What is Behavioural Data And How Do You Use It?

Behavioural data lies at the intersection of all customer touchpoints and channels.

Websites, mobile apps, CRM systems, marketing platforms, call centres, customer service desks and payment/billing systems are all sources of behavioural data. 

Behavioural data powers recommendation engines and helps optimise web and app design – UI and UX – to improve customer experiences. For marketers, behavioural data automates marketing strategies to increase acquisition, activation and retention and reduce customer churn. 

Analysing behavioural data – or behavioural analytics – seeks to examine the “how”, “what” and “where” to understand the customer’s “why”. 

It helps answer questions such as; “why do customers unsubscribe after 4 weeks?”, “why are sales higher on Wednesday morning, is there a connection to something else we’re doing?”, “why is X customer so much more likely to stay with us in the long-term?”

Here, we’ll be discussing some fundamental concepts and processes in the field of behavioural data and behavioural data analysis. 

The Benefits of Behavioural Data

Behavioural data equips businesses and organisations with an understanding of their customers.

Behavioural data is a glimpse of the past as well as the future – it’s possible to analyse a customer’s journey through all the touchpoints of every product and service.

A strong conception of where customers come from, how they interact with your business or brand and what they end up doing, e.g. churning or becoming loyal customers, allows businesses to predict what other customers will do in the future.

Once you use behavioural data to optimise campaigns and improve KPIs, you’re well on your way to flexing your brand like a muscle that expands and contracts with your customer’s behaviour.

Practically speaking, behavioural data is crucial for building and optimising recommendations engines, marketing campaigns, sales channels, social media strategy and content.

---

The Pros of Behavioural Data

• Anticipate what customers expect and build recommendations engines
• Optimise promotional campaigns for when they are most effective
• Know when to send emails and/or push notifications for maximum impact
• Improve your understanding of how customers move between touchpoints and channels
• Predict what customers will do; e.g. when they tend to exit your brand and desubscribe

---

The Cons of Behavioural Data

• Collecting and analysing the data is arduous when many touchpoints are involved
• Optimising based on behavioural data can have unexpected consequences, particularly if such data is used to train unsupervised machine learning models
• The collection of certain behavioural data poses a regulatory and ethical problem, e.g. you will have to comply with GDPR
• Behavioural data is most effective when there is plenty of it, scarce or low-quality data is a problem
• Incorrectly interpreted customer data results in incorrectly interpreted customers

---

Google Analytics: The Fundamental Behavioural Analysis Tool

Google Analytics includes under Audience tabs labelled “Behavior” and also “User Explorer” amongst other functions which delve into aspects of user behaviour. It does what it says on the tin, allowing the user to analyse and track metrics associated with their site(s). 

Today, Google Analytics has become more evolved and offers several options for behavioural data collection, analysis and reporting:

• Site Content: Somewhat similar to an SEO audit
• Behavior Flow/Events: Tracking actions that aren’t linked to clicking through links (e.g. playing a video)
• Experiments: For A/B comparisons and optimising UI/UX
• AdSense: Data analysis for AdSense
• User Explorer: For exploring users and their interests 

The tool also provides insights into your audience’s interests inside and outside of your niche(s). It lets you take a look at the psychology and behaviour of your audience – but we must be careful to not over-assert any conclusions based on this data alone. 

Correlation may not infer causation when it comes to behavioural analysis and it may not be wise to make simple assertions about why your audience/users do x or y. 

As our knowledge of behavioural analysis has broadened in recent years, and since we’ve become equipped with new tools, we’re now able to tap deeper and deeper into what users want and why. 

Here are 3 key areas in behavioural data that businesses often target for analysis:

• Acquisition 
• Activation 
• Retention 

---

Using Behavioural Data For Customer Acquisition

Potential customers have to be open to interacting with your brand or product prior to you acquiring them.

Behavioural data is the key to measuring customer acquisition. 

It might be a case of tracking the traffic generated by a campaign; the conversion point might be a sign-up, subscription or other forms of interaction like downloading a product or booking a demo or appointment. This doesn’t necessarily mean that customers will onboard successfully, nor that they’ll be retained. 

Still, this sort of data can help you home in on the customers you really want to attract to your business. What if you’re able to optimise your campaigns to attract customers with the highest retention rates and customer lifetime value (LTV)? 

You might also optimise your campaigns for those who are more likely to actually complete onboarding. 

---

Using Behavioural Data For Customer Activation 

Customers activate their interest in your product(s) when they onboard and begin to derive value from the product. Activation follows onboarding. Onboarding strategies are important for more complex products that have a few stages between initial interest and activation (e.g. retrieving and inserting a code into an app). 

Behavioural data will reveal who successfully activates their product and can assist you in ascertaining why some are more likely to activate than others. 

---

Using Behavioural Data For Customer Retention

Retention and customer churn are massive problems today, largely because of the volume of competitors that surrounds the customer in most sectors/industries/niches. This makes it exceptionally easy for customers to bounce from one competitor to the other.

• Recurly suggests an average churn rate of 5% monthly. 
• Business2Community states similarly that businesses should expect to lose 6% to 8% of customers each month. 
• McKinsey discovered that 40% of subscribers cancel eventually, as many as a third cancel in 3 months. 

Behavioural data can reveal why customers do this, when, and how to prevent it. For example, you can run different promotions to track whether or not they reduce churn, or learn what campaigns are most efficient for retaining customers and encouraging future purchases. 

---

Behavioural Data: The Tools

Let’s take a look at some tools (other than GA) that can help you with the above common business problems. 

These are broken down into several categories as there are several stages to building scalable, robust and omnichannel-capable behavioural data systems. 

---

1: Data Collection and Storage

Data collection and storage sits firmly towards the engineering end of the professional data spectrum. The aim is to collect and organise data at its closest proximity to the source. Sources might include social media, Google, offline data, marketing tools, CRMs, etc. 

One popular tool for customer data collection and storage is the customer data platform (CDP). This aims to centralise network-wide data in one customer-oriented database, which is perfect for behavioural analysis. 

Key names in CDPs include: 

• Segment 
• MetaRouter 
• mParticle 

With these CDPs, it’s easy to pool data from a variety of sources using simple APIs. For example, you could pool data from SaaS platforms and sync it back to the tool, allowing you to operate a headless customer data warehouse with multiple integrations across web app, mobile and websites. 

You could also use ETL data pipelines to ingest data, load it into warehouses and transform it for use.

Key ETL providers include:

• Stitch 
• Matillion 
• Fivetran 
• Meltano 
• Airbyte 

You can read more about data pipelines on the blog here.

Either CDPs or ETL pipelines can build a single view of customer data sources, though CDPs are rising in popularity lately. 

---

2: Marketing Attribution 

Marketing attribution allows you to attribute specific actions (e.g. conversion) to certain campaigns. It’s the main method used to assess the performance or ROI of different channels and campaigns.

It’s all well and good tracking sales data, but at which exact point does a customer convert? Is it quite early into their journey (e.g. after clicking on a paid ad), or do they need to be exposed to multiple ads and/or promotions?

Multi-source marketing attribution is the best way to model multi or omnichannel behaviour. It accounts for the user or customer’s entire journey from point to point. Points might include mobile apps, website landing pages, social media ads, subscription content – anything that creates a loggable event in the customer’s journey.

Here are some typical MTA models you might find when analysing attribution data:

• Linear. This is the simplest MTA model which provides equal credit to every touchpoint in the customers’s journey.
• Time decay. A time delay between first and last touch points might suggest that the last touchpoints are more effective, since they’re closer to conversion.
• U-shaped. A U-shape emphasises the first and last touch points. This suggests and ‘in and out’ journey where a customer joins at one touch and converts/activates at another, which is a fairly easy behaviour to measure. There may be some interaction with touches in the middle. For example, a customer may download an app, see some ads on social media, then make an in-app purchase.
• W-shaped. The same as above but with an additional defined touch point.
• Full path. Full path is a more evenly distributed customer journey that involves multiple major and minor touchpoints.
• Custom. There is nothing to say that an MTA model will fit these models, especially if there are many touchpoints (e.g. at enterprise level).

Google Analytics can handle basic attribution but Rockerbox and Wicked Reports enable a more granular investigation into attribution metrics. 

By creating reports on various campaigns, it becomes simple to examine the connections between campaigns and their associated behaviours. 

---

Want to improve your data skills?

See the best data engineering & data science books

Shop Now

3: Product Analytics

Product analytics tools are the mainstay of behavioural analytics and there are a handful of key actors and brands in the space, including:

• Mixpanel 
• Amplitude
  Rakam 
• Heap 
• Indicative 

These platforms focus on segmenting customers, analysing their journeys and interactions and understanding their decisions using funnel reports and cohort analysis. 

Once set up with a CDP or data warehouse, product analytics tools are pretty straightforward to operate, so long as you’re working with clean, good-quality data to start with. 

---

4: Business Intelligence

Business intelligence and product analysis tools crossover in many ways, but BI tools are perhaps less product-focused. There are many ways to analyze behavioural data for business intelligence purposes, such as social media analysis, sentiment analysis, anti-fraud and cybersecurity purposes, etc. 

Some key tools include

• Tableau 
• Domo
• SiSense 
• Looker

---

5: Qualitative Analysis

Not all data is easy to quantify. Qualitative data analysis inducts data from feedback, heatmaps, surveys and session information. This can provide a nuanced layer of insight into highly human actions that are not easily described with typical data. 

Some key names include: 

• FullStory 
• Mouseflow
• CrazyEgg
• Hotjar

---

The Anatomy of Behavioural Data

---

There are 3 main categories of variables to track and study when it comes to behavioural analysis:

• Events: Such as sign-ups or purchases. 
• Event properties: The product or action in question. 
• User properties: Names, addresses, other specific information about the user. May also include their device, browser, ISP provider and numerous other human and non-human properties and identifiers. 

These data points assess the user journey from intent and action through to the characteristics of the event and how this tallies with measurable human variables, whether that’s age, gender, location, order history, search history, etc. 

Prior to strategizing a behavioural data analysis plan, you’ll need to make a tracking plan. This will establish the KPIs you want to measure and ultimately influence. 

Some common problems you might want to provide answers for include:

• Optimising customer acquisition via the analysis of channels and campaigns. This will involve segmenting customers and measuring how they interact with different channels, e.g. discovering that buyers acquired via social media are of a certain demographic and respond better to advertisements optimised to this demographic.
  
• Increasing customer lifetime value – LTV. This will involve honing in on the shared characteristics of loyal customers, learning how and why they’re motivated to stay with your brand so you can obtain higher LTV across different demographic segments.
  
• Maximising retention and reducing churn, by discovering when and how to target promotions and incentives. 

Of course, behavioural data is by no means curtailed to commercial industries. Behavioural data plays a fundamental role in all manner of scientific study, particularly in psychology, neurochemistry, neurology and other biological, geographical and life sciences. 

Apps like Spotify use behavioural data to show recommendations of different music depending on the user’s search history, mood, time of day, etc. 

This doesn’t necessarily sell the product in a rudimental sense, but it allows Spotify to create value for the user. This obviously benefits the business and the user. 

Behavioural data analysis, therefore, involves cross-pollination between behaviours and interaction – it enables businesses to feedback in a way that the user appreciates. 

Modern behavioural data platforms allow for granular monitoring, enabling teams to quickly alter marketing material on the fly, to give one example. 

In essence, the purpose of behavioural data analysis and monitoring is a simple one. It resolves the question of “why do humans interact with X in a certain way”. 

---

Issues with Behavioural Data Analysis

Behavioural data is no commodity, it pertains to people’s individual and at times personal lives. Privacy is a major issue here, as is discrimination which can become a real problem with static behavioural modelling.

For example, whilst car insurers may be able to obtain a large dataset that attributes higher risk to young men aged between 18-25, thus causing them to raise their premiums for this demographic, a static dataset here fails to address the nuances that underlie such a pattern. 

Dynamic behavioural data analysis enables businesses and organisations to tailor this sort of risk assessment to a wider range of dynamic traits and behaviours rather than the static and potentially discriminatory demographic of young men. 

It enables insurers to react when the risk posed by young men, in this example, is not tangibly inflated to say, adults living in inner-city areas. 

• Behavioural data analysis, when conducted properly, can expose wrongful assumptions – assumptions that may also pose legal or regulatory risks.
  
• This also ties in with the issue of privacy which goes right to the core of behavioural data analysis. Digital fingerprinting provides a potent backdrop for debate on how behavioural data is readily usable for more nefarious or invasive purposes. 

The key here is to be extremely cautious when evaluating certain human characteristics – the user properties. Whilst some level of characterisation may seem reasonable, delving too deeply into everything from sexual orientation, race and even gender, age and location, can pose ethical problems. 

In many ways, behavioural data is becoming too good. For most businesses looking for simple ways to increase acquisition and conversion whilst lowering churn, this may not seem like a big issue. 

However, for businesses that have maximal access to data, using personal data in questionable ways is not likely to remain totally unmitigated, both in law, regulation and public mood. 

---

5 Steps Towards Your First Behavioural Analysis Strategy 

Here are 5 typical steps to running your first behavioural data analysis project or strategy:

1. Select Important KPIs and Metrics

Firstly, choose goals and examine what KPIs are associated with those goals. 
The goal might be simple:

• Increase customer acquisition 

However, this is not so much a goal as it is a general business objective. Broken down into a goal for behavioural analytics, this might be something more like:

• What channels acquire the most customers, and what do these customers have in common?

Here, you have a measurement – the channels that acquire the most customers – and a feature that questions the behaviours behind that measurement.

2. Define the Different User Journeys

Each user journey should be defined. You’ll need to discover and track journeys via each channel, segmenting customers that interact with your brand via email, social media posts, organic SEO, ads, etc. 

This will help you locate the most effective channel for acquiring customers, allowing you to drill down into who those customers are. You’ll then be able to compare this data to other KPIs, like LTV and churn.

3. Develop a Tracking Plan

Your tracking plan requires both events, event properties and user properties. The events might be simple, like clicking on a signup box or link. The event properties might include the end result of this interaction; whether or not they continued with the action and ended up signing up, making a purchase, etc. 

Tracking plans often include temporal journeys, particularly in the case of subscriptions or other forms of engagement that evolve throughout time. 

Users, events and properties will have to be broken down, named and organised to begin tracking. 

4. Create Identifiers

Users and/or groups of users will need to be identified using random strings or names. 

5. Begin Tracking and Start Analysis 

Your tracking period could be ongoing or it could have a defined length, e.g. if you want to measure the effectiveness of a campaign. Once you’ve collected enough data, you’ll be able to view how your KPIs are influenced, by whom and hopefully, why. 

---

What Next: Using Behavioural Data

What you decide to do next will depend on your findings. There are many strategies you can implement to captivate more engagement, drive conversions, increase LTV and reduce customer churn. 

For example, if you run a subscription box business then your data might show that you tend to lose customers 2 to 3 days before the end-of-month payday each month. You could target promos for this time to reduce churn and help ensure customers don’t terminate their subs. 

You might also find that certain social media campaigns drive conversions across certain demographics. This enables you to look into what these demographics have in common, which often turns into a qualitative analysis – what are they discussing on your social media posts? What do they like/dislike about your content from a qualitative perspective?

If you’re optimising a recommendations model, you can use behavioural data to target different product recommendations at their optimal demographics. For example, if it’s getting towards summer then buyers who regularly dwell in the outdoors section of the website might be easily converted into sales with some promos. 

---

Summary: Behavioural Data Analysis

Behavioural data is a powerful tool that can expose links between actions, interactions, engagement, intent and their end results. 

Whilst behavioural data can be very general or wide, looking at the effectiveness of a broad campaign, it can also delve down into increasingly nuanced user and event properties. As always, there are pros, cons, benefits and limitations to using behavioural data. 

Businesses and organisations must be careful to not over-extrapolate or over-assert conclusions from their behavioural data. If any data is being used to train systems, e.g. ML models, then extreme caution should be exercised over the user properties to avoid discrimination and other forms of prejudice. 

FAQ

---

---