Store speed and Core Web Vitals work that starts with measurement rather than a checklist. The bottleneck is usually not where the team assumes, which is why the guesses have not worked.
In one paragraph
How do you actually make an eCommerce store faster?
By measuring first. Server response, database queries, cache hit rates, index freshness, third-party scripts, image delivery and render-blocking resources are all candidates, and on any given store one or two of them dominate while the rest are noise. The work is finding which, fixing those, and measuring again. On the record: site speed improved by 60% through Varnish, Redis, full-page cache, CDN and database tuning. The checklist approach, applying every known optimisation, usually moves the number a little and costs a lot.
What this is
Almost every slow store has already had things tried on it. Images compressed, a CDN added, a caching extension installed, lazy loading switched on. Sometimes it helped a little; usually the number barely moved, because the actual bottleneck was somewhere nobody looked.
That is the whole argument for measuring before optimising. A store where the database is doing a full scan on category pages will not be fixed by better images, and one where twelve third-party scripts block render will not be fixed by a faster server.
What this page is not
This is diagnosis and fixing. If the storefront architecture itself is the ceiling, the honest answer is sometimes a decoupled frontend rather than more tuning, and that is a bigger decision.
What usually brings this in
Core Web Vitals will not improve no matter what we try.
Paid traffic is landing on a page that converts below what it should, and the spend is unchanged.
The site is fine until traffic arrives.
It falls over on exactly the days that matter, and the response has been to over-provision and hope.
Category pages take seconds to load.
Shoppers abandon before the products appear, and the catalogue the business has invested in is effectively hidden.
Someone added a caching extension and nothing changed.
Money spent on a fix for a problem that was never diagnosed, and now less confidence that anything will work.
What gets built
- Measurement before anything else
- Server timing, query profiling, cache hit rates, index state and real-user Core Web Vitals. The deliverable of this step is knowing which two things matter.
- Backend and database
- Query profiling, missing indexes, N+1 patterns and the queries that only become a problem at your data volume.
- Caching architecture
- Full-page cache, Varnish, Redis and the invalidation rules, because a cache that is correct but never hit is a cache that does nothing.
- Search and indexing
- Elasticsearch configuration and reindex strategy, which is where large-catalogue stores usually lose their category pages.
- Frontend delivery
- Render-blocking resources, image formats and sizing, font loading, and the third-party scripts nobody has audited since they were added.
- Measured again, and written down
- Before and after on the same conditions, so the improvement is a number rather than an impression. This is the part that makes the next decision easier.
Technologies
- Caching
- Varnish
- Redis
- Full-page cache
- Cloudflare
- CDN
- Data
- MySQL profiling
- Elasticsearch
- Query optimisation
- Indexing
- Frontend
- Core Web Vitals
- LCP
- INP
- CLS
- Image optimisation
- Measurement
- Lighthouse
- CrUX
- Real user monitoring
- New Relic
How delivery runs
Measure the real thing
Production conditions, production data, real devices and real networks. A local profile of a store with twelve test products tells you nothing about one with two hundred thousand.
Find the two that matter
Rank the candidates by how much time each actually accounts for. On most stores one or two dominate and the rest are rounding errors.
Fix those, in order
Largest first, measured after each. Changing six things at once means not knowing which one worked, which matters the next time something slows down.
Report the before and after
Same conditions, same measurement, written down. A performance engagement with no number at the end is indistinguishable from one that did nothing.
Leave the monitoring behind
So the next regression is visible when it happens rather than when someone notices months later.
Relevant builds
- Magento 2 · Marketplace
Multi-category marketplace
A scalable Magento 2 platform powering a wide multi-category retail catalogue.
- Magento 2
- Multi-category
- Scale
- Custom platform · B2B
Procurement and approvals platform
Scalable B2B procurement platform streamlining purchase requests and approvals.
- B2B workflows
- Approvals
- Integrations
- Headless commerce · Fashion
Headless fashion storefront
High-performance headless commerce architecture for India’s fastest-growing men’s fashion brand.
- Headless
- Storefront API
- Performance
Questions
- How much faster will our store get?
- I will not quote a number before measuring, and anyone who does is guessing. What I can say is what the record shows: 60% improvement on a large multi-store Magento platform through caching, CDN and database work. Your store has its own bottleneck and its own ceiling.
- Is a performance audit worth it on its own?
- Often yes, and it is the cheapest way to find out whether the work is worth doing. The audit tells you what is wrong and roughly what fixing each thing costs, and you can then decide, including deciding not to.
- Will this help SEO?
- Core Web Vitals are a ranking factor, so improving them helps, and it is rarely the largest factor. The more reliable benefit is conversion: a faster checkout loses fewer people who had already decided to buy.
- Do we need to replatform to be fast?
- Usually not. Most slow stores are slow because of configuration, queries and third-party scripts rather than because the platform is incapable. Replatforming for speed alone is an expensive answer to a question that has not been diagnosed.
Describe what is in front of you and you get a written scope and a figure, usually within a few days of a short call. No charge for either.