Insight & Foresight

Exploring predictive AI/ML capability across research planning and customer journeys

  • Leading B2B bioscience provider of research products

  • Trusted by global researchers via validation data, peer reviews and citations

  • ~750,000 researchers / ~100K products / ~130+ countries

“Graph UI - Complex data visualisation, enabling future trends and outcomes.”

Overview

Data Science, UX and Product Management collaboration: Exploring a predictive ML Graph UI that integrates into customer journeys, using experiment attributes to surface relevant products and adjacent protein pathways, highlighting future areas of study. Assessement of how to scale and introduce new customer capability without disrupting existing revenue-critical journeys.

 

  • Partnered with Data Science leadership to align strategy and customer experience requirements

  • Developed customer user flow concept - scaling from one to a collections of tools

  • Explored POC UI alongside Data Science & Design Ops

  • Conducted stakeholder engagement and risk assessment for search conversion impact

  • Collaborated with Product Management and Engineering on POC refinement

  • Advocated for success metrics and phased integration to de-risk existing primary revenue journeys

Customer time challenge

  • Time-limited research grants, creating a high-pressure environment to get published

  • Planning, identifying products, execution, results analysis, abstracts and citations all take time

  • Discovery involves navigating large catalogues of products and data

  • Complex protein associations can be prone to missed opportunities

Objectives

  • Leverage ML graph UI tool to identify associations between products for experiment planning

  • Simplify and accelerate complex product exploration

  • Prototype journeys using Graph-UI to explore planning phase, and potentially product decisions

Why predictive modelling

  • Customer: Single antigens can lead into multiple pathways, simplifying analysis and product choice changes the ‘getting published’ timeframe

  • Business: Predicting demand early enables investment ahead of the curve

Key outcomes at a glance

  • 75% average reduction in planning time / ~3 weeks → ~3–5 days

  • First venture into exploring a future digital product portfolio alongside traditional manufacturing

  • POC of predictive analytics powering customer journeys to identify future product trends

  • Offering customers digital services that can be scaled across manufacturing portfolio

  • Improved cross-team collaboration - UXR, Design Ops, Product Management and Data Science

  • Exploration and experimentation to drive customer value

  • Improved customer journeys across existing digital channels

 

Associated products - Introducing purchasing

Interaction: Explore pathway data across the UI

Auto-centre on attributes of interest

Research & discovery

“Future opportunities for experience improvements and growth are often found in existing customer feedback data”

Methods

  • Mapping existing search journeys and behaviours

  • Designing new flows - from existing to new capability

  • Simplification of early prototypes

  • Evaluating internal and external data sources:

    • PubMed, product reviews, UniProt database

  • Assessing end-to-end journey and complexity as a digital search experience

  • Feedback sessions on both tool capability and the journey flow

Design approach & solution exploration

Concepts & prototyping

  • Simplification to minimum required UI input to return results - encouraging easy adoption

  • Potential to upscale and build upon results with further inputs for complex models

  • User flows: AI tools, landing pages, product quick view, purchasing journey

Iterations & Trade-offs:

  • Speed & stability: critical to adoption, any delays risked abandonment

  • Novelty: new interaction patterns created friction, so required careful onboarding

  • Complexity: risk of overwhelming users with too many options - we knew from existing search journeys many filters and options were left unused

 

All gestures across active Graph UI

Next stage: Simplification of complex terminology

Results

  • Researchers valued exposure to associated pathways not previously considered

  • ≈ 75% average reduction in planning time / (From ~3 weeks → ~3–5 days)
    Faster experiment planning compared to multi-source manual approach

Business Impact (Pre-release)

  • Demonstrated potential to shift from reactive product launches to proactive, trend-aligned portfolio planning

  • Positioned AI/ML as an enabler of digital-first discovery experiences

Learnings & next steps

Successes

  • Revealed deeper connections between experiment attributes - becoming an ‘advanced search’ alongside universally accepted method of e-commerce product search

  • Exposing customers to adjacent research opportunities helping them discover optional or future experiment potential

  • Demonstrating feasibility of predictive journeys within existing channels

Challenges & Learnings

  • Risk of friction if introduced as sole way to discover products, would need to compliment existing capability with option to explore deeper associations

Next Steps

  • Work with PM & Marketing on position as planing tool

  • Identify optimal positioning, exploring planning phase content journeys over search

  • Small scale testing in live environment

  • Explore Phase 2: AI Assisted interactions / Voice over manual

 

Early POC

Early desktop concept

Phase 2 - Planning & ideation

Right Product, First Time
Expanding search capability, improving efficiency and increasing relevancy

View case study