Slow website fix: the page speed problems on marketing sites that matter most

By Ranjit Bhinge, founder of Blur Studio · Updated · 9 min read

Short answer: On most marketing sites, slowness comes from a few things: a large hero image, too many tracking and chat scripts, heavy fonts, and slow hosting. Start by reading your field data (how real visitors experienced the page) in PageSpeed Insights, Google's free page analysis tool, to see which of the three Core Web Vitals fails. Google's thresholds at the 75th percentile are: Largest Contentful Paint (LCP, how fast the main content appears) 2.5 seconds or less, Interaction to Next Paint (INP, how fast the page reacts to a tap) 200 milliseconds or less, and Cumulative Layout Shift (CLS, how much content jumps around) 0.1 or less.

Ranjit Bhinge

Your marketing site feels slow or scored badly in a speed test, and you want to know which fixes matter and which are noise.

I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.

Do not start with the Lighthouse score. Start with which metric fails for real visitors, then fix in a set order: the main image, the tracking tags, the layout jumps, the fonts, then hosting. That order is my recommendation, based on Google's own web.dev guidance listed in the sources. The steps work on any platform. Menu names differ between builders and hosts.

A word on promises: speed fixes help visitors, but no speed score guarantees rankings or leads, so measure the fix by the metrics below, and by what happens to your form submissions and calls.

Step 1: Read the right numbers first

Lab data is one test run in a controlled setting with an empty cache. Field data comes from real Chrome users over the last 28 days, and it is taken at the 75th percentile, meaning 75% of visits were this good or better. Google's web.dev guidance says field data is what you should use to prioritise, and lab data is for debugging.

  1. Go to PageSpeed Insights (pagespeed.web.dev) and enter your page address. Pages must be public and crawlable.
  2. Read the top section, "Discover what your real users are experiencing". That is field data from the Chrome User Experience Report (CrUX, Google's collection of real visit data). If your page lacks enough visits, PageSpeed Insights falls back to data for the whole site. If neither exists, you only get lab data.
  3. Note which of the three metrics fail on mobile.
  4. Only then read the lab diagnostics lower down for the cause.
MetricPlain meaningGood (75th percentile)
LCP, Largest Contentful PaintWhen the biggest thing you see (often the hero image or headline) finishes loading2.5 seconds or less (poor is above 4.0 seconds)
INP, Interaction to Next PaintHow long the page takes to respond after a tap or click200 ms or less (up to 500 ms needs improvement)
CLS, Cumulative Layout ShiftHow much the page jumps while loading0.1 or less

About the score: Lighthouse, the lab tool inside PageSpeed Insights, weights its score as follows in version 10: First Contentful Paint 10%, Speed Index 10%, LCP 25%, Total Blocking Time 30% and CLS 25%. INP is not part of that score. Scores of 90 to 100 are green, 50 to 89 orange, 0 to 49 red. Treat the score as a debugging guide. The three field metrics are the real test.

Step 2: The biggest causes, in the order I would check them

OrderCauseMetric it usually hurtsEffort
1Hero image too big, in an old format, or lazy-loadedLCPLow
2Too many third-party scripts (analytics, ad pixels, chat, heatmaps, consent banner)LCP, INPLow to medium
3No space reserved for images, embeds and late contentCLSLow
4Heavy or late-loading fontsLCP, CLSLow
5Slow server response, no caching or CDNLCPMedium
6Heavy JavaScript running at loadINPMedium to high

Fix 1: The hero image and the LCP element

Find your LCP element: in PageSpeed Insights, open the diagnostics and look for "Largest Contentful Paint element". It is often a big image. Web.dev's guide to optimising LCP says the time splits into four parts: time to first byte (how long the server takes to start replying, about 40% of the budget), resource load delay (under 10%), resource load duration (about 40%) and element render delay (under 10%). It calls these guidelines, not strict rules, and warns that fixing one part can push time into another.

  • Shrink the file. Resize the image to the largest size it is shown at, compress it, and use a modern format your platform supports.
  • Do not lazy-load it. Web.dev says not to lazy-load images likely to be in view when the page loads, especially the LCP image. Lazy loading is for images below the fold.
  • Make it easy to find. The image should be in the page's HTML so the browser finds it early. If it only appears through CSS or JavaScript, preload it.
  • Raise its priority. Add fetchpriority="high" to the one or two images likely to be the LCP element. Use fetchpriority="low" for offscreen ones.
  • Cut redirects and use caching. Each redirect adds waiting time. Avoid unique URL parameters that bypass the cache.

Builders vary. Many compress and resize images automatically, but a huge original uploaded as a background image can still slow the page. Check your builder's help for the image settings.

Fix 2: Third-party scripts and tracking tags

A third-party script is code from another company that runs on your page: analytics, ad pixels, chat widgets, heatmaps, booking widgets and consent banners. Web.dev's guidance says that if you have optimised your own code and the site is still slow, third-party scripts are probably the reason. The HTTP Archive Web Almanac 2024 found that 92% of pages use at least one third party, with a median of 27 third parties on the top 1,000,000 sites (66 on the top 1,000). Marketing sites collect them over time.

Google Tag Manager is a tool that loads many tags through one script. The container is not the problem, the tags inside it are. Every tag you add and forget stays on every page.

Run a tag inventory

List every script on the page. Open Chrome DevTools, go to the Network tab, reload, filter by JS, and note each domain. Then fill in this table.

TagWho needs itStill used? (yes or no)Action
AnalyticsMarketingYesKeep, load async
Ad pixel for a campaign that endedAdsNoRemove
Chat widgetSalesYesDelay until after load or after a click
Heatmap toolDesignRarelyRun for a month, then remove
Consent bannerLegalYesTest with it on and off

The actions, in order of impact: remove what nobody uses, use async or defer on the rest (web.dev says always use these for third-party scripts unless they are needed for the critical rendering path), and load non-essential tags after the page is interactive or after consent. The cost per tag varies by tag and site, so measure your own with the before and after test below.

For consent banners, see cookie banners and consent for marketing sites. Do not remove a consent tool to save speed if the law or your policy needs it. Load it efficiently instead.

The before and after test

  1. Run PageSpeed Insights on the page and note LCP, INP proxies (Total Blocking Time in lab) and CLS.
  2. In Chrome DevTools, use the Network tab's request blocking (open the command menu and search for "block request URL") to block one vendor's domain.
  3. Reload, measure again, and record the difference.
  4. Repeat for each vendor. The biggest gain is your first removal candidate.

Fix 3: Layout jumps (CLS)

Web.dev's CLS guidance gives these fixes:

  • Set width and height (or CSS aspect-ratio) on images and video so the space is reserved.
  • Reserve space for ads, embeds and late content with a minimum height.
  • For fonts, use font-display: optional, a fallback font with similar size, and preload critical fonts.
  • Animate with transform, not with top or left.

On marketing sites the usual culprits are a cookie banner that pushes content down, a chat widget, a sticky bar that appears late, and a hero with no reserved height.

Fix 4 and 5: Fonts, hosting and caching

Use fewer font families and weights, and choose a font-display value other than auto or block so text appears at once. For hosting, time to first byte is the largest slice of LCP, so a slow server or a missing cache hurts everything. A CDN and efficient caching help, and so does prerendering pages that do not change. If a page seems slow to start even when it is small, ask your host about caching. For hosting choices see website hosting for small business: speed and security and 10 hosting options.

If INP is the failing metric

INP is the time from a tap to the next update on screen: input delay, processing time and presentation delay. Web.dev says long tasks at load, such as heavy script evaluation, increase input delay, and that breaking up long tasks (yielding) helps. In practice, check which scripts run during load. The first suspects are the tags from Fix 2, then page-builder widgets such as sliders and popups. Remove or delay them, then retest.

A worked example (made-up numbers)

A plumber's marketing site shows these field results on mobile: LCP 4.2 s, INP 180 ms, CLS 0.22. INP passes, so the owner ignores it. The LCP element is a 3.4 MB photo, and the tag manager holds seven tags, including an ad pixel from a finished campaign and a heatmap tool left on since last year.

ActionResult in lab test (illustrative)
Resize and compress the hero photo to 180 KB, remove lazy loading from it, add fetchpriorityLCP drops from 4.2 s to 2.9 s in the lab
Remove two dead tags, delay the chat widgetLCP 2.4 s in the lab
Add width and height to the logo and gallery images, reserve a banner spaceCLS 0.04

The owner then waits for field data to catch up. CrUX uses a rolling 28-day window, so the real numbers move slowly. Lab results show the direction immediately, but only field data confirms it.

Is speed worth chasing?

A 2019 study by Deloitte for Google, covering 37 European and American brand sites and over 30 million mobile sessions, found that a 0.1 second improvement in four speed metrics lifted retail conversion by 8.4%, and lead-generation progression from the first form step to submission by 21.6%. That is old data from big brands, so do not expect the same on a small site. The sensible reading: fix the clear problems (a huge hero, dead tags), and do not spend a month squeezing the last 100 milliseconds.

Retest and keep it fast

  1. Same day: re-run PageSpeed Insights on the same pages, same device setting, and compare lab results.
  2. After 28 days: read the field data again. This is the pass or fail.
  3. Pass: LCP 2.5 s or less, INP 200 ms or less, CLS 0.1 or less at the 75th percentile, for mobile.
  4. Then keep it: ask that nobody adds a tag, widget or embed without checking the speed effect. Regressions usually come from a new tag.

If the site speed is fine but forms or the mobile layout look wrong, see the mobile fix checklist and the forms not sending checklist.

Copy-ready checklist

  1. PageSpeed Insights on your top 3 pages. Write down mobile field LCP, INP and CLS.
  2. Identify the LCP element.
  3. Resize, compress and prioritise that image. Never lazy-load it.
  4. List all third-party tags. Remove unused ones. Use async or defer on the rest.
  5. Set width and height on images and video. Reserve space for banners and embeds.
  6. Limit fonts and set font-display.
  7. Check caching, CDN and redirects.
  8. Retest in the lab the same day, and in the field after 28 days.

When to ask for help

Call your developer or host if time to first byte is the largest slice and the host offers no caching, if INP fails and you cannot tell which script causes it, or if a fix needs changes to the theme or build process. Send them the PageSpeed Insights report links, the tag inventory and your before and after numbers.

How I'd approach this

I would never start by installing a speed plugin. Spend the first hour on the field data and the LCP element, since that usually shows one clear fix. In the first week, shrink the hero, clean up the tags, reserve space for anything that pops in, and then leave the page alone for 28 days. If the three metrics pass for mobile, stop. If they do not, work down the table in Step 2. Skip the tuning if your visitors mostly come from a desktop on fast wifi and your numbers already pass. For most small business sites, the big wins are a smaller image and fewer tags.

Blur Studio builds custom sites and cares for them after launch, including speed checks when new tags are added. See what that includes if you want that handled for you.

Questions people ask

Why is my website slow?

On marketing sites the usual causes are a large hero image, many third-party scripts (analytics, ad pixels, chat, heatmaps), heavy fonts and slow hosting. Check your field data in PageSpeed Insights to see which Core Web Vitals metric fails, then fix the matching cause.

What are good Core Web Vitals scores?

Per Google's web.dev, at the 75th percentile of page loads: Largest Contentful Paint 2.5 seconds or less, Interaction to Next Paint 200 milliseconds or less, and Cumulative Layout Shift 0.1 or less. Check mobile and desktop separately.

What is the difference between lab data and field data?

Lab data is one controlled test run, such as a Lighthouse test. Field data comes from real Chrome users over the previous 28 days. Web.dev advises using field data to decide priorities and lab data to debug.

Do I need a Lighthouse score above 90?

No. A score is a debugging aid, not a target to chase. Aim to pass the three Core Web Vitals in field data. Lighthouse 10 does not even include INP in its score.

Does Google Tag Manager slow my site?

The container itself is small. The tags it loads are the cost, and they add up when nobody removes old ones. Audit the tags, remove unused ones, and use async or defer. For consent tools, see cookie banners and consent.

How long until my speed improvements show in PageSpeed Insights?

The lab result changes right away. The field data uses a 28-day window, so expect a few weeks before the real numbers reflect your fixes.

Will a faster website bring more leads?

It can help visitors stay and finish forms, but no speed fix guarantees leads or rankings. A 2019 Deloitte study for Google found gains for large brands, but a small site may see different results. Measure your own form submissions and calls before and after.

Is my slowness caused by my host?

Possibly, if time to first byte is high even with a small page. Ask about caching and a CDN. Our guide to hosting speed and security lists questions to ask.

More in this topic: Websites: fixing migration, DNS, forms, speed and CRM problems (6 guides) →

Sources

Ready for the next step?

A 30 minute call. If you do not need me, I will tell you.

Ranjit Bhinge You talk to me, Ranjit. I reply within a few hours.

Not sure what you need? Tell me what is going on. I will point you the right way, even if it is not me.

Blur Studio · 30 minute call

Book a 30 minute call

You talk to me, Ranjit. If you do not need me, I will tell you.

  • We talk about your business and what is going on.
  • You leave with a clear next step, even if it is not me.
  • I reply to every message within a few hours.
Loading the calendar…