Skip to main content
Swift Page
Swift Page

Stage 03 of the work

Re-testing and the report

The same pages measured again under the same conditions, set out beside the first reading, so the difference is a comparison rather than an assertion.

scope: the pages audited · needs: the work published · gives: both readings, side by side

You end up holding

A report with both readings side by side, the conditions printed against each of them, and the request waterfall from each run.

How it runs

In this order, every time. A stage that is done in a different order on different jobs cannot be compared with itself afterwards.

What is needed from you

  • The changes to be live, or a staging address you are happy to be measured instead.
  • A note of anything else that changed on the site in the meantime, because it belongs in the report.
  1. 01

    Repeat the conditions exactly

    The same pages, the same connection profile, the same device profile, the same tool and the same number of runs as the first reading. A comparison against different conditions is not a comparison.

  2. 02

    Measure again

    Several runs, the middle one reported, and the waterfall kept from each so the second report can be checked rather than believed.

  3. 03

    Set the two readings side by side

    Each figure with its conditions against it, and each change in the waterfall matched to the item from the list that caused it.

  4. 04

    Say what did not move

    Items that were carried out and changed nothing measurable are reported as such. A report that only shows what improved is an advertisement.

  5. 05

    Say what is left

    Anything from the audit that was not carried out, and anything the platform or the hosting puts out of reach, is listed again at the end so you know where the remaining time is going.

The comparison

Two readings, one set of conditions

A second reading only means something if it repeats the first one exactly: the same pages, the same connection profile, the same device profile, the same tool, the same number of runs. Change any of those and the difference is between two test setups, not between two versions of your site.

0 1 2 3 4 5 Before 4.0s
The main picture finishes arriving, and the page finally looks like itself

The same event, twice

Both arcs are reading the moment the largest thing on the first screen finishes arriving, on the same scale, under the same conditions. That is the only comparison a report of this kind is entitled to draw.

0 1 2 3 4 5 After 1.4s
The same picture, the same test, once the work in the waterfall below was done
the page as it arrives now the same request after the work
  • The page itself 48 KB → 62 KB

    Left alone, and a little larger: the styles the first screen needs are now inside it

  • styles.css 186 KB → 94 KB

    Rules no page uses removed; the rest stops holding up the first paint

  • slider.js 142 KB → nothing

    Removed. The one slider on the site was replaced with plain markup

  • Six web font files 540 KB → 60 KB · now two web font files

    Four weights nothing used dropped, the two that are left asked for early

  • hero.jpg 1.4 MB → 180 KB · now hero.webp

    Sized for the space it occupies, saved as WebP, asked for before anything else

  • Six product photographs 1.2 MB → 300 KB

    Resized, and no longer asked for until the reader scrolls down to them: in this run that put them after the first screen had finished

  • analytics.js 96 KB → 96 KB

    Same file, moved behind the page instead of in front of it

  • Chat widget 320 KB → nothing

    Loads when somebody opens it, which on this page nobody had yet

seconds from opening the address
0s 1s 2s 3s 4s 5s
A worked example of the comparison table. The weight and the font count are worked out from the waterfall above rather than typed in.
Reading Before After
Largest Contentful Paint When the biggest thing on the first screen has finished arriving 4.0 s 1.4 s
Total Blocking Time How long the page ignored taps and scrolls while it was busy 610 ms 120 ms
Cumulative Layout Shift How far the page jumped about under the reader as it filled in 0.21 0.01
Fetched before the first screen was ready Everything the browser had already asked for by the moment above 3.8 MB 0.4 MB
Font files fetched How many separate font files the page waited on 6 2

Measured under these conditions

page
The home page of a five-page brochure site
connection
Throttled to a 4G profile: 1.6 Mbps down, 150 ms round trip
device
Mobile profile, processor slowed four times to stand in for a mid-range phone
tool
Lighthouse in Chrome, cache empty, no browser extensions
runs
Three runs each side, the middle one reported

Everything on this screen is a worked example of the report format, filled in with example figures so the layout can be read. It is not a result offered to your site: what your pages read before, and what they read afterwards, is not known to anybody until it has been measured.

What goes in besides the good news

A report you cannot argue with is not worth having. These four go in every time, including the ones that make the work look less impressive than it was.

Still not promised

  • A score. No number out of a hundred is offered as a target, and no work here is aimed at one: a score is a summary of measurements, and the measurements are what is worked on.
  • A position in search results. Loading speed is one signal among many that nobody outside a search engine controls, and anyone who tells you otherwise is guessing.
  • More sales, more enquiries or more sign-ups. A faster page removes one reason to leave. What a visitor does once they have stayed is down to the offer, the price and the writing.
Items that changed nothing
Carried out, measured, and no movement worth reporting. They stay in the report with that written against them. A report showing only what improved is an advertisement with a table in it.
Items that were not done
Anything from the audit you chose not to have done, or that a platform put out of reach, is listed again at the end with what it is still costing.
Anything else that moved
If the site changed in other ways between the two readings - new content, a new plugin, a different hosting plan - it goes in the report, because it is part of the difference.
What the reading does not cover
One page on one connection profile on one device profile. Other pages, other connections and other tools will read differently, and the report says which pages it stands for.

What this stage is not

  • No target figure is agreed beforehand, because the honest answer is not known beforehand.
  • Measurements taken by other tools, on other connections or on other devices will not match, and are not what this report is compared against.
  • Ongoing monitoring is not part of it. If you want the site watched over time, that is separate work.

The other two stages

What a stage costs, how long it takes and how many rounds of changes it includes depend on the site, and they are set out in a written quotation in pounds sterling before anything starts. Nothing is owed for asking.

Request a website speed review