Turning thousands of customer comments into a clear to-do list
How a large online retailer stopped reading comments one by one and started seeing what customers kept raising, saving over 2,700 hours of manual review.
Case studies
The problem, how it was solved and what changed, told without the jargon. These projects come from our founder's earlier roles. Client names are kept private.
Showing shoppers the right products first
How a luxury resale marketplace used past shopper behaviour to decide which products to show first, and cut its product page exit rate by 6%.
Suggesting what to buy next, and adding 4% to revenue
How a luxury resale marketplace built a recommendation system from product details, helping customers find more they liked and lifting revenue by 4%.
Speaking to buyers and sellers differently, and making marketing 35% more efficient
How grouping buyers and sellers by behaviour let a resale marketplace send more relevant messages and improve its marketing efficiency by 35%.
From messy app data to numbers everyone agrees on
How reorganising tangled app data into a clean structure cut data discrepancies by 90% and made reporting around 50% faster for a consumer app.
Fixing the tracking behind a rewards store, and lifting engagement by 20%
How rebuilding event tracking for a loyalty rewards store showed what shoppers were doing and helped lift engagement metrics by 20%.
Giving a delivery operations team back more than 600 hours a year
How automating manual reporting saved 624 hours a year and faster data pipelines made queries over 80% quicker for a global e-commerce operations team.
Choosing the right customers to ask, and lifting conversion from 2.5% to 4%
How a predictive model replaced guesswork in a customer advocacy programme, lifting conversion from 2.5% to 4% and cutting the effort of keeping it current.
Finding what customers buy together, and adding $20,000 in a month
How looking at which products customers bought together produced upsell rules that brought in an extra $20,000 in a single month.
Setting the right price for ad space, and automating weekly client reports
How data-led minimum prices for ad slots and an automated weekly reporting tool helped a digital advertising business and saved 208 hours a year.
Spotting orders that were likely to be cancelled, before they were
How a prediction model for an online marketplace flagged orders at risk of cancellation early, so the team could step in instead of finding out afterwards.
Forecasting demand so planning could start from numbers, not hunches
How a time series forecast, tested against a plain baseline and shown with a realistic range, gave an e-commerce team a better starting point for planning.
Using a language model to read what customers write, at any volume
How a language model, guided by themes the business defined and checked by people, turned reviews, questions and comments into decisions.
Putting AI assistants to work in an analytics team, with guardrails
How an analytics team used language-model tools to speed up routine drafting, such as queries and report commentary, while people stayed responsible for the numbers.
Have a similar problem?
Tell us what you are trying to solve. A short conversation often shows whether a project like one of these could help.