Testing tools simulate a visit under chosen conditions, which is useful for diagnosis and is not what your visitors experienced. Field data from real visits is the measurement that matters, and both are worth reading for different reasons.

Two kinds of measurement

They answer different questions and are constantly confused.

A simulated test loads your page under conditions the tool chose, from a location the tool chose, on a connection the tool simulated. It produces a repeatable result useful for diagnosing what is slow.

Field data is collected from actual visits by actual people on their own devices and connections. It is what your visitors experienced.

A page can score well in a simulation and perform poorly in the field, usually because the simulation ran on a fast connection from a nearby location on a device better than most of your visitors carry.

Why the score is the least useful part

Tools produce a headline number because people want one, and it is a weighted summary of several measurements.

Two problems follow. A single number cannot tell you what to fix, and chasing it produces work on whatever the weighting emphasises rather than on what visitors notice.

The measurements underneath are the useful part: how long until something readable appeared, how long until the page stopped moving, and how long until it responded to a tap.

Those correspond to things a visitor actually experiences, which the score does not.

Testing conditions that resemble reality

The setting that changes the result most.

Test on a mobile profile with a throttled connection, from a location near your customers, with the cache cleared. That is closer to a first-time visitor on a phone outside than any default desktop test.

Most tools default to something more flattering, which is why businesses are surprised when a good score does not match how the site feels.

Run it more than once, since a single test can be affected by transient conditions, and take a median rather than the best result.

Reading the waterfall

The most informative view and the one nobody opens.

It shows every request the page made, in order, with how long each took and when each started. That is where the actual answer sits.

What to look for: the largest items, anything from a domain that is not yours, anything that starts late because it was waiting on something else, and long gaps where nothing is happening.

A single oversized image, or a third-party script blocking everything behind it, is visible immediately in a waterfall and invisible in a score.

Field data, and its limits

Where real-visit measurements come from and what they can and cannot tell you.

Search console and some measurement tools report data collected from real visits, aggregated over a period. That is the honest picture of what people experienced.

The limits: it lags by weeks, it needs enough traffic to report at all, and a small site may have insufficient data for individual pages. Many small business sites see only site-level figures or nothing.

Where field data is unavailable, a realistic simulated test is the best proxy, which is why the testing conditions matter so much.

The measurement nobody automates

Worth doing regardless of tools.

Open your site on your own phone, on mobile data, away from the office, with the cache cleared, and count. How long until you could read something. How long until you could tap the phone number without the page moving.

That takes a minute and it is closer to your customer's experience than any dashboard. It also catches things tools do not report, such as a layout that is technically fast and practically unusable.

What to do with the results

  1. Record the numbers before changing anything, so improvement is demonstrable.
  2. Fix the largest item in the waterfall first, which is usually an image.
  3. Retest under the same conditions, or the comparison means nothing.
  4. Ignore small movements, since test results vary between runs.
  5. Check the field data months later, since it lags.

The third is where most speed work goes wrong. A before figure from a desktop test and an after figure from a mobile test is not a measurement of anything.

When to stop

A practical limit worth setting.

Once a page shows readable content quickly on a phone on mobile data and does not shift while loading, further optimisation returns very little for a small business site.

Pursuing a perfect score past that point is work with no customer-visible benefit, and the effort is better spent on what the pages actually say, which is the argument set out in what slowness actually costs you.

Frequently asked questions

What is the difference between test types?

A simulated test loads your page under conditions the tool chose. Field data comes from actual visits by real people, and is what your visitors experienced.

Why is the score the least useful part?

It is a weighted summary that cannot tell you what to fix, and chasing it produces work on whatever the weighting emphasises rather than on what visitors notice.

How should I set up a test?

Mobile profile, throttled connection, a location near your customers, cache cleared. Run it several times and take a median rather than the best result.

What should I read instead of the score?

The waterfall. It shows every request in order with timings, which is where a single oversized image or a blocking third-party script becomes immediately visible.

What are the limits of field data?

It lags by weeks, needs enough traffic to report, and many small sites see only site-level figures or nothing at all.

When should I stop optimising?

Once a page shows readable content quickly on mobile data and does not shift while loading. Beyond that the effort returns very little for a small business site.

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.

Chasing a score that does not match how the site feels?

We measure under conditions your customers actually experience and read the waterfall, which tells you what to fix rather than how you scored.

Start a Conversation