Case Study
Increasing growth by 38%
The Modern Milkman is a subscription-based milk and grocery delivery service operating across the UK. By 2022, the product had strong brand recognition but a conversion problem: thousands of users were signing up and never placing an order.
My role was to diagnose why, and design an onboarding experience that turned sign-ups into genuine subscribers.
Product Designer
Q1-2 2022
Product Design
Problem statement
Sign-up numbers looked healthy on paper, but a growing cohort of users we called "Sign-Up-No-Purchase" (SUNP) were masking the real picture. These were people who'd registered but never placed an order, inflating our user base while contributing nothing to revenue.
A secondary issue emerged in the data: basket values dropped noticeably mid-week, disrupting delivery logistics and directly affecting driver income and shift availability.
The business needed answers to three questions:
1. Why weren't new users converting?
2. How could we build habitual repeat purchasing?
3. Could design intervention address the mid-week slump?
Research
I started by mapping what we already knew. Analytics showed the sharpest drop-off point was immediately after registration, users were reaching the product catalogue and abandoning. But data alone couldn't explain why.
I observed sessions with the Customer Service team, who were fielding complaints daily. Two patterns emerged quickly: users didn't know how much milk their household actually needed, and they found the open catalogue overwhelming without a starting point. Nobody knew where to begin, so many didn't.
Shadowing CS also surfaced something the analytics had missed, a significant number of SUNP users had contacted support asking basic questions ("how does the delivery work?", "can I pause?") and received no response before churning. The product wasn't answering questions the user had before they'd even built a basket.
The insight that shaped everything
Users didn't need more choice, they needed a confident starting point. The job wasn't to improve the catalogue, it was to remove the blank canvas.
Design Rationalle
Before moving to wireframes, I mapped three possible intervention types:
Guided tooltip walkthrough?
Step-by-step instructions overlaid on the existing UI.
Ruled out:
It teaches the interface, not the decision.
Simplified starter catalogue?
A curated subset of products for new users.
Ruled out:
It still requires the user to make every choice from scratch.
Pre-filled basket based on household size
A suggested first order, tailored and editable. This addressed decision fatigue directly, gave users a confident starting point, and created an implicit product education moment as they reviewed and adjusted items.
Three steps to customer success
01
Users sign up as normal (name, address, email) with one addition: household size.
This single data point does the work of a much longer preferences quiz.
02
Based on household size and real-time product availability, the user lands on a pre-filled basket rather than an empty one.
Suggested items are distributed across the delivery week, exposing users to the subscription model naturally rather than explaining it with copy.
03
The user can accept, edit, or remove items directly from this screen.
Critically, they never have to browse the full catalogue to get started, but the door is open if they want to.
This respects confident users while protecting anxious ones.
Build phase
Wireframes
Early wireframes explored how much information to surface at the pre-fill screen. The first version showed full product detail: name, price, delivery day, quantity. User testing revealed this was still overwhelming. We stripped back to name and delivery day only, with price visible on expansion.
Iteration 1
By the second iteration, we introduced a simple progress indicator at the top of the flow. Testing showed users who could see "Step 2 of 3" were significantly more likely to complete the flow than those who couldn't gauge their progress.
_2x.png)
High Fidelity
The final design used The Modern Milkman's existing component library with one new component: the editable basket card. I worked closely with engineering during this phase to ensure the pre-fill logic (household size × product availability × delivery day distribution) could be implemented within the existing order management system.
Results
The design was handed off during a company restructure before I could see it through to launch, a frustrating moment, but one that taught me a lot about documentation and handoff rigour.
The team implemented the designs several months later and the subsequent numbers validated the approach.
The mid-week result was particularly satisfying, it wasn't directly targeted by the design, but fell out of the pre-fill logic naturally. That's usually a sign you've solved the right underlying problem.
+38%
Overall growth
Large reduction
In SUNP users
Uplift
across all household size segments
Reflection
What I'd do differently
I'd push harder for a post-launch observation period before handoff. The SUNP reduction told us the onboarding worked, but I'd want qualitative data on why users accepted the pre-filled basket versus edited it heavily. That behaviour pattern would have informed the next iteration significantly.
What this project taught me
Subscription products live or die on the first five minutes. If a user doesn't feel confident placing their first order, the product never gets the chance to prove its value. The best thing design can do in that moment isn't explain; it's decide, provisionally, on the user's behalf.
