Analytics measures what a browser will report to it. Phone calls, blocked scripts, cross-device journeys, and referrals that arrive without a source are all invisible or misattributed, which means the picture is systematically incomplete in predictable ways. Knowing which parts are missing is what stops the numbers being over-read.

The direct traffic bucket

Direct is presented as people who typed your address or used a bookmark. In practice it is where anything without an identifiable source ends up.

That includes clicks from documents and desktop applications, links opened from messaging apps, some links from email, traffic where a security setting stripped the referring information, and increasingly, visits arriving from AI-generated answers that pass no source along.

So a large direct figure is not necessarily brand strength. It is partly a category of last resort. Treating it as people who knew your name and came looking is one of the more common misreadings, and the AI portion of it is examined further in tracking AI referral traffic.

Blocked and limited measurement

A meaningful share of visitors block analytics scripts through browser settings, extensions, or network-level filtering. Those visits are not undercounted, they are absent.

Two consequences. Your true traffic is higher than reported, which is mildly encouraging and not very actionable. And the missing visitors are not a random sample, since people who block scripts skew toward certain devices and demographics, so the composition of your reported audience is slightly off in ways you cannot correct.

The practical response is to treat analytics as an index rather than a census. Changes over time remain meaningful because the bias is reasonably consistent. Absolute figures should not be quoted with confidence.

The phone call

The largest gap for most service businesses.

Analytics can record that someone tapped a telephone link. It cannot know whether the call connected, how long it lasted, whether it was a real enquiry, or whether it became work. A tap is a signal of intent and it is not a conversion.

Someone who reads your site on a phone, writes the number down, and calls from a landline an hour later is invisible entirely. So is anyone who saw your number on a vehicle after visiting the site.

This is why the four-second question on the phone, asking how they found you, is worth more than most analytics configuration, as covered in who answers your phone.

Cross-device journeys

A realistic path: someone finds you on a phone during the day, looks again on a laptop that evening, and calls the next morning.

Analytics generally sees two unrelated visitors and one of them converting, or neither. The research visit gets credited to nothing and the phone call is invisible. For considered purchases, which most trades work is, this pattern is common rather than exceptional.

Small numbers

A local business site may receive a few hundred visits a month. At that scale, ordinary variation looks like a trend.

A conversion rate calculated on eight enquiries moves substantially if next month brings ten. That is noise, and it will regularly be presented as a forty percent improvement. Reporting tools also suppress figures below certain thresholds to protect privacy, so some rows simply do not appear.

The discipline is to compare longer periods, quarter against quarter rather than week against week, and to distrust any percentage calculated on a small base.

What to do about it

Why this is worth knowing

Not to discourage measurement, which remains far better than guessing. It is to prevent two specific errors: cancelling something that was working because analytics could not see its effect, and expanding something that appears to work because it happens to be measurable.

Channels differ enormously in how visible they are. Paid advertising reports itself precisely. Word of mouth, vehicle signage, and referral relationships report nothing at all. A business that funds only what it can measure will systematically underfund the things that produce work quietly, which for most local service businesses is a substantial share of it.

Frequently asked questions

What does direct traffic actually mean?

Anything without an identifiable source, not only people who typed your address. It includes clicks from documents, messaging apps, some email, traffic where referring information was stripped, and visits from AI answers.

Are my analytics numbers accurate?

They are an index rather than a census. A share of visitors block measurement entirely, so real traffic is higher than reported. Trends over time remain meaningful because the bias is reasonably consistent.

Why can I not see phone calls in analytics?

Analytics can record a tap on a telephone link but not whether the call connected, how long it lasted, or whether it became work. Anyone who noted the number and called later is invisible entirely.

How do I track customers who use several devices?

Largely you cannot, without a login. A visit on a phone and a follow-up on a laptop usually appear as two unrelated visitors, which is common for considered purchases rather than exceptional.

Why do my percentages swing so much month to month?

Small numbers. A conversion rate based on eight enquiries moves substantially when next month brings ten, and that is noise. Compare quarters rather than weeks and distrust percentages on a small base.

What is the most reliable measurement for a small business?

Counted enquiries from your inbox and call log, combined with asking callers how they found you. Both exist outside analytics and cover the offline half that scripts cannot see.

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.

Making decisions on numbers that only tell half the story?

We combine analytics with your call log and enquiry records, so the channels that work quietly do not get cut because they were harder to measure.

Start a Conversation