Why is analytics important?
On the face of it, the simple question of why data itself is important seems to have a simple answer. Data informs us. Data tells us what is working and what is not working. In short, data trumps opinion – it tells us when we are wrong. The problem is that in reality it doesn’t. The nice soundbites about data trumping opinion are wrong.
The problem isn’t the data, though, it’s all the humans in the way. Humans have opinions, and opinions are emotionally developed and held firmly. People’s opinions are rarely swayed by being confronted with the truth, as recent political polarisation illustrates. In fact, human nature is to look for other explanations as to why data doesn’t agree with a deeply held opinion more than to change that opinion. This is why many religious people reject overwhelming evidence supporting the theory of evolution. This is the “devil put dinosaurs here” effect.
The problem is that numbers are dull. Stories are exciting. Stories give people something to believe in. Once someone believes the story, it becomes difficult to change their opinion. Almost no one recognises this human trait in themselves, but it is true of everyone. People’s beliefs about the nature of themselves are implicit and therefore difficult to self-review. This is why it is so difficult to persuade people to change their opinion.
A good web analytics consultant controls the narrative, writes the stories and helps people build their own new opinions based on the truth. Analytics is where storytelling marries data science. Analytics is not just data. Analytics is insight.
First, what is analytics?
Analytics is a catch-all term that covers a multitude of disciplines, not all of which are digital. It roughly breaks down into two categories – data analysis (i.e. a view of the past), and predictive analytics (i.e. an estimate of the future). Leaving aside some aspects of data science and analytics, such as operational optimisation, signal processing, image and speech recognition and symbolic reasoning, we will focus on business analytics itself.
Business analytics crosses over with data science, but is not entirely encapsulated by it. Business analytics includes qualitative aspects such as business case studies and qualitative surveys, which are not really data science (although machine learning may change this in the coming years). Qualitative business analytics often focuses on users’ needs and experiences, and is a discipline with closer ties to UX.
Where business analytics does meet data science, though, there is a great range of potential activities. The most basic of these is simple data reporting – i.e. being able to access raw data collected about a business or service. By processing this data, we enter the sub-field of Business Intelligence, which includes data visualisation, dashboards, and insights. Layering machine learning onto business intelligence adds the potential for real-time insights and dashboards.
All of these are looking at past data (i.e. analysis). If we take the raw data we can also undertake forecasting (predictive analytics) from this, either manually or using machine learning. Machine learning can also be used to assess real-time data to prevent fraud, aid client retention, or automate personalised marketing activities.
Much of the data used in modern business analytics is web analytics data, often Google Analytics although there are other similar tools. However, many data sources contribute to true business analytics, including CRM, sales, social media, and many other tools.
Understanding the distinctions between these different aspects of business analytics is important because these build the foundations for an organisation’s data maturity, and such maturity is critical in making the transition from having data to having stories. That is to say, from having the truth, to get people to listen to the truth. The devil put the data here – we need to create the environment that we need to be able to tell its story.
Principles of a good web analytics strategy
There are many ways to skin a cat, although many will leave you covered in scratches! At Storm we align our first principles with the UK’s ‘Government Digital Services’, and Scotland’s ‘Digital First Service Standard‘. These both describe four key measurements that apply to any digital service:
- Digital take-up – i.e. how many service users engage with the digital service as opposed to an alternative (e.g. call centre, on foot, etc.)
- User satisfaction – how happy are users with the digital service, what frustrations do they have, etc.
- Completion rate – what ratio of users that start to use a service get to the endpoint
- Cost per transaction – what is the cost to get a user to the endpoint of the service (and how does this compare to users on an alternative such as call centre)
While these KPIs are derived with public services in mind, they translate well to most applications including commercial ones. Commercial services might also add Revenue as a KPI. It is important to consider that the KPIs themselves are not simple metrics, but are the headlines of the data stories, and are the product of many data points from many sources.
Taking cost per transaction as an example, to calculate this, and have important context, we must know:
- What is the total cost to deliver the service, including hosting, marketing, development, staffing, etc.
- How many transactions have been serviced
- What is the total cost to deliver the service in other channels (e.g. call centre), and how many transactions have been serviced through those channels
This means that if we know it has cost, for example, £30 to support each digital transaction and £70 to support each call-centre transaction, then we can justify increasing support for the digital service. Additionally, we can propose hypotheses to take actions that reduce the cost, and because it is measured in context, we can see the impact of continuous improvement of the service.
There is no more convincing a story than “each time we do X, Y happens”, and a good web analytics strategy enables this storytelling approach. Web analytics, then, brings value not simply through enabling the collection or reporting of performance data, but through the context with which that data is delivered. By delivering compelling and irrefutable data stories to an organisation, web analytics is able to steer decision making based on analysis of the truth rather than unfounded opinions.
What skills are required from a web analytics consultant?
There are a number of different consultancy roles that are related to web analytics, each of which brings unique value and skills.
Roles in traditional and big data management
Traditional data and big data differ only in complexity. Big data does not mean high volume data, but complex and unstructured data, although by its nature it typically has a greater volume than traditional structured data. Both traditional and big data require management through data architects and engineers, and traditional data also requires database administrators.
These roles help to ensure data is collected accurately and that pre-processing, such as data cleansing and class labelling, is done effectively. They build the datasets on which analysis can be undertaken. For example, in relation to Google Analytics, this includes management of tracking code, the architecture of account structure, and set up of Google Tag Manager, etc. For other data types, skills might include the use of SQL, R or Python, or tools like Google BigQuery or Microsoft Azure.
Roles in business intelligence
Business intelligence refers to the analysis of data in the past, and to present it in the form of metrics, KPIs, reports and dashboards. Roles in business intelligence include BI analysts, BI consultants, and BI developers. These roles are concerned with extracting and telling data stories from the data collected, helping organisations to optimise services and pricing, manage inventory, and manage costs.
In relation to Google Analytics, basic skills include the ability to interpret the reports and draw insights from them. Advanced users will also export and undertake further analysis on the data using tools such as Microsoft Excel, and create visualisations using tools like Google Data Studio, Tableau and Power BI. For other data types, analysts may rely on SQL, R, Matlab and Python, as well as those tools mentioned above.
Roles in predictive analytics
Predictive analytics is the process of making predictions based on data. These roles include data scientists, data analysts, and machine learning engineers. They help organisations forecast future performance, detect fraud, retain clients and undertake informed continuous improvement.
Currently, Google Analytics does not make any real attempt to predict future performance, and there is no direct tool to help an analyst make simple predictions using Google Analytics. Data needs to be exported and processed using statistical techniques such as regressions analysis, clustering and factor analysis in order to identify behaviours that are predictable, and therefore make predictions for the future.
How does a consultancy align with your objectives?
Using a web analytics consultancy service helps an organisation utilise the many varied roles and skills that are necessary for a fully featured analytics team where these don’t all exist within the organisation. Additionally, consultancies bring a wealth of experience on implementing data strategies, extracting and analysing data, visualising it, and storytelling to help the organisation extract the real value from data.
While day-to-day analytics practitioners have the technical skills to analyse data, a consultant also brings the soft skills to turn that analysis into presentable data stories.
A consultant understands the needs of the digital team and develops innovative solutions that tackle their problems head-on. A consultant capitalises on an organisation’s existing datasets and knowledge base, while recognising new business opportunities for the organisation.
A consultant can also help restructure problems from ambiguous and complex to structured and clear, and will help empower a digital team to make confident predictions and report accurately.