Personalisation

How Smart Sampling Cut Manual Work by 95%

Unlocq.ai helped a leading retailer automate weekly product sampling across BigQuery, Airflow, and ML-based personalisation.

Leading ANZ-based Retail Group Retail / Consumer Goods
A retail campaign planning dashboard showing automated product sampling allocations and customer targeting.

95%

Reduction in manual allocation tasks

3x

Faster campaign turnaround time

500K+

Members served with automated sampling

100%

Automation across AU/NZ campaigns

The automation Unlocq.ai built using Airflow and collaborative filtering has completely transformed how we manage sampling: fast, accurate, and hands-free.

Senior Product Manager, Sampling Automation

Challenge

A leading retail group in Australia and New Zealand was still running a growing product sampling program through manual coordination. Each campaign required teams to match products to eligible members, apply allergen and store rules, and stay within budget across multiple markets.

That process could not keep up with the scale of the program. Launches were slow, manual effort was high, and the team had little room to expand into more campaigns or smaller customer segments.

Approach

Unlocq.ai built an automated sampling engine that handled the end-to-end campaign workflow.

The solution combined:

  • BigQuery for campaign data preparation and eligibility logic
  • Airflow and Google Cloud Composer for orchestration across weekly and ad hoc runs
  • ALS-based collaborative filtering to score member-product affinity
  • Salesforce integration to push allocations back into downstream campaign operations

The retailer could move from spreadsheet coordination to a repeatable system that applied business rules, generated allocations, and supported both high-volume and niche campaigns.

Results

  • 95% reduction in manual allocation work
  • 3x faster campaign turnaround
  • 500,000+ members served through automated sampling
  • AU/NZ coverage across the full campaign program

Want to pressure-test a similar use case?

Let's talk through the data, delivery path, and likely commercial upside.

Get in touch