If you use Google Tag Manager, you’ll probably have noticed that there is now the capability to track how far down the page your visitors scroll. It’s been possible for a long time with custom JavaScript, but now it’s available out of the box and I can’t overstate how useful this is.

Tracking how far down a page your visitors scroll is a great metric as it lets you understand how engaged people are with your content. Are they reading all of your articles or are they bailing as soon as they read your intro? Is your landing page turning people off instantly, or is it just not the right time for them to buy? Scroll depth tracking will let you know.

But scroll depth tracking isn’t without its controversy: unless you change one specific setting, it changes the way Google Analytics reports on your bounce rate.

Since events are seen as an interaction, a user that only visits one page but triggers the scroll depth event won’t be counted as a bounce, even if they only view one page. This becomes more of an issue when scroll depth triggers are set at low thresholds like 10 or 25% since users with decent sized screens will often trigger those as soon as they view the page so even if they do bounce, they won’t be reported as such.

There are a couple of schools of thought on this.

The Case For Changing The Bounce Rate

Some argue that if a user has scrolled, they have interacted with the page, even if they’ve left afterwards. This is relevant for blogs and other content sites like recipes, where visitors typically land on an article, read it and leave. It’s not uncommon for these sites to have very high bounce rates because of this, so why wouldn’t you want to count a scroll as a proper visit and what advertiser wouldn’t want to be on a site with a low bounce rate?

The Case Against Changing Your Bounce Rate

There are two key reasonings against changing this: firstly, you won’t have historic data to compare against, since this will only work from the moment you install it, so if you make improvements, you won’t be comparing like for like. Secondly, it’s down to accuracy.

As we’ve mentioned, there is the problem of screens automatically counting visits as a scroll if they have seen enough of the page. A lot of desktop screens will show a quarter or more of the page upon a view, so even if they bounce, they'll be counted as an interacting visit. This is where the case against changing your bounce rate comes in. Aside from the event triggering early, there are pages where you want to send people to other pages.

Your homepage, for example, should be used to funnel people through your site. If visitors are scrolling but not doing anything else on a page such as this, they should still be counted as a bounce.

What Should You Do?

Before you change anything, you should consult your measurement plan and decide if this is the right thing for you and your business’ KPI's. Do you want your bounce rate changed on certain pages, such as your blog? If so, set up different events for those pages and make sure your triggers are configured appropriately. If not, click the “non-interaction hit” slider as shown below.

If you’ve never heard of measurement plans or everything we're talking about today sounds new, swing by our digital analytics page to find out more about what we do in this field and get in touch to see how we can help your business understand its performance.

Sign up for our newsletter at the bottom of this page and follow us on Facebook and Twitter for latest updates.

One of the most common questions I’m asked about Google Analytics is the difference between a segment and a filter and the main use case of each of them. I’m often asked why you would ever use a filter when a segment does the same job and vice versa.

In today’s post, I’m going to briefly run you through what segments and filters are, how they work and the reasons for using each of them.

WHAT IS A GOOGLE ANALYTICS SEGMENT?

A segment in Google Analytics lets you view your metrics based upon specific criteria, for example only organic or paid traffic. They allow you to change your data on the fly and you use the whole of the Google Analytics interface just focusing on that data and, crucially, they do not change your data the way a filter does.

A segment can be applied retroactively, so you can see how your organic performance was last year and so on, and you can also create your own segments based on certain specific conditions. You can even share those custom segments with other Google Analytics users.

You can apply a segment to your Google Analytics like so:

Click the Add Segment button and you’ll see the list of pre-configured ones. As you can see, there’s a lot to play with and with the ability to import new segments from the Google Analytics gallery and create your own, there’s plenty of flexibility there to investigate your data from a variety of perspectives.

Segments are great and an essential part of your Google Analytics arsenal, but they’re not without their weaknesses.

Weaknesses Of Segments

As handy as it is being able to alter your data on the fly, there is inherently some lost functionality compared to filters. Firstly, there is less flexibility in what you can do with a segment than a filter – you cannot exclude a specific IP address or series of IP addresses with a segment, for example.

They also have a habit of triggering sampling within Google Analytics, where the data shown in a report is less than 100% accurate. If your dataset is small, you should be OK, but segments do bring this on much sooner

WHAT IS A GOOGLE ANALYTICS FILTER?

A filter is applied to a Google Analytics view and permanently changes the way that the data is collected for that view, rather than changing the way it’s reported on the fly. Unlike a segment, a filter will not change your data retroactively.

Filters offer a great deal more functionality than segments – as well as just replicating the capacities of segments, which would be prudent if you have a high amount of traffic, you can also make sweeping changes to the way your data is collected, processed and reported. You can use a filter to rewrite the URLs in your page reports, for example, or to double-check the hostname or simply to exclude a section of traffic which you know is not relevant (your own team, for example, or bots). You can also unleash the power of regular expressions to really take control of your data.

Filters are a far more powerful solution than segments, but they don’t offer the same flexibility. You would use a filter for a specific task within a reporting view (excluding your own office’s traffic, for example), rather than using it to check the performance of a specific metric in most cases.

Weaknesses Of Filters

With the power of filters comes responsibility in their use. They permanently change the data in a view from the moment they’re applied to the moment you remove it. There’s no going back. They also can’t be applied retroactively in the same way a segment can. It’s this permanence, plus the additional Google Analytics knowledge required to set up a filter that is the key weakness of them.

In line with best practice, you should always have a completely unfiltered “All Website Data” view, to ensure data continuity and to use for checking that your data is coming through properly. You should then have other filtered views depending on the kind of requirements your site has.

At the very least, we suggest having the All Website Data view and a view which filters out your own IP address and the IP address of any partner agencies/ other offices etc, although we would typically go much deeper than this with a Google Analytics setup.

WHEN TO USE SEGMENTS & FILTERS

A segment is the best way to isolate a certain metric, channel or device in your reporting view and apply that to your historic data. If you want to see how many people have come to your site over the last three years from Facebook on their tablets, a segment is the way to go.

If you need to permanently change the way your data is collected, such as excluding your IP address, removing bots, or rewriting your URLs so that they’re easier to read in reports, you’ll be looking for a filter.

The key thing to understand about filters vs segments is that there is really no “vs” at all. They’re different tools for different tasks and a good setup uses them together. For most reports, you’ll be relying on segments to isolate and highlight different metrics, but to ensure that your data is as clean as it can be, you’re going to need filters to be involved.

Unsure of how well your Google Analytics setup stands up to best practice? Get in touch through the contact form and let us see what we can do to help.

Follow us on Facebook and Twitter for the latest updates.

Measuring and optimising the performance of their investments is the goal of every marketer. To do this, a new concept has emerged: algorithmic attribution.

What is attribution?

Attribution allows one to understand the contribution of each facet in the path of conversion of a user (purchase, lead, reservation, etc.), in order to be able to measure its impact and its performance.

In practice, attribution creates a model by assigning weight to each interaction between a prospect or customer and a brand. These interactions can be online: SEO, SEA, Display, Social, .... Or offline: television, call centres, physical stores. Thus, the assessment goes beyond the reductive (and inaccurate) practice of assigning conversions to the last click of the conversion path. Attribution models are used to measure the profitability of each interaction and maximise the return on investment (ROI) of the advertiser. Attribution is a tool appropriate for all advertisers: small and big budgets, pure player or traditional actor, and to any type of industry: insurance, travel, e-retailer, humanitarian association, etc…

What is attribution for?

In practice, attribution provides answers to many questions:

• What is the true performance of my budgets and campaigns?
• I have an extra budget, where do I invest it to boost my sales?
• I have to reduce my investments by $10,000, where do I have to disinvest?
• What is the impact of my online investments on the offline and vice versa?
• How can I avoid duplicating the same sale to different channels?
• What is the actual impact of display ads on my performance?
• Visualise and understand synergies between channels; How to exploit them?

How does algorithmic attribution work?

The goal of attribution is to attribute a share of the conversion to the channels involved in the path, according to different rules. To do this, two main categories of models have emerged: "ad hoc" models and algorithmic models.

Ad hoc models are user-defined models, which use assumptions to build their model that reflect their understanding of customer dynamics. For example, an ad hoc model may consider that a click on a product page should weigh more compared to a click on the home page. These assumptions more often reflect the intrinsic biases of the designer's opinions than the reality of the customer journey.

Conversely, algorithmic models abstain from any ab initio hypotheses and allow the computer to use all available data to find the model that best represents the customer path. Although much more complex to implement, algorithmic attribution represents the best assessment of the weight of each step in the customer journey. It is important to pay close attention to how these algorithms function, and to understand the variables on which they are based to ensure that they evolve according to the media mix and time. Certain metrics and characteristics are essential and must be present as foundations of the algorithm:

• Consider all channels.
• Take into account the nature of the point of contact (click, print or real view).
• Take into account on-site user engagement.
• Take into account how soon after the interaction a conversion is gained.
• Adapt according to the evolution of the media mix.

Integrate as much data as possible with the model: Why? How?

Algorithmic attribution does not only apply to digital channels. If you are able to recover structured and comprehensive offline data, you will be able to measure the performance of your media mix as a whole. The allocation model will then combine the online and offline data and give you a holistic view of your different channels as well as existing synergies between channels.

In particular, granular integration of offline media investments will make it possible to measure return on investment (ROI). Similarly, a cross-device reconciliation or the integration of your CRM data will allow for a more precise understanding of customer paths towards the act of purchase. It is therefore necessary to maximise the use of connectors in order to automatically integrate all this data within the same platform. Once the data is collected, it is then possible to measure, understand, and improve its media mix as a whole. As a result, you will be able to control, via a single tool, all the channels you use.[box]

What are the key success factors?

When starting up an attribution project, several key success factors need to be given particular attention: first, the quality and completeness of the data used, in order to create the model that best reflects the reality of the impact of each marketing channel. Using false or partial data creates an inaccurate model. Second, you should measure your return on investment by analysing the lifetime value of your customers rather than a discrete event, such as a purchase, that only partially reflects the interaction of your customers with your brand.

Finally, it must be taken into account that algorithmic attribution is a technically complex subject. Choosing the technology that suits your needs and being guided, temporarily or over time, by subject experts is an investment that will pay over time.

For more information on how you can incorporate multichannel attribution into your marketing strategy, tweet at us @ESV_Digital_UK or follow us on LinkedIn.