Check the hostname report for test addresses and old copies of the site. Add a view filter that includes only your real domain, keep one unfiltered view, and remove the tracking code from test copies.

Visits rising in a quiet season

During a redesign, the owner opens Google Analytics to see visits rising week after week, in a season that is usually quiet. Enquiries have not changed at all.

The new site is not live yet. It is being built at a test address, and the extra visits come from there: the designer, the owner and anyone else reviewing pages.

Every page they open is recorded as a visit to the real site, sometimes for months, so the figures used to judge the new design start out wrong.

How a test copy reports to the real account

Designers usually start from a copy of the live site at a test address: a subdomain, a folder on their own server or the host's temporary address.

The copy carries the same tracking code, with the same account number. That code does not check where it is running; it reports every page view to the account it names.

So a designer checking a menu at the test address is counted exactly like a customer reading a services page on the real site.

Who visits a test copy

The visits to a copy come from people who are not customers:

Each page they load looks like a customer's visit in every report, from the visit count to the time spent on the site.

Old copies that keep counting

Test copies are often left in place after launch. A copy from an earlier redesign, or one left on a former designer's server, can keep sending visits for years.

Some are found by search engines, so real people land on an outdated version with old prices and a form that emails nobody, and their visits count too.

The hostname report shows the source

Analytics records the address each page was served from, called the hostname. Open Audience, then Technology, then Network, and choose Hostname at the top of the table.

The list should hold the real domain, with and without www, and little else. A test subdomain, a temporary address or an unfamiliar domain is a copy carrying your code.

Set the date range back a year. Old copies appear as small lines running month after month, and the test copy as a burst that matches the redesign.

A filter that accepts only the real domain

In the Admin settings, a filter on the view can include only hits whose hostname matches your domain. Hits from any other address are then left out of that view.

Write the pattern to cover the domain with and without www, and check the next day that visits are still arriving. A mistyped pattern can exclude everything.

Keep one view with no filters, as the raw record. A filter works only from the day it is saved, and what it leaves out is gone for good, so the untouched view is your only check on it.

A worked example

A bicycle tour company in the Okanagan runs guided winery rides and multi-day trips along old railway grades. Bookings for the summer start in the spring.

Over the winter it had its site redesigned. The designer built the new version at a test subdomain, copied from the live site, tracking code and all.

Visits climbed from early January while enquiries stayed at their usual winter trickle, and the owner assumed people were planning trips earlier this year.

The hostname report showed the test subdomain behind much of the winter's rise, and a second, smaller line: the previous site, still online at a former designer's address.

The code came off the test copy, which now sits behind a password. The former designer deleted the old copy, and the main view now includes only the real domain.

An unfiltered view keeps the raw record, and an annotation marks the day the filter began, so the spring figures will describe people looking at tours.

Fixing it at the source

The filter cleans the reports, but the copies still need dealing with. Ask the designer to leave tracking code out of test copies, or to give them a separate account.

Put every test copy behind a password, so search engines cannot find it and strangers cannot use its forms, and delete it once the new site is live.

On launch day, check that the live site carries the right tracking code. A launch with test settings, or with no code at all, leaves a gap no filter can fill.

Reading the months already recorded

The inflated months stay in the account. A filter cannot change them, but a segment can, because segments apply to past data as well.

Build a segment that includes only the real hostname, and use it whenever you compare this winter with last. A raw month beside a filtered one makes any change look bigger or smaller than it was.

Note when the redesign started and ended beside any report you share, so nobody reads the winter as a change in demand.

The counter-case

Not every redesign touches the live account. A designer who builds on a local computer, or who strips the tracking code from every copy by habit, leaves nothing to clean up.

On a busy site a few weeks of designer visits may be too small to matter, and an owner who never decides anything from the visit figures may not notice.

A hostname check is still worth the few minutes it takes, since it often turns up a forgotten copy. The filter and the raw view cost little and protect every report that follows.

Keeping the count clean through a redesign

  1. Open the hostname report for the past year.
  2. List every address that is not your domain.
  3. Keep an unfiltered view as the raw record.
  4. Add an include filter for your domain.
  5. Annotate the date the filter started.
  6. Remove the code from copies and lock them with a password.
  7. Check the tracking code on launch day.

Step three comes before the filter for a reason. A filter changes what is recorded from that moment on, and without an untouched view a mistaken pattern loses data for good.

With the count describing real visitors again, a goal funnel can show where they abandon a long form, as the registration form where parents give up explains.

Frequently asked questions

Why does a test copy of my site count as visits?

It carries the same tracking code as the live site. The code does not check where it runs, so every page viewed on the copy is reported to your account.

Where do I see which addresses sent visits?

In the hostname report: open Audience, Technology, Network, and choose Hostname. Anything other than your domain, with or without www, is a copy carrying your code.

What filter keeps the test visits out?

An include filter on the view that accepts only hits whose hostname matches your real domain. Check the next day that visits are still arriving.

Does the filter fix the past months?

No. Filters work only from the day they are saved. To look back, use a segment that includes only your real hostname, since segments apply to past data.

Why keep an unfiltered view?

What a filter leaves out cannot be recovered later. An untouched view keeps the raw record, so a mistaken pattern never costs you data you cannot get back.

What should happen to old copies of the site?

Remove them, or ask whoever controls them to. Copies left online keep sending visits, and search engines can send real customers to outdated prices and forms.

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.

Visits rising while the new site is built?

Give us read access to your Analytics account, and we will review the hostnames and set up the filter and a backup view.

Start a Conversation