Using data science to deliver revenue-enhancing insights to bar owners

An automated data-science platform powering a digital product for Carlsberg customers. It automatically ingests, cleans and analyses multiple datasets – Carlsberg, POS and third-party – and generates tailored recommendations in simple, human language.

Approach

Primary research uncovered that the savviest outlet owners were already analysing spreadsheet data to make better planning decisions. Carlsberg could help them do this far more effectively – and serve those who weren't yet working this way.

Using data science to deliver revenue-enhancing insights to bar owners
Using data science to deliver revenue-enhancing insights to bar owners

As a small product team – designer, product manager, engineers, data scientists and business stakeholders – we validated the concept with customers using a range of prototyping techniques across three key markets in Europe and Asia. Despite being a data product, we led with customer need rather than available data.

A small product team – designer, product manager, engineers, data scientists and business stakeholders
A small product team – designer, product manager, engineers, data scientists and business stakeholders
We validated the concept with customers using a range of prototyping techniques across three key markets in Europe and Asia
We validated the concept with customers using a range of prototyping techniques across three key markets in Europe and Asia

We worked in short iterative cycles, prototyping through to build, shipping new features every two weeks. An 'insight sprint' was introduced to experiment with new data sources – geo, time, event, weather, social, POS – weekly, using a bespoke in-app scoring system to gather customer feedback rapidly.

The result

An app that turns insights into conversational recommendations that are easy to scan, interpret and action – designed for people running busy, noisy bars.

An app that turns insights into conversational recommendations that are easy to scan, interpret and action
An app that turns insights into conversational recommendations that are easy to scan, interpret and action

Outcomes

  • +240k insights, automatically generated monthly, most from a previously untapped data source
  • 10 diverse data sources ensuring solutions for all scenarios
  • Resulting in high engagement - 89% of customers returning in the past 30 days and an average insight rating of 5.8/7
  • In Denmark, we’ve seen an average 3150DKK ($475) incremental spend per month of Adapt customers since February 2019
  • In Switzerland, we’ve seen an average 2100DKK ($315) incremental spend per month of Adapt customers since March 2019

Achievements

  • Continuous user testing in four markets across the globe
  • Guided team decision-making and established new processes and methodologies, including insight sprints
  • Sub-brand development, including a brand values and principles workshop with senior stakeholders
  • Co-created the proof case for the continued investment business case
  • Presented designs and product strategy to steering groups, the CEO and at board level
  • Led two teams concurrently