Site Speed and Conversion: What to Fix First
Speed matters, but not uniformly. Going from painfully slow to acceptable changes behaviour. Going from good to excellent usually changes a score. Knowing which situation you are in saves a lot of wasted effort.
Find the actual problem before optimising
Run your site through PageSpeed Insights and read the field data section — real measurements from real visitors — before the lab score. Lab scores are useful for diagnosis and terrible as a target.
Then check on a mid-range phone over mobile data, not the desktop you built it on. Most small business traffic is mobile, and most performance problems are invisible on a fast machine and a fast connection.
Where the wins usually are
- Images. Still the biggest single cause. Uncompressed hero images of several megabytes are routine. Converting to WebP and sizing them for their display size typically cuts page weight more than every other optimisation combined.
- Third-party scripts. Chat widgets, tag managers, tracking pixels, embedded video players. Each is someone else's JavaScript on your critical path. Audit them and remove what nobody looks at.
- One enormous JavaScript bundle. If every visitor downloads the code for every page, route-level code splitting means they only download the page they asked for.
- Fonts. Two weights of one family is usually plenty. Every extra file blocks text from appearing.
A note on splitting code
Code splitting is worth doing, but it can be overdone. When we split our own site we initially produced 48 JavaScript files, 23 of them under 3KB — a separate chunk per icon. That is more network round trips, not fewer, and it makes deployment fragile: a partial upload leaves the site broken in ways that are hard to diagnose. Consolidating dependencies into a single cached vendor chunk brought it down to 10 files.
The goal is fewer, well-sized files — not the largest possible number of small ones.
What speed will not fix
If visitors arrive and leave, the cause is more often clarity than latency. An unclear offer, a hidden price, a form asking for ten fields when three would do, or no obvious next step. These cost more conversions than a second of load time, and they are cheaper to fix.
Measure before you optimise, and be honest about whether performance is really your bottleneck.
Digova offers conversion rate optimization and performance work for businesses across the GTA.