Use case · Engineering & R&D

AI consumer insights for product development

Language models and machine learning mine product reviews, social posts, support tickets and sales data to show what customers want, so product teams pick stronger concepts and reach launch sooner.

Maturity
Emerging
Complexity
Low (our assessment)
Department
Engineering & R&D
Technology
Generative AI & LLMs, Machine learning
Solves
Long lead times, Defects reaching customers, Excess inventory

Evidence at a glance

1documented deployment
1manufacturer
1independently assessed
Independently assessedManufacturer-reportedVendor-reported

Reported results

1 quantified outcome across 1 deployment

Independently assessedManufacturer-reportedVendor-reported

Each dot is one reported improvement, as published by its source; the black tick is the median. Figures come from different baselines and methods, so read them as the range of reported results, not a forecast.

Who has done it

Kenvue · Shanghai

Big data and AI mine consumer feedback for product-fit insights in new product introduction.

Independently assessed
Shanghai, China · 2023
50% Time to market (new product introduction)
View the data as a table
ManufacturerEvidenceOutcome as reported
KenvueIndependently assessed50% Time to market (new product introduction)

What it is

AI consumer insights for product development uses language models and machine learning to read large volumes of customer feedback, such as online reviews, social media, retailer comments, warranty claims, support tickets and survey responses, and turn it into structured findings about unmet needs, complaints, emerging preferences and competitor gaps. Product, R&D and marketing teams use those findings to choose which concepts to develop, which features or flavors to prioritize, and which problems in existing products to fix first.

It matters because new product introduction in consumer-facing manufacturing, such as food and beverage, household goods, appliances and personal care, is slow and many launches miss. Traditional research relies on surveys and focus groups that take weeks and capture a small sample. Customers already describe what they like and dislike in their own words, in large volumes, but few teams have time to read it. Automated analysis lets teams test ideas against real feedback earlier and with less guesswork.

How it works

  1. Frame the product question

    Product and R&D define what they need to learn: why a line is losing share, which attributes matter in a category, what complaints recur for a product, or which adjacent needs competitors are not meeting.

  2. Gather the feedback

    Reviews, social posts, retailer data, support tickets, warranty claims and internal survey results are collected for the company's own products and for competitors, within the terms of each source and applicable privacy rules.

  3. Structure the language

    Language models classify comments by product attribute, sentiment and theme, and group them into topics such as taste, durability, packaging, ease of use or price, while linking each theme back to example quotes.

  4. Connect to the business

    Themes are compared with sales, returns and quality data to separate loud but rare complaints from issues that actually move purchases, ratings or warranty cost.

  5. Feed concepts and testing

    Findings shape concept briefs, formulation or design targets and test plans, and the same analysis is rerun after launch to check whether the new product answered the need.

Benefits

Faster early research

Teams can answer category and concept questions from existing feedback in days rather than waiting for a new survey or focus group round, which shortens the front end of development.

Concepts grounded in customer language

Briefs built on recurring themes and verbatim comments give engineers and formulators concrete targets instead of loosely defined marketing goals.

Earlier warning on product problems

Complaints about a defect, packaging failure or formulation change show up in reviews and tickets before they appear in returns data, which gives quality and engineering a head start.

Fewer launches that miss

Testing concepts against what customers already ask for helps teams drop weak ideas before committing tooling, packaging and inventory to them.

Data & prerequisites

Data you will need

  • Online product reviews and ratings for own and competitor products
  • Social media and community discussion relevant to the category
  • Customer support tickets, call notes and warranty claims
  • Survey and concept test responses
  • Sales, returns and quality data by product and region

Before you start

  • A clear product question or category focus, rather than a general request to analyze everything
  • Rights and processes for collecting external data in line with platform terms and privacy law
  • Product, R&D and marketing teams willing to act on findings in concept and design decisions
  • Access to internal sales, returns and support data to validate what feedback suggests

Common pitfalls

  • Treating online reviews as representative of all customers, when they skew toward strongly positive or negative buyers
  • Accepting model-generated summaries without checking them against the underlying comments
  • Producing insight reports that never reach the people writing specifications and formulations
  • Collecting external data in ways that breach platform terms or privacy rules

Questions to ask vendors

  1. Which data sources do you cover, and how do you collect them within each platform's terms?
  2. How do you classify themes and sentiment for our category, and can we see the comments behind each finding?
  3. How do you handle fake reviews, duplicates and sampling bias?
  4. Can you combine external feedback with our own support, warranty and sales data?
  5. How do our product teams query the data themselves rather than waiting for a report?

What to measure

  • Time from concept idea to approved development brief
  • Time to market for new products
  • Share of launched products meeting first-year sales or rating targets
  • Average ratings and complaint themes for launched products
  • Warranty claims and returns tied to issues seen first in customer feedback

Frequently asked questions

Is this only useful for consumer brands?

It is most common in food and beverage, household and personal care products, and appliances, where public reviews are plentiful. Industrial manufacturers can apply the same methods to support tickets, distributor feedback, field service notes and warranty claims, which often contain the most specific product feedback available.

Can AI replace surveys and focus groups?

It reduces how often they are needed and makes them more targeted, but it does not fully replace them. Existing feedback shows what current customers say about current products. Testing new concepts, prices or formulations with target buyers still requires primary research.

How do we know the summaries are accurate?

Good tools link every theme and statistic back to the underlying comments so analysts can read examples and check the classification. Teams should sample and review results regularly, especially when the model is applied to a new product category or language.

Related use cases

Supply chain & procurement

Demand forecasting

Machine learning combines order history with external signals to forecast demand more accurately by product, customer and location.

Machine learning
12 documented deployments4 software companies

Engineering & R&D

Generative design & engineering simulation

AI generates design options from requirements and replaces slow simulations with fast surrogate models so engineers explore more designs.

Machine learningGenerative AI & LLMsOptimization & planning
12 documented deployments3 software companies

Sales, quoting & service

Field service & aftermarket support

Use connected-equipment data and AI to diagnose customer machines remotely, plan service visits and sell the right parts and services.

Machine learningGenerative AI & LLMsIndustrial IoT & edge
0 documented deployments0 software companies

Quality

Root-cause analysis for quality & yield

Machine learning sifts process, material and equipment data to narrow down which factors are actually driving defects and yield loss.

Machine learningGenerative AI & LLMs
13 documented deployments8 software companies

Put AI to work

Find the AI opportunities hiding in your plant.

Genvo, our implementation partner, works with manufacturers to pick the right first project, connect the systems you already run and put AI agents and tools into daily use, with people approving what goes out.

Assessments and consultations are provided by Genvo. How our partnership works.