The decreasing cost of data storage and the arrival of new cloud-based technologies have lowered the barrier to entry for big data projects. Many companies are jumping in, often with the same goal: develop an in-depth understanding of customers to better anticipate and address their needs. And when it comes to optimising the customer experience, web analytics is leading source of data used, according to a 2015 study from Econsultancy.
In this interview, we’ll hear from Marion Joffre, product manager and specialist in all things big data at AT Internet. She shares insights on analytics challenges and tools as they relate to big data projects and discusses the central role of digital analytics in the data ecosystem.
Why is digital analytics data essential in big data projects?
Digital analytics data plays a significant part in these types of projects, firstly because digital analytics data offers a detailed view of everything a customer might do on a website or mobile app. This is exclusive information about how visitors are using a site or app, and there are extremely rich learnings to be extracted. Digital analytics data provides a super-detailed view of user behaviours and browsing patterns – down to each individual visitor – no matter the device used to browse. All this information is available via exhaustive data from digital analytics tools that don’t use sampling, such as our Analytics Suite.
Another reason digital analytics data is so crucial for big data projects is its reliability: Digital analytics data is the product of both highly specific calculation rules and visitor identification, enabled by user login. As a result, digital analytics data is consistent over time. It’s therefore possible to retrace a user’s journey across several devices over a specific period, if that person has logged in.
Lastly, digital analytics data enables us to answer the following questions: What did my visitors do before buying one of my products? Which content or campaigns led them to finalise the purchase process? Prior to making a purchase, how frequently did my visitors come back to my site?
Can you start a data science project if you only have digital analytics data?
Yes! Digital analytics data alone offers substantial potential for rich analyses within the framework of any big data project: unifying the customer journey as we just discussed, establishing visitor profiles by engagement levels, creating custom attribution models, and more… According to a 2015 study from Econsultancy, web analytics is the number one source of data used to optimise the customer experience:
These results support the idea that digital analytics has a key role to play in unifying the customer journey and improving the user experience.
But beyond web analytics data, the whole idea of big data projects is to unify different tools and data sources to identify new learnings that we wouldn’t be able to find otherwise.
I’m thinking specifically of CRM data, which contains a goldmine of information. If we combine CRM data with digital analytics data, we can see everything that has happened up to conversion, as well as after! With this kind of data combination, the possibilities are endless: customer and prospect scoring, segmenting down to the individual level, more precise targeting, and better personalisation.
Which sources of digital analytics data should be used?
If you’ve adopted a “test-and-learn” approach, you can advance step by step and validate hypotheses as you go. With this approach, you can make quick adjustments in case of any discrepancies. A good starting point is analysing visits per visitor, with details of sources, devices used, and the types of actions taken during each visit. This kind of analysis will provide a view that’s already very rich in learnings.
Once intuitions have been confirmed, you can proceed with a standardisation process. But it’s risky to want to move too quickly. A study from Forrester shows that most companies use only 12% of collected data. It’s therefore not necessary to try to retrieve everything from the start – it’s a better idea to advance progressively.
Beyond the type of data to be used, it’s also important to think about the format. We often hear about raw data, but in digital analytics, raw data is simply the hit sent upon each action being tracked by the analytics solution. At this point in time, the hits have not yet been processed, and data has not yet been enriched – information is missing about the visit, the visitor, the device used, and geolocalisation, for example. This data is therefore not directly exploitable for big data projects, because crucial information is still missing. In this type of project, the necessity really lies in having access to extremely detailed data which has not been aggregated, so that it can be reprocessed, and so that specific calculation rules can be applied.
Do you have any examples or uses cases to share?
Across all sectors of business activity, we observe the need to recreate customer journeys… whether it’s in e-commerce, to understand how visitors convert using extremely deep analysis of navigation, or in the media sector, to understand the correlations between different types of consumed content.
The same goes for the banking sector, given the high volumes of logged-in users we see in this sector – there are fantastic opportunities for unifying the customer journey across all possible channels of interaction (physical branches, call centres, mobile apps, etc.). It’s therefore now possible to know for any banking product purchased, what the upstream interactions were, via which channels and at which frequency, in order to continuously optimise the marketing actions taken on future prospects for this same product.
Which skills are essential to have for people exploiting this data?
If we’re talking about data scientists, in theory, it’s a very well-rounded profile which must combine several very specific skillsets. In addition to (necessary!) technical knowledge, a data scientist can only do his or her job correctly if he or she has a clear and precise view of the business’ challenges. The data scientist should also be able to communicate widely within the company to evangelise internally and help others understand the added value of his or her role. The data scientist must be able to find simple, clear ways of explaining phenomena that are sometimes very complex. Without these skills, it will be difficult to truly exploit the data, and the project will likely not be as profitable.
Nonetheless, we must keep in mind that these new sorts of profiles don’t necessarily originate in the digital sector, and are therefore not necessarily familiar with all the details of calculation rules specific to digital analytics, which – as we know – can be numerous, and very specific. It’s therefore a good idea to seek expert guidance, in order to ensure the quality and accuracy of the properties used in your various queries.
How can digital analytics providers like AT Internet help analysts and data scientists with these kinds of big data projects?
This past summer, we launched a new API for data extraction (Data Flow) designed to support these kinds of projects. This tool combines a powerful, flexible API with a very intuitive interface for flow configuration, so our users can easily create and manage data export flows themselves. Data can be extracted from the system on an hourly basis, which means our users can retrieve events practically in real time.
Also, our digital analytics consultants are true specialists regarding data usage and interpretation. They’re experienced in advising our customers on how to best use AT Internet digital analytics data and can help ensure correct data manipulation.