Website Optimization: The Order I Actually Fix Things In

The old version of this post promised you 200-400% ROI within 6 months from technical optimization. It had a table for it. I don't know where that number came from, and I wrote the post.

That's the problem with most "complete guides" to website optimization, mine included. They list six areas, attach a percentage to each, and send you off to work on all of them at once. You end up compressing images, rewriting headlines, installing a heatmap tool and A/B testing button colors in the same week. Then nothing measurably changes, because you changed everything and measured nothing.

So I threw the old version out. This is the order I actually work through when someone asks me to "optimize" their site. It's shorter and has fewer promises in it. The Core Web Vitals part is also no longer wrong.

Start With One Number, Not Six Areas

The 2019 version was built around a table of six optimization areas: technical performance, UX, SEO, conversion, mobile, content. Every one of those is real. None of them is a goal.

A goal is one number. For a local service business it's usually "contact form submissions" or "calls from the website." For a shop it's completed orders. For a SaaS landing page it's signups. Pick the one action your site exists for and write it down. Everything else in this post gets judged against that number.

Then check that you can actually count it. This sounds like a formality and it's where a lot of optimization projects quietly die. In Google Analytics 4 the action needs to be set up as a key event, and someone needs to submit a test form and see it show up. If your cookie banner sends half your visitors to "declined," your analytics only sees a sample. I wrote about how badly a cookie banner can gut your analytics, along with seven other leaks that never show up on a dashboard.

If you can't count the thing you're optimizing, stop here and fix that first. Every change you make before that point is a guess you'll never be able to grade.

Core Web Vitals in 2026: LCP, INP and CLS

The old post listed First Input Delay (FID) with a 100 millisecond target. FID is gone. Google replaced it with Interaction to Next Paint (INP) in March 2024, and if you're still reading guides that mention FID, they haven't been touched in years.

The three Core Web Vitals right now:

Metric	What it measures	Good	Poor
LCP (Largest Contentful Paint)	When the biggest visible element finishes loading	2.5 s or less	over 4 s
INP (Interaction to Next Paint)	How fast the page reacts to taps, clicks and key presses, across the whole visit	200 ms or less	over 500 ms
CLS (Cumulative Layout Shift)	How much the layout jumps around while loading	0.1 or less	over 0.25

The difference between FID and INP matters more than the name change. FID only looked at the delay before the browser started handling your first interaction. INP looks at every interaction during the visit and reports one of the slowest. A site could pass FID easily and still feel sluggish every time you opened the mobile menu. INP catches that.

Two details most guides skip. Google judges these at the 75th percentile of real visits, split by mobile and desktop. So "it's fast on my laptop" means nothing; three out of four of your real mobile visitors need to have a good experience. And the number that counts is field data from real Chrome users, not the Lighthouse score you get from a lab test. Lighthouse can't even measure INP, because there's no real user clicking anything. It shows Total Blocking Time as a stand-in.

For the testing tools and how to read them, I've got a separate guide on how to test your website's load time, so I won't repeat that here.

For context on how common failing is: April 2026 field data showed only 49% of WordPress sites passing all three Core Web Vitals, against 67% of Astro sites. I broke down where that Core Web Vitals gap between Astro and WordPress comes from in another post. If your site fails, you're in a lot of company.

Fix the Metric That's Failing, Not the Score

Open Google Search Console, go to the Core Web Vitals report, and look at which metric is failing and on which group of URLs. That tells you far more than a single performance score out of 100, because each metric fails for different reasons.

LCP failing usually means the hero image or the server is the problem. Check whether the big image at the top is a 3 MB JPEG, whether it's lazy-loaded (the hero should never be lazy-loaded), and how long the server takes to send the first byte. On cheap shared hosting, that first byte alone can eat most of your 2.5 seconds before the browser has anything to draw.

INP failing almost always points at JavaScript. Chat widgets, tag managers with fifteen tags in them, cookie consent tools, page builder scripts, sliders. Each one competes for the main thread, and when a visitor taps a button, the browser has to wait its turn. The fix is rarely clever code. It's removing scripts nobody uses and loading the rest later.

CLS failing is the most fixable of the three. Images and embeds without width and height attributes, a cookie banner that pushes content down instead of overlaying it, ads or "announcement bars" that load late, web fonts that swap in with different dimensions. Reserve the space, and the jumping stops.

One fix at a time, then wait. Search Console's field data is based on a rolling 28-day window, so a fix you ship today won't fully show up there for about four weeks. If you change five things in that window, you won't know which one worked.

The Conversion Path: Walk It Yourself on a Phone

Speed gets people to the page. Whether they do the one thing you wrote down earlier depends on the path, and the fastest way to find problems there costs nothing.

Take your phone, search for what a customer would search for, click your result, and try to complete the action. Fill out the form. Call the number. Buy the cheapest product. Do it on mobile data, not Wi-Fi. Then check whether the form email actually arrived, and whether it landed in spam.

What you're looking for along the way:

• Can you tell what the business does and where it operates within the first screen?
• Is there one obvious next step, or five buttons competing for attention?
• How many fields does the form have, and does every one of them need to be there?
• Does anything cover the content: a chat bubble over the submit button, a newsletter popup after three seconds?

Most of what I'd tell you to change after that walk is boring. Clearer headline, fewer form fields, one call to action instead of four, proof that's from this decade. I wrote a whole post on these boring fixes that beat a redesign, and none of them require a developer.

If your problem is that people land and leave without doing anything at all, that's a separate diagnosis. My guide to reducing your bounce rate covers what GA4 counts as a bounce now and what usually causes it.

What I Cut From the Old Version

The 2019 post had a "Future of Website Optimization" section with voice search, Progressive Web Apps, AI-powered personalization and "privacy-first analytics." I deleted all of it. None of those is why a small business site loses customers, and putting them in a guide just gives you more things to feel behind on.

I also cut the ROI tables with their "expected impact" percentages next to every button color and form field. Same for the timeline promising "significant traffic and conversion gains" by month six. I can't back those numbers up. Results depend on how broken the site was to begin with. Fixing a contact form that silently stopped sending emails can matter more than a year of tweaking headlines, and no table predicts that.

And the "three-click rule" for navigation is gone too. It sounds precise, and it isn't a rule anyone measured into existence. What matters is whether each click makes it obvious where to go next.

How Often to Check Again

Optimization isn't a project you finish. A plugin update changes how scripts load, or someone uploads a 6 MB photo to the homepage. Either one can undo a month of work without anyone noticing.

What I'd do: look at the Core Web Vitals report and your key event numbers once a month, and walk the conversion path on your phone once a quarter. If you want the full routine, my monthly SEO maintenance checklist covers what else to look at in the same sitting.

Where This Leaves You

Website optimization done right is less exciting than the guides make it sound. One number to move. Core Web Vitals checked against real-user data, with INP instead of the FID most old guides still mention. One fix at a time, with enough time in between to see what it did. And a walk through your own site on a phone, which will probably show you more than any tool.

I can't promise you a percentage. What I can do is look at your site, tell you which of these steps is actually your bottleneck, and fix it. Let's talk if that sounds useful.