Skip to main content
Swift Page
Swift Page

Stage 01 of the work

Speed audit

Your site measured as it stands, under conditions written down, with everything that is costing time listed and ranked by what removing it would save.

scope: one site · needs: the address · gives: a written report · changes nothing

You end up holding

A written report: the measurements, the conditions they were taken under, and a ranked list of what is slowing the page down with an estimate of what each item is worth.

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 address of the page or pages you want measured.
  • The platform the site is built on, and the hosting if you know it.
  • Anything you have already been told about the site being slow, and by whom.
  1. 01

    Agree what is measured

    Which pages, which connection profile and which device profile. A home page on a desktop connection and a product page on a throttled mobile one are different measurements, and a report that mixes them proves nothing.

  2. 02

    Measure the site as it is

    The agreed pages, several runs each, cache empty, everything recorded: the timings, the request waterfall, the size of what is fetched and the order it arrives in.

  3. 03

    Find what the time is being spent on

    Pictures larger than the space they fill, font files nothing is set in, scripts that hold up the first paint, third-party embeds, work the server is doing before it answers at all.

  4. 04

    Rank it by what it would save

    Every item gets an estimate of the time it is costing and a note on how disruptive removing it would be, so you can choose rather than be told.

  5. 05

    Hand the report over

    It is yours. You may act on it yourself, give it to whoever maintains the site, or ask for a quotation for the second stage. Nothing about the audit obliges you to the rest.

Five questions a reading has to answer

A number that cannot be repeated is not a measurement

The same site, measured twice in the same hour, will give two different answers if the page, the connection, the device or the tool changed in between. So the conditions are settled before anything is measured and printed against every figure afterwards.

The conditions used in the worked example on the home page

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
Which page is measured?
Whichever pages you choose, named in the report. A home page and a checkout are different pages with different problems, and a single figure covering both hides whichever one is worse.
On what connection?
A throttled profile that stands in for a mobile connection, held the same for every run. A measurement taken on office broadband flatters a site that most of its readers are waiting on.
On what device?
A slowed processor profile standing in for a mid-range phone. Scripts that feel instant on a laptop are the ones that make a phone stop responding.
With what, and how many times?
A named tool, cache empty, several runs, the middle one reported. One run of anything is an anecdote: browsers, networks and third parties all vary between runs.
Compared against what?
Against the same site later, under the same conditions. Not against a competitor measured some other way, and not against a number somebody else published.

What the audit looks for

Where the time actually goes

The report walks the same five headings every time, so two audits of the same site a year apart can be laid beside each other. Under each heading, what was found on your site, what it is costing, and what removing it would take.

Images

What each picture weighs, what size it is actually displayed at, what format it is in, and whether the page has reserved room for it.

reported as: what was found · what it costs · what changing it would take

Fonts

How many files are fetched, which weights and styles anything on the site is actually set in, where they are fetched from, and what the text does while they are on their way.

reported as: what was found · what it costs · what changing it would take

Scripts

What each script costs to fetch and to run, what holds up the first paint, what runs on a page that never uses it, and what a third party is loading on your behalf.

reported as: what was found · what it costs · what changing it would take

Loading order

What the browser asks for first, what it discovers late because something else had to be read first, and what is competing for the connection while the reader is looking at nothing.

reported as: what was found · what it costs · what changing it would take

Caching

What your server says about how long a file may be kept, whether a returning reader fetches the same things again, and how long the server takes to answer before any of it starts.

reported as: what was found · what it costs · what changing it would take

And the answer before any of it

How long your server takes to answer at all. If a page spends a second being built before a single byte is sent, nothing done in the browser will make up for it, and the report says so instead of tidying round the edges.

hosting is a decision, not a setting · the report names it either way

What this stage is not

  • No change is made to your site. An audit is a reading, not a repair.
  • No access to your site is needed, beyond the public address.
  • It is not a security review, an accessibility audit or a design critique, although anything glaring that turns up is mentioned.

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