08 October 15 -

6 ground rules for effective web analytics.

Modern marketing is complicated. We’re using multiple devices to interact with brands and make purchases. We’re using multiple apps. Search behaviour is changing. The search market is changing. Advertising is getting more and more complex.

But most importantly, purchase behaviour has changed. As consumers, we’re in almost complete control over our purchase decisions. We jump in and out of websites, check out reviews, read blog articles, sign up for newsletters, use price comparison sites, and compare multiple providers at the same time. The traditional purchase funnel is dead, and has morphed into something enormously complicated.

With all this increasing complexity, it’s getting harder and harder for us to answer questions like, “What should we do next?”, and, “What strategies and tactics are going to help our business achieve our goals?”

To make matters worse, marketers rarely make decisions based on data.

According to an ongoing study by CMO, also mentioned in JD’s recent article on our blog, less than a third of projects use marketing analytics. What’s more, the number seems to be in decline:

Percentage of projects using marketing analytics

All of this raises the questions: If things are getting more complicated, and we rarely make decisions using data, how on earth are we making decisions on what to do to improve the performance of our websites?

Are we just going on gut feel? Do we just copy what everyone else is doing? Or are we choosing tactics at random?

These days, almost every website owner (with any sense) has some sort of analytics software installed on their website. For most, this means Google Analytics.

It’s certainly no panacea, but Google Analytics is incredibly powerful. And, unless you are using the premium version, it’s also completely free. The problem is, I come across very few Analytics accounts that have been set up properly, and rarely see evidence of clients using it to its potential.

That’s what this article is all about: helping you get more insight from your Google Analytics data, so you can make better decisions.

And all it takes is 6 ground rules.

Note: This article is aimed at people who have a basic knowledge of Google Analytics. Maybe you’ve got it installed on your website, and you know how to use the standard reports, but haven’t gone much further than this. If this sounds like you, keep reading…

Ground rule #1: Set yourself up for success.

I think it’s fair to say, Google Analytics is an amazing piece of software. Out of the box, you get detailed information about your audience, how they found you and how they behaved.

But Google Analytics can do so much more than this. I’m constantly amazed at how often I see Analytics accounts that are only set up with the basic ‘out-of-the-box’ setup.

When you just use Analytics out-of-the-box, you don’t control the way in which data is collected. Nor do you configure how that data should be processed and reported on. You’re only benefiting from a certain amount of information.

When you turn everything else on, however, you start to get data that will give you serious insight into your users. And oftentimes, it only requires very simple changes in how you collect data and configure your account.

  • Track events. At its core, Google Analytics is a pageview tracker. It tracks when people go from one page to another, and gives you information about those interactions. But you can also track individual interactions that happen within a page, not just the pageview itself. Examples might be PDF downloads, video plays, scroll depth, link clicks. In fact pretty much anything a user does with their keyboard or mouse on your website, in theory, can be tracked. The only thing you shouldn’t track is people’s personal information.
  • Track conversions. Do you have a contact form, a newsletter signup, gated content, a shopping cart? All of these things might be considered ‘conversions’. These aren’t automatically turned on in Google Analytics – you have to configure it to track these things. Tracking conversions is absolutely critical to effective analysis – make sure you set up conversion tracking!
  • Get data from other Google platforms. If you use Google’s Search Console, or are doing any advertising through Google Adwords, you can pass the data from those platforms into Google Analytics.
  • Track clicks from campaigns. If you’re running any kind of online campaigns, such as eDMs, you can pass data from these campaigns into Google Analytics.
  • Get internal site search query data. Do you have a search box on your website? Would you like to know what people are searching for? Google Analytics’ ‘Site Search’ feature can give you that information. But, again, it’s not turned on by default; you have to turn it on.
  • Learn the demographics and interests of your visitors. Want to know the age, gender and interests of your visitors? Google Analytics can tell you this information. But – you guessed it – it’s not turned on by default; you have to turn it on.
  • See how you compare against other similar sites. For the more competitive among you, you might be interested in the ‘Benchmarking’ feature in Google Analytics. See how you compare against other similar sites on metrics such as visits, session duration and more. It’s not turned on by default…
  • Pass in your own online data. Depending on your website, this could be a powerful feature for you. Any data on your website can be passed into Google Analytics (provided it doesn’t personally identify your visitors). If you are running a multi-author blog, you might want to pass in information such as the author and the blog category. This information isn’t automatically tracked by Google Analytics, you have to turn it on.
  • Pass in your own offline data. If you are collecting data from an external system (such as your own CRM system), you might want to join and analyse that data in conjunction with your Google Analytics data. Using Google Analytics’ data import feature, you can do exactly this – pass in your own offline data and analyse it right inside Google Analytics.

You might say data for data’s sake is pointless. My opinion? It makes sense to track as much as possible, so if the time comes and you need to make a decision about something, data takes opinion out of arguments. You can’t get insight about your visitors when you only have limited data.

Set yourself up for success. Track everything!

Ground rule #2: Experiment with different date ranges.

When you first log into Google Analytics, by default, you are presented with the past 31 days of data.

And if you compare to the previous period, you’re comparing back another 31 days.

Now imagine you log in to Google Analytics on a Monday morning, click on ‘compare to past’, but you see all of your top-level metrics are looking pretty bad:

Google Analytics compare last 31 days

You think to yourself, “Darn it, our traffic has gone down. This is a crappy start to the week”.

The problem is, the current date range contains an extra weekend. You’re not looking at a fair comparison.

If you’re only looking at a short time period such as a month, I always recommend you set up your date ranges in multiples of seven; for example, 28 days instead of 31. You’ll see now that the graphs line up, and you have an equal number of weekends in each time period:

Google Analytics compare last 28 days

You’re now looking at a fair comparison, and when you look at those top-level figures, you realise your traffic is actually up, not down.

But there’s a problem with 28 days too (or any short time period for that matter).

If you only looked at 28 days in isolation, you might think to yourself, “hey our traffic is looking on the up, this is great!”

Traffic trend 28 days

In fact, if you’d looked at a longer time period, you’d have realised your website has been experiencing a steady decline for the past three to four months:

Traffic trend 4 months

To overcome this problem, look at longer time periods such as six months, or even a year. This way you’ll be able to see how your current stats compare to other periods throughout the year; not just compared to last month.

But there’s a problem with longer timeframes too: seasonality. Almost every business will experience seasonality of some sort.

If you’re looking at six months of data and you compare to the previous six months, you’re probably not comparing apples with apples. For many businesses, there will be a drop in traffic in December around Christmas holidays.

It makes more sense to compare back to the same time the previous year, not just to the previous six months.

A little tip – whenever you’re looking at Analytics, make sure you’re looking at enough data to make an informed decision. If you only look at really short time frames, there’s a chance that whatever you’re looking at could have occurred by complete chance, and might not be representative of what’s really happening.

Looking at short time frames might make you jump to the wrong conclusion, leading to poor decisions about what to do next.

So, experiment with different date ranges. See how the different date ranges affect trends. Don’t restrict yourself to a single month of data.

Ground rule #3: Focus on meaningful metrics.

Image-6

Remember hit counters? The accepted standard of web tracking in the ’90s. You wouldn’t dream of using a hit counter as a measure of success these days, right?

So, why do we measure success using things like ‘pageviews’, ‘visits’ and ‘followers’?

These are examples of vanity metrics. They make us feel good, but don’t tell us much.

There is a place for these types of metrics, but they definitely aren’t the end game. Yet so many people still use them as their main measures of marketing success.

Why do we obsess over these types of metrics?

  • They’re really easy to measure. Whenever you log into Analytics, what’s the first thing you see? Sessions. Google Analytics rams this metric down your throat. It is the default metric for pretty much every graph in Analytics. So unless you deliberately change the graph metric to something else, you’re going to default to sessions.
  • ‘Up and to the right’ makes us feel good. It makes us feel warm and fuzzy inside when we see our number of Facebook likes, or Twitter followers go up. The thing is with these types of metrics is that they are cumulative. So unless people are actually un-liking or un-following you, the number will always go up. It makes you feel great, but doesn’t actually tell you what’s going on.

So, what’s a good metric?

Here’s a really good way of looking at things. A metric should change your behaviour. When you look at a metric, it should be so relevant to your business that any changes in that metric should trigger you to take some sort of action.

So, how do you choose meaningful metrics? I think it helps to start with these two questions:

  1. What type of business are you?
  2. What do you want people to do?

Your business type and your objectives will have a fairly big bearing on what you choose as metrics.

Let’s look at some examples depending on the type of business you are…

  • Ecommerce. You want people to buy your products, and come back and buy again. Meaningful metrics might be number of transactions, revenue, revenue per customer, and customer lifetime value.
  • Lead generation. You want people to get in touch with you. Meaningful metrics might be number of leads, cost per lead, and cost per acquisition.
  • Publisher. You want people to stay active on your website. The more impressions you get, the more ad revenue you receive. Meaningful metrics might be pages per session, average session duration, unique pageviews, and ad revenue.
  • Informational site. You want people to find information easily. Meaningful metrics might be user feedback, file downloads, and number of contact us requests (with fewer being better).
  • Online software. You want people to sign up, stay active and upgrade. Meaningful metrics might be new signups, cost per acquisition, active users, revenue per customer, and churn.
  • Brand. If the aim of your website is to do none of the above, but simply to promote your brand, you probably just want people to fall in love with you. Meaningful metrics might be things like shares, comments with positive sentiment, and loyalty (repeat visitors).

As you can see, apart from ‘Publisher’, metrics like ‘pageviews’ don’t feature anywhere. They’re important, but not the end game.

Align metrics to the customer journey

Another really good way of looking at metrics is to think about what you are currently trying to achieve.

If you’re trying to build awareness, then those ‘vanity metrics’ I talked about might be useful. Likes, followers, sessions and so on are all indicators that your awareness is building.

However, if you are getting lots of traffic to your site but it isn’t converting well, measuring success on things like followers and sessions isn’t going to help you.

Instead, you’d be sensible to look at metrics in the consideration and purchase stages of the purchase funnel, to see what’s happening there. Examples might be ‘time to purchase’, or ‘shopping cart abandonment’. Or you might want to look at the type of traffic you’re sending to your website.

So, there’s ground rule #3: focus on meaningful metrics. Zero in on what really matters to your business, not on those distracting and meaningless vanity metrics that don’t change how you behave.

Ground rule #4: Segment your data.

Whenever you look at data that represents all visits, you’re looking at what’s called ‘aggregate data’.

Whenever you look at data that represents a subset of the overall visits, you’re looking at ‘segmented data’.

If you load up any ‘overview’ report in Google Analytics, the metrics you are looking at are an example of aggregate data:

Google Analytics Overview Report

It’s nice to know whether metrics such as total visits, total revenue, total pageviews, time on site, bounce rate etc are improving, but they tell you nothing about why things are changing, or what to do to improve.

Segmenting your data is the key to success. Segmentation lets you dig deep into your data to discover what’s happening. In fact, I’d go as far as saying the only way you’ll get insight from your data is through segmentation.

Luckily, Google Analytics actually does some segmentation for you. Whenever you load up any of the standard reports in Analytics (except for the overview reports), you’re looking at a version of segmented data.

Google Analytics All Traffic Report

In this example, each of the different channels in this report, be that organic, direct, referral, paid etc is an example of a segment.

Segmentation lets you go one level deeper from aggregate data, giving you a little bit of insight.

But many people stop there. They rely on the standard reports in Analytics, and that’s as far as they go. In fact, you can further segment that data, and you can do it based on almost any dimension or metric in Google Analytics.

  • Do females behave differently to males?
  • Do people behave differently depending on whether they used a desktop, tablet or mobile device to browse the site?
  • Do single-visit users have any distinct characteristics from multi-visit users?
  • Is there anything different about converters versus non-converters?

Segmentation allows you to compare different user groups, helping you to pinpoint differences. This deeper segmentation is where you’ll get critical insight into your visitors and their behaviour, and this is the key information you should draw upon to influence your marketing decisions.

How do you create more segments?

The best way to segment data in Google Analytics is to use the feature at the top of almost every page, aptly named ‘segments’.

Google Analytics Segments

This is where the magic happens.

Within the segments dialog, you’ll be able to select a pre-configured segment, or alternatively create your own segment.

One of the most useful ones for ecommerce websites – I think – is to compare users who didn’t make a purchase, with users who did make a purchase.

Imagine you’re looking at the demographics overview and you’re looking at the age of your visitors. When you don’t have any segments set up, it looks something like this:

Google Analytics Demographics Overview

When you set up those two segments I just mentioned – ‘did not make a purchase’ and ‘made a purchase’ – you can start to analyse whether there are any differences in those two user groups.

Google Analytics Demographics Overview Segmented

Here’s another example. Is there any difference in the people who don’t make a purchase versus those who do make a purchase, depending on what device they use?

Mobile users segmented

Or, is there any difference in the people who don’t make a purchase versus those who do make a purchase, depending on their interests?

Audience Interests Segmented

So, that’s segmentation. Don’t fall into the same trap a lot of companies fall into. If you try and target everybody – by definition – you end up targeting nobody.

Segmentation helps you pinpoint the people who are most likely to positively influence your goals, then lets you align your marketing tactics to those people.

Ground rule #5: Learn how to attribute conversions.

For most businesses, online conversions will often involve more than a single touchpoint.

Unless you’re in an industry like emergency plumbing, it is unlikely that someone will come to your site and convert on the first visit.

You can see this when you look at the ‘top conversion paths’ report in Analytics, which shows you the different paths people took before they converted.

Multi-Channel Top Conversions Paths

What’s this ‘attribution’ thing all about?

Imagine a typical customer journey like this. A user comes to the website first by clicking on a display ad. They leave the site without buying, then return a few days later via a Google search. Again, they leave the site without buying, but then return a few days later via a direct click, and this time they make a purchase:

Typical Conversion Path

This typical journey involves three channels – display, organic and direct. The display and organic clicks would be referred to as ‘assisting interactions’, whereas the direct click would be referred to as the ‘last interaction’.

The problem with most Analytics software (including Google Analytics), is that many of the reports within that software will attribute conversions only to the ‘last interaction’.

So, when you look at a report, such as your channels report, you’ll see something like this:

Display Channel 10 Conversions

In the above example, display advertising has only resulted in 10 transactions, close to $1800 worth of revenue, and a conversion rate of only 0.43%.

You might look at that data and think to yourself, “display advertising is doing nothing for us, let’s turn it off”.

The problem is, you’d be discounting the fact that, although display advertising wasn’t the last click before the conversion, it did actually contribute to bringing customers to the website. It acted as an assisting interaction.

When you are analysing conversions in Google Analytics, I recommend you take a look at the ‘assisted conversions’ report:

Display Channel 88 Assisted Conversions

In this example, although display advertising only resulted in 18 ‘last click’ conversions, it actually contributed to more conversions as an assisting interaction.

Had you not known about this report, your previous decision to turn off your display advertising may have been a little hasty.

Attribution gives you a mechanism to give credit to different channels, even if they weren’t the step immediately before a conversion. And it’s important to do this analysis, because it differs depending on industry, geographic location, and even the size of your business.

Google have actually done a huge amount of research into this area, analysing millions of consumer interactions to see which channels are more likely to act as an assisting interaction, and which ones are more likely to act as a last interaction.

Assisting and Last Interaction Channels tool

They found that it varies depending on the size of a business, the industry, and even the location.

Attribution modelling

So how much credit should assisting interactions be given? Should they be given as much credit as last interactions?

This is where something called ‘attribution modelling’ comes in. There are many different types of attribution models, and you can even create your own. But here are five common attribution models you should be aware of.

  • Last interaction model. The last interaction gets 100% of the credit for the conversion. Not that useful if you ask me.

Last Interaction Model

  • First interaction model. The first interaction gets 100% of the credit. Again, not that useful.

First Interaction Model

  • Linear model: Each interaction is given an equal amount of credit for the conversion. This seems much more logical compared to ‘last’ and ‘first interaction’ models.

Linear Model

  • Time decay model: Interactions closer to the purchase get the most credit, which ‘decays’ for interactions earlier on in the journey. Again, this seems logical.

Time Decay Model

  • Position-based model: The first and last interactions get a lot of credit, and the rest is distributed between all other interactions. This also seems logical.

Position-based Model

Which attribution model do you choose?

It really depends on your business, and also on each type of conversion.

The best thing to do is play around with the different models and see what works for you. There’s a really good tool in Analytics that allows you to do exactly that. It’s suitably named the ‘model comparison tool’:

Attribution Model Comparison Tool

So play around with it and see what works.

Without attribution, you could end up throwing away money. Or you could end up ignoring a channel that is actually helping to make you money, albeit indirectly. Learn how to attribute conversions, so you can properly analyse the impact of your different marketing channels.

Ground rule #6: Raise questions.

Ground rule #6 is the shortest to explain, but probably the most important.

I want to talk about dashboards.

Google Analytics Dashboard

There’s definitely a place for dashboards. They’re especially good for monitoring key metrics (like the ones I outlined in ground rule #3), and to help you quickly get to data that you want to review on a regular basis. Dashboards are great for monitoring the health of the business, and help you to quickly identify any anomalies that may need to be investigated.

However, some people only look at their dashboards. They don’t dig deeper, meaning they only see a small snapshot of their overall data and miss out on interesting and important insights.

If you want to get real insight from your data, don’t rely solely on your dashboards. Instead, raise questions.

Such as:

  • “Are any of our products too expensive?”
  • “Are people struggling to find our blog?”
  • “Do our eDMs generate any sales?”
  • “Should we make our website responsive?”
  • “Are we losing potential customers during the checkout process?”
  • “Is anyone using our slide-out navigation on mobile?”
  • “How often do people view the individual slides within our home page carousel?”
  • “Do people read all the way to the bottom of our blog articles?”
  • “Who are our biggest spenders… and how do we find more people like them?”

It’s easy to set up a dashboard and say you are ‘doing analytics’, when in actual fact you may be doing nothing to change the way you behave or improve your business.

Dashboards are helpful; keep using them. But don’t only use dashboards. In addition to them, raise specific questions, then start digging into your data to find the answers.

***

So, those are my six ground rules for effective analytics. Jump in to your analytics account and start playing around. Set your account up to track everything important to your business. Play around with the date ranges to spot trends and anomalies. Think about metrics that are meaningful to your business. Segment your data, and learn how to attribute conversions.

But most of all, start raising questions. Reduce your reliance on dashboards and start seeking out insights.

Because they’re not going to jump out on their own.