Key takeaways:
- Universal Analytics stopped measuring in summer 2023. From 1 July 2024, all data will be deleted.
- Even if you already have year-on-year comparison data in the new Google Analytics 4, we recommend backing up older data.
- Back up traffic data at least on a monthly basis.
- Evaluate other data and save at least overviews and reports.
- For smaller websites, exporting the most important data is sufficient; larger projects will benefit from help from data specialists.
Why back up old data
While some companies were slow to transition to the new Google Analytics 4, others have been using it for years. If you're in the latter camp, you already have year-on-year comparison data. Even so, we recommend backing up at least the Universal Analytics data as CSV. Older data can still be useful for evaluating trends, for historical business context, and for strategic decision-making and future planning.
„There are plenty of companies that left the transition to Google Analytics 4 to the last minute and only started using it from July 2023. For them especially, backing up previous data is important. But other companies may also find it useful to look a few years back — for example, to track how a trend has developed or to compare with the situation before COVID. Historical data offers many important business insights, and it would be a mistake to lose it."
We should note that comparing data from the old and new Google Analytics is not straightforward. They use different measurement methodologies — for example, a website visit is counted differently. You can't simply take older data and compare it directly. But you can keep it stored in CSV format, and whenever you need to make a comparison, you can interpret and compare the data with the current state.
What Universal Analytics data to back up
While we consider data important, we also notice the bad habit of measuring everything. Original Analytics accounts can therefore be full of data you don't actually need. When backing up, decide which data is important to you and which you'll actually use.
If you're not sure what to back up, we at least recommend data on traffic and your website's primary goal (transactions or leads) broken down by source and optionally by campaign. Back up other data based on what you measure, use, and compare. For e-commerce websites, this might be the performance of individual products or transactions including IDs. For lead generation websites, it could be the performance of individual landing pages, other goals, and events. Data on traffic by individual pages or by the device used to visit the website may also come in handy.
Back up data at a minimum on a monthly basis — that's often the only viable approach. Having data aggregated by year is practically useless. Daily data will be more useful, but it's often also far too large.
„When exporting daily traffic data with an average of 50 different sources over 10 years, you end up with 182,000 rows. For individual products it'll be even more. So monthly data can be the quickest option. For everything else, you can evaluate it and save just the report instead of the raw data — and that should serve you well going forward."
If you do an annual data review for your website, that may be sufficient as a backup. Adding at least the basic minimum of data not stored elsewhere (for example, in your CRM) will make it even better.
Export options
Exporting data from Universal Analytics
Directly from the Universal Analytics interface, you can save data in CSV, PDF, XLSX, or Google Sheets format. You can then evaluate, interpret, or visualize the data, for example in Looker Studio.
This method is straightforward and can be done by anyone without technical skills. Log in to your Universal Analytics account, view the data you want to export and the time period you want to save, then download it in the required format.
However, you'll encounter limitations — only 5,000 rows can be exported at a time, so if you have more data you'll need to save it in batches. The entire process can therefore take considerably longer. There's also the issue of sampling, where the system calculates values based on only a portion of the data, introducing distortions. The only way to avoid this is to download data in smaller chunks.
We therefore recommend data export mainly for smaller websites with a smaller amount of data.
Backup via BigQuery
If you have a larger website and use the paid Google Analytics 360, you may already have a connection to the BigQuery cloud storage, where your data is automatically backed up. Otherwise, you'll need to upload it manually.
By the way, also set up BigQuery for the new Google Analytics 4, where you'll have nearly raw data that you can freely model and calculate.
Backup via Google Analytics Spreadsheet Add-on
Slightly more complex than a direct export is the solution using an add-on from Google itself, which connects Universal Analytics with Google Sheets. A template is available for the backup, so you just create a table in Google Sheets and connect it to the Google Analytics app using the add-on. You then create a new report directly in the sheet, and the data from Analytics will be exported into it.
The advantage is that you have the data directly in a sheet, which you can connect to Looker Studio, just like Universal Analytics was connected before. But again, sampling is an issue — for larger websites, the add-on isn't suitable, or you'll need to download data in batches to avoid sampling. The other consideration is the size of the data itself — Google Sheets has a limit of 10,000,000 cells and even smaller datasets will significantly slow it down.
CSV export via API
You can also download data using the Google Analytics API V4, which you can access, for example, via an R script (which can be run from RStudio). It works similarly to the add-on backup, but the output is not a sheet — it's a CSV file, for example — and all it takes is a few lines of code.
Rscript
# Install package if not already installed
if (!requireNamespace("googleAnalyticsR", quietly = TRUE)) install.packages("googleAnalyticsR")
library(googleAnalyticsR)
# Google login
ga_auth()
# Google Analytics view ID
view_id <- "25073918"
# Select metrics and dimensions
metrics <- c("sessions", "users", "bounceRate", "goalCompletionsAll", "transactionRevenue")
dimensions <- c("date","sourceMedium")
# Create request
ga_ua_data <- google_analytics(viewId = view_id,
date_range = c("2022-01-01", "2023-01-31"),
metrics = metrics,
dimensions = dimensions,
anti_sample = TRUE) # Split queries due to sampling
# Print sample output
head(ga_ua_data)
# Storage path and add view ID to filename
filename <- paste("C:/Users/ondrej.svec/Desktop/EXT/P&R/ua_backup/", view_id, "_ga_ua_backup.csv")
# Save to CSV
write.csv(ga_ua_data, filename, row.names = FALSE)
Downloading data this way is relatively reliable and happens without sampling. For larger websites with millions of rows of records, downloading may take a while. The exported CSV can be saved and used later as needed.
Another option is downloading Universal Analytics data via the Keboola interface, if you use it. In Keboola, you can easily select the dimensions and metrics you want to back up, set anti-sampling, and save the data wherever you need it.
Back up, but don't overthink it
If you have a larger website with a data analytics team behind you, don't worry about the backup — they've most likely already backed up the data via BigQuery (or another cloud service) or via R. Similarly, if you work with an agency like ours, they've probably already recommended and implemented a suitable solution.
For smaller websites, back up the necessary data using file export or the add-on. If you've had Google Analytics 4 deployed since 2021, you really only need the absolute minimum of monthly data for the past few years, which is manageable via export.
We also recommend doing annual reviews and evaluating the data — when making comparisons, you may well find that just the overview is enough, without the raw data itself.
Data matters and you don't want to lose it. But there's no need to go overboard and back up absolutely everything. Either way, you need to decide by the end of June.
