Set up the replacement now so it has a full year of data by July. Waiting until spring means no year-on-year comparison until 2024.

Two dates, and the second one matters more

The stated deadline is the first of July next year, when the old analytics stops processing new data for standard properties.

The practical deadline is much earlier, and it is the one nobody acts on.

The replacement does not inherit your history, so it only knows what it has collected itself.

Which means a property started in June has no comparison data in September, and a property started this month does.

Why the parallel year matters

Since this is the whole argument for acting in December rather than in spring.

Almost every useful question in analytics is a comparison: is this month better than last year, did the season arrive early, is that page doing more than it was.

Answering any of those requires the same tool to have been collecting during both periods.

Start now and you have a full twelve months by next December, covering a whole seasonal cycle.

Start in June and your first genuine year-on-year comparison arrives in mid-2024.

The difference between those two positions is an hour of configuration, done now.

What to do this month

The third is worth stating explicitly. Both should run together until July, since the old one still holds the comparison you need between now and then and turning it off early gains nothing.

Check it is actually collecting

Because a property created and not verified is the common failure.

A great many businesses set one up when the announcement came in March, saw no errors, and assumed it was working.

Open it now and look at whether it has data for the last week, which takes a minute.

Where it shows nothing, the tracking was never installed correctly and nine months of data has not been collected.

That is recoverable if found today and considerably less so if found in June, which is the reason to look this month.

Check on both a phone and a desktop, since a tag installed in one template and not another is a common half-failure.

A worked example

A business created the new property in April after being told to, and did nothing further.

Checking in December, it had recorded almost nothing, because the tracking had been added to the site's header in a way that a later theme update had removed.

They reinstalled it properly, verified it on a phone and a desktop, and set up two conversions.

Eight months of data had been lost, which they could not recover, and the twelve months before July were still available.

Had they checked in June instead, they would have had a few weeks of data and no comparison at all for the following year.

The check took four minutes.

Export what you want to keep

Since the old data does not move across and will not remain available indefinitely.

Decide what you would actually want in two years, which is less than everything.

Monthly traffic and enquiries by year, your main landing pages and their figures, and the sources traffic came from is usually enough.

Export those as a spreadsheet, once, rather than attempting to preserve the whole account.

A single page of annual figures is more use than an archive nobody can open, and it takes half an hour.

Do it before July, and there is no reason not to do it in January while it is quiet.

The numbers will not match

Worth expecting now so it is not alarming later.

The replacement counts differently, so its figures will not agree with the old ones for the same period.

Sessions, users, and bounce rate in particular are defined differently, and comparing across the two tools produces confusion rather than insight.

Which is another reason to run both: you can see the gap for your own site and know what to expect.

After July, compare within the new tool only, and treat the historical figures as a separate record rather than a continuous series.

Do not over-configure it

Since the common failure at the other extreme is a setup nobody understands.

Two or three conversions is enough: an enquiry, a call, and a sale where relevant.

Everything else can wait, and a property with forty configured events is one nobody reads.

Get it collecting, get the few things that matter marked, and leave it alone until you need it.

The aim in December is that the data exists next year, not that the reporting is perfect.

Whoever set it up may not be around

A practical point that decides whether any of this is possible in June.

Analytics accounts are frequently owned by a developer or an agency, created under their own login, with the business having no access at all.

That is fine until the relationship ends or somebody stops answering, at which point the history and the new property both belong to somebody else.

Check now whose account it sits in, and make sure somebody at the business has full administrative access rather than a view-only invitation.

Do the same for the search console, since the two are usually set up together and lost together.

The counter-case

Some businesses barely use this at all.

A trade whose work all comes from the phone and from referrals may look at analytics twice a year, and for them the deadline is a small administrative matter.

There are also alternatives to the standard tool, some simpler, and a business unhappy with the replacement is free to move rather than migrate.

And the date may move, since these dates sometimes do.

Create it, verify it is collecting, run both until July, and export the figures you would miss.

This month

  1. Open the new property.
  2. Confirm it has data for last week.
  3. Check on a phone and a desktop.
  4. Fix the tracking if it is missing.
  5. Mark two or three conversions.
  6. Add a second person to the account.
  7. Export the old figures in January.

Step two takes a minute and is where the unpleasant discoveries happen, since a property created in spring and never verified has frequently collected nothing at all.

What was announced in March is covered in Google set a date to switch off Universal Analytics.


Frequently asked questions

What is the deadline?

The first of July next year, when the old analytics stops processing new data for standard properties. The practical deadline is much earlier.

Why act in December?

Because the replacement does not inherit history, so it only knows what it has collected. Start now and you have a full year by next December; start in June and your first comparison is mid-2024.

Should I turn the old one off?

No. Run both until July, since the old one still holds the comparison you need between now and then and turning it off early gains nothing.

What is the common failure?

A property created after the March announcement, never verified, collecting nothing because the tracking was not installed correctly or was removed by an update.

How do I check?

Open it and look for data from the last week, on both a phone and a desktop. That takes a minute and is where the unpleasant discoveries happen.

What should I export?

Monthly traffic and enquiries by year, your main landing pages, and traffic sources. A single page of annual figures beats an archive nobody can open.

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.

Set up the new analytics in the spring?

Open it and check it has data from last week. A property created and never verified has frequently collected nothing.

Start a Conversation