Faragiri

Learn What speed is worth Speed

A Devon cheesemonger passed its speed assessment and moved half a place in search, while the same fix added 31 completed orders a month.

4Sections
4Min read
4Depth

Speed is a tiebreaker between comparable pages, not a way past a better one. Cullompton Cheese sells cheese by post from Devon, fixed the images and the layout shift on its shop template in March 2026, and passed the assessment in April. The site passed and moved from 6.4 to 5.9 on its main query.

Half a place. In the same two months the same fix added 31 completed orders a month at the checkout, where a payment form had been moving under people's thumbs while they typed. One of those numbers is worth mentioning in a meeting and it is not the ranking.

Why the search effect is small

A page that answers badly and loads instantly still loses. Speed is one input among many and it acts almost entirely between pages an engine already considers similar; where one page answers the question and another does not, no amount of speed reverses the order, and the pages Cullompton competes with for its main query are three national retailers with far more links.

Ranking weight

When page speed moves a ranking, and when it cannot

Speed acts as a tiebreaker between comparable pages; it never lifts a weak answer past stronger competitors.

Is the page already equal to rivals on answer quality and links?
  • No (weaker match or links)
    Speed changes nothingNational retailers with more links stay ahead regardless of load time.
    • Improve the page answer first
    • Earn real links before tuning milliseconds
  • Yes (closely matched pages)
    Speed breaks the tieMovement stays small, typically around half a position on contested terms.
    • Small edge between similar pages
    • Zero shift where you already lead clearly

A fast page that nobody links to does not rank; speed work is easier than earning links, which is why it gets oversold.

The effect is largest where two pages are otherwise equal, which is a narrow band and it is where the half place came from. On a query where the shop was already clearly the best answer, nothing moved at all, because there was nothing to break a tie about.

A fast page that nobody links to does not rank, and that sentence is worth keeping next to any speed proposal. The work is real and its size is real, and a firm doing speed work instead of earning links without buying them has chosen the easier of two jobs.

Where the value actually was

The commercial effect is usually larger than the search effect and easier to prove. Cullompton's layout shift was on the payment step: the card fields moved down by about 90 pixels when a delivery estimate loaded, and people were typing into a field that was no longer where they had put their cursor.

What it bought

The same fix, weighed two ways

One pan holds a soft half place measured against a noisy background, and the other holds a count of orders at a shop where nothing else changed.

  • What it did in searchHalf a place on the main query, and nothing measurable on the rest.
    • 6.4 to 5.9 on the main query
    • No movement where the shop was already the best answer
    • Assessment passed in April
  • What it did at the checkoutThe card fields stopped moving under people's thumbs.
    • 31 more completed orders a month
    • Same weeks, nothing else changed
    • A 90 pixel shift on the payment step

One template fix, March 2026

The tilt is drawn because the numbers tilt it, not because speed work is being argued for or against. At a larger shop the crawl effect would put a third weight on the left pan.

The 31 orders are measured by comparing completed against started checkouts either side of a dated change, at a shop where nothing else changed that month, which is as close to clean as a small business ever gets. The half place in search is measured against a background of everything else that moves a position, and it is the softer of the two numbers despite being the one the work was commissioned for.

Where this stalls: selling it as a ranking gain

Speed work is worth selling on the checkout, not on the ranking. Proposals for it are almost always written as search work, because that is the department it lands in, and the result is a firm that spends four thousand pounds expecting a ranking change and gets half a place.

Where proposals stall

Selling checkout fixes as ranking work kills the second phase

When speed work is pitched solely for search rankings, a real commercial gain looks like failure on the wrong scorecard.

One line in the project brief measuring orders as well as rankings changes what the entire piece of work is judged on.

The stall is the disappointment that follows, which is what stops the second piece of work being funded. Cullompton's owner had been told in 2024 that speed was a ranking factor, which is true and was sold at the wrong price, and the 31 orders were discovered by accident afterwards.

Measure the orders as well as the position, because only one of them will move. That is one line in a brief and it changes what the work is judged on.

The other thing it buys

There is a second effect on a large site and it is not about ranking either. A slow site is fetched less often, so a catalogue with thousands of pages can find that speed is limiting how much of it is stored at all, which is what a slow site costs you in pages indexed. At a cheesemonger with 90 products it does not arise.

What to measure

Three places a speed fix can show up, and what reads each one

Search position, checkout completion, and large-catalogue crawling each require a different check rather than one catch-all score.

Post-fix evidence
  • Pages indexedMatters on large catalogues; does not apply to a 90-product shop.
  • Search Console positionFree since replacing 2011 keyword reports, but noisy on one query.
  • GA4 checkout rateReplaced Universal Analytics in 2023; compares completed to started orders.

Compare before and after a single dated template change

Only one of the three typically moves cleanly on a small site, and it is almost always the checkout comparison.

Search Console gives the position either side of a dated change, free, and it replaced the keyword data Google removed from analytics in 2011. Reading it for one query across two months is the whole of the search measurement available and it is a soft one.

Google Analytics 4 gives the checkout number, free, and it replaced Universal Analytics, which stopped processing data on 1 July 2023. The comparison that matters is completed against started, in the same weeks, at a shop where nothing else moved.

Which of the three measurements the fix was aimed at is the three measurements, and the thresholds Google uses, and telling this month's search effect apart from everything else happening in a month is knowing whether any of it worked. Both belong to how fast it loads, and what that is worth in search, and half a place is the honest answer to the second half of that sentence.

Alongside this

Above this