questions already exists in your review corpus. The question is whether you have the tools to ask it.

Interested in what your review data could tell you?

ORRO builds bespoke text analytics solutions for organizations with complex, unstructured data challenges. Every engagement is custom — we work with your data, your questions, and your domain.

Contact us to start a conversation.

© 2026 ORRO AI Genius, LLC. All rights reserved.