HeyRosie

Overview

Hey Rosie is an AI-powered call answering service that ensures businesses never miss an inbound call. It replaces traditional voicemail with intelligent, automated call handling - delivering a faster, more reliable experience at a fraction of the cost of traditional answering services.

I was brought in to run a performance audit and implement targeted optimisations, with a specific focus on improving Core Web Vitals on the homepage (mobile), where performance issues were impacting user experience and technical SEO signals.

The challenge

The homepage mobile variant was struggling with Core Web Vitals, particularly with INP (Interaction to Next Paint) and CLS (Culumative Layout Shift). The desktop version was doing fine, but there was still some room for improvement.

Mobile CWV at the start of the collaboration:

LCP → 1.4s // INP → 243ms // CLS → 0.18

Desktop CWV at the start of the collaboration:

LCP → 2s // INP → 115ms // CLS → 0.01

My role

After a full technical audit, I implemented the following optimisations:

1. Fixed CLS by adding a min-height to elements that were loading at 0px and then expanding
   
   

2. Decreased INP without removing any tracking scripts by manually deferring the loading of a big embed that was keeping the main JS thread busy.
   
   

3. Further optimised INP by swapping computationally-heavy elements for a much lighter alternative without changing the design.
   
   

4. Implemented a custom solution to lazy load images below the fold that weren't being lazy loaded due to Framer limitations.

In other words, we kept the design and marketing stack the exact same - while still being able to improve CWV metrics across the board thanks to targeted yet invisible changes.

Outcome

The Core Web Vitals report can take up to 28 days to fully clear out so changes take a bit to show up. After around 20 days, here are the results:

Mobile CWV after three weeks:

LCP → 1.3s (-0.1s)

INP → 187ms (-56ms)

CLS → 0 (-0.18)

Desktop CWV after three weeks:

LCP → 1.3s (-0.7s)

INP → 114ms (-1ms)

CLS → 0.01