Do you trust your Google Analytics data?

Watch a re-run of our Fresh Thinking Live! analytics-focused webinar from 29/07/2020.

large-dot

Watch the recording where the Fresh Egg team members why it is important to maintain trustworthy Google Analytics data.

Taking part in the discussion were: Dwight Thomas, Business Director; Graham Marsh, Senior Web Analyst; Dinah Alam, Head of Conversion Services; Julian Erbsloeh, Head of Analytics

1. What skills and experience do I need my team to have to be able to maintain my analytics framework and the integrity of my data?

Julian: One of the most common challenges we see is not skill or experience but time. Like all other areas of your business or organisation, analytics requires investment. While the tools we use are often free, you need to assign ownership over the data quality to someone and make sure they have the time to look after it.  

The data collection side of web analytics can be quite technical, so your analytics champion should not shy away from using the console to debug tags. They don’t need to be a developer but know when to ask one for help.

 

DinahYes, I agree, learn from a course, another person or from resources online to get you going and then continue to learn, and always good to get some orientation and to get ahead.

However, I believe you learn the most from real experience, so a course can only take you so far. 

I think it’s important to say that some of what we have been talking about today are very technical and that might be a skillset area which is difficult to fill. Not because the discipline is too hard, but because it means you might be looking for a wide-ranging skillset in one person. 

Therefore, play to your team’s strengths and enhance where needed and be flexible to the idea that they will need to learn all the time.

Experience is excellent, of course, but from my experience curiosity, resourcefulness, and a desire to understand what is happening on the website are equally valuable, if not more so. Most people in your team, with some training and a robust maintenance process, can do the job but they need to be curious to be good at it so find the person that shows an interest.  

2. How much of the solution to my internal analytical talent gap should come from “formal” external training vs on the job training?

Julian: It depends on where you are starting the journey. If you are starting from zero or minimal experience, structured training at the beginning is vital. Like skiing, you can probably teach yourself, but it will take a long time, and you are bound to make mistakes which will later be difficult to correct.

Training can also be useful for the wider team, so the owner of the data quality as well as the users of that data, as it helps understand the full potential of the platform and aligns language and understanding.

The on-the-job training part never ends. There is not a week where I don’t learn something new, and I have been working with GA for over ten years now. All the platforms are constantly evolving; the pace of change is accelerating, so we have to make sure we dedicate time to staying in touch and testing new features and functionality.

DinahYes, I agree, learn from a course, another person or from resources online to get you going and then continue to learn, and always good to get some orientation and to get ahead.

However, I believe you learn the most from real experience, so a course can only take you so far. 

I think it’s important to say that some of what we have been talking about today are very technical and that might be a skillset area which is difficult to fill. Not because the discipline is too hard, but because it means you might be looking for a wide-ranging skillset in one person. 

Therefore, play to your team’s strengths and enhance where needed and be flexible to the idea that they will need to learn all the time.

Experience is excellent, of course, but from my experience curiosity, resourcefulness, and a desire to understand what is happening on the website are equally valuable, if not more so. Most people in your team, with some training and a robust maintenance process, can do the job but they need to be curious to be good at it so find the person that shows an interest.  

3. At what stage do I need to consider budgeting for a paid analytics tool?

Graham: That is a question of data maturity. The problem occurs by the fact that you're pushing the limits of free tools. For example, Google Analytics has limits on the amount of data you can collect as well as sampling limits that can cripple the ability to analyse data for high traffic sites effectively.

If your organisation has a large ad spend, it can make more sense to upgrade to a paid tool because it's so easy to show the direct ROI. GA 360 has integrations with DoubleClick, and other ad platforms and the ability to be able to do more advanced data analysis can pay for the cost of the tool. Using Google Analytics as an example, the ability to access granular hit-level data enables you to do propensity modelling. By doing this, it allows you to calculate the likelihood of your visitors to convert, which you can feedback into your advertising platform and use to create effective remarketing strategies.
The key is that you need the resource/talent to be able to maximise the value you're getting out of your tools.

Dinah: Funny, if I go back 8-10 years the question would have been "When can I ditch my expensive paid analytics tool for the free ones. When will they be good enough?"

It's not some evolutionary step that everyone needs to make.

Is it possible to resolve the problem? Understand if the issue that's stretching the free tool is fixable, e.g. with other data extraction methods. Don't make software decisions lightly.

4. So many of our key data points are still offline. What can I be doing to better understand overall performance?

Dinah: Joining up journeys on and offline is challenging, but often it’s just a case of starting to find out what touchpoints and reporting exist already, perhaps in other systems as that will be a clue to how you get the data into your reporting.

And you are not alone – even the most sophisticated of data reporting set-ups can be lacking in this area. I have found that data can be a lot harder to find on retention journeys.

Typically more effort is put into the acquisition. As a lot happens online, it is common to observe a lot of data around that but hardly any focus on retention. Remember to expand your reporting into this area if you want a good measure of your business.

Graham: Understanding performance is a common challenge, but there are solutions to the problem. The first thing I would recommend is to take a strategic approach to measurement. Map out all of the data points you need to track throughout the user journey. Once you have these, you can ensure you’re collecting data in all those areas. Then it’s about working out the best way to bring data from various sources together.
There are simple ways to do this, for example, by using data studio to create dashboards with data from multiple sources or sending Google Analytics events in from other systems. But you can also combine data in cloud data storage platforms like Google Bigquery, where you can also process and transform that data and run advanced analysis on it like forecasting and propensity modelling.

5. How regularly should I be reviewing my Google Analytics set up to ensure my data is validated?

Graham: The answer is it depends on the complexity of the setup and how often your website or app is changing.

There are tasks that we recommend carrying out at least monthly, e.g. checking for PII, checking your goals and e-commerce tracking are working. Other jobs might only need checking quarterly or annually.

We use a maintenance schedule to keep track of all the tasks and tend to start with our standard template and then adjust the frequency within the first few months once we get a feel for it.

If you can automate some aspects of it by creating tools like the Analytics Troubleshooter, then that frees up more time for more valuable tasks like analysing the data and taking action off the back of it!

6. How does “sampling” work for the free version of Google Analytics? Does sampling really matter?

Graham: Google Analytics uses sampling to create a mix of accuracy versus speed when loading reports. For built-in reports you will rarely see sampling but when you start interacting with these reports by adding filters, segments or secondary dimensions Google Analytics (free version) will sample the reports if the underlying data for them is more than 500k sessions. For some organisations, this means even month on month or week on week reporting can suffer.

Whether it matters or not depends a lot on the level of sampling applied, which you can see why clicking the orange shield at the top of reports. If the information is sampling at 80-90%, then you can be confident this is a realistic picture. As this level starts to drop, you can see considerable variations in metrics for unsampled versus sampled data.

7. Why is there sometimes a small discrepancy between my Google Analytics report and the same data in Data Studio?

Julian: Small discrepancies between data sources can be frustrating, so it is essential to understand when it is worth spending the time to investigate and fix it and when to learn to live with it. 

Discrepancies between GA and GDS fall into the former camp for me, while trying to match numbers on different platforms, using different data and attribution models falls into the latter.
If you are sure that the configuration matches between GA and GDS (think filters, segments, data source, date range, etc.), the answer is most likely ‘sampling’. Data Studio sampling thresholds are the same as Google Analytics, but the two may not be using the same sample of data to extrapolate your numbers.
The threshold for sampling is 500k sessions within the time frame you are looking at, but sampling can impact your numbers much sooner if filters or segments with multiple conditions are applied to the data.

Look out for a small ‘Show Sampling’ link in the footer of Google Data Studio – if that link appears, your data is sampled.