November 5, 2021
Data analytics has long been an integral part of the life of any business with a digital presence. It is so indispensable to us that we forget that it is used not only for products and their sales but also for social studies, medicine and all kinds of scientific research.
Today, however, we are going to look at the very ordinary analytics: the business-related ones. We will go over the most common types of data analytics platforms and their applications.
According to Investopedia (why not?), business intelligence is the procedural and technical infrastructure that collects, stores, and analyzes the data generated by a company’s activities.
It collects data from all departments (marketing, logistics, production, accounting) and makes it homogeneous and fit for business analytics. Business analytics is what the human mind has to do with this data (not without digital means, of course). The goal of business analytics is to assess the current situation and make the appropriate changes.
And, again, data analytics is the umbrella term for everything BI (but BA can be done without Big Data, although that’s rare these days)
First, the data “happens” where it naturally resides. Then, you add data sources such as:Product database, CRM, various marketing platforms, custom data warehouses, anything from Microsoft Azure to a CSV file.
All data is merged, cleansed, re-collected and sent to a single storage (internal or external).
BI uses algorithms and automation to identify regularities and trends. These are typically descriptive, diagnostic, predictive, and prescriptive types of analysis.
This is where the user sees the data in the form of graphs, charts, screen displays, etc., organized in dashboards. This is where the obvious magic happens and where data analysis is accessible to all users.
There are platforms that require coding to reach this level, as well as zero-code tools, but the result must be
Usually applicable: Today, data is stored in places other than where it is processed, but that could change in the future.
Here is the tricky part: most BI platforms do not do all of these steps at once and focus on some of the functionality, partly because the processes described require different database structures and it’s more efficient to focus on an average of 3 steps.
Note that you may see different combinations of these steps, such as Extract – Transform – Load – Visualize or Pull – Save – Visualize, abbreviated as ETLV, ETL, PSV and so on. You do not need to memorize these steps immediately, but should get a basic understanding of their meaning/application.
Most platforms provide you with a combination of such features (from Tableau’s point of view, and we agree):
One of the goals of BI is to make business analytics accessible to people from different departments and with different levels of technical expertise. This has been achieved to a certain extent, but using the information presented is still a major problem.
Further development of the field is heavily focused on AI predictions and recommendations, as humanity has perfected the previous processes.
What happens when we break all-in BI into different business use cases?
Let us look at the most common types of data analytics platforms, starting with the simplest.
The most common definition of analytics that almost every business owner knows. The unit of analysis is solely the website. The basic functions of data analytics platforms are:
Modern website analytics platforms extend their functionality to provide a 360° experience focusing on specific types of features.
In addition to Google Analytics, some of the most popular services include: Matomo (an open-source and privacy-friendly counterpart to GA, which even has a tag manager)
Mixpanel (focusing on user behavior), Adobe Analytics (customer journey and ROMI), GoSquared (minimalistic, but with many integrations), Clicky (marketing and SEO).
The main focus of product analytics is the user experience: from the website visit to the information on how your product is later used. This kind of information allows you to make product hypotheses, add new features, find and eliminate bottlenecks. It is said that product analysis has three levels: the big picture, specific funnels and individual user journeys, and all of them usually require your attention.
Why PA? You define the look and feel, product-user interactions and their number, as well as the development roadmap, advertising budget, and advertising goals by understanding what works better based on key marketing and user metrics. You can also say you are answering 2 questions: “What” (what are users doing with your product) and “Why” (why are they doing it in the first place).
It can be compared to Business Analytics (we learn how the product feels, as opposed to how the entire company feels), with one key difference: BA is based on Big Data and strategic solutions, PA is almost entirely real-time: you evaluate the current, short-term data, make suggestions, test or A/B them. We base suggestions and hypotheses on debugging, and you need to know that proving them wrong is a much more valuable insight than verifying that something works.
How do I set it up? This includes a measurement plan with key KPIs, metrics and their values to track how well the KPIs are being met. A marketing specialist may not be enough for this purpose because defining the level of a target event can be tricky.
Ideally, you need a balanced team for the aforementioned task to avoid overkill: Technicians can either track a user’s every finger movement or vice versa (track only the biggest things), but what you should avoid at all costs is exaggeration. If you overdo it, you run the risk of ending up spending too much time and effort setting it up and feeling embarrassed about not using the valuable data. Measure what’s really important.
Most popular product analytics tools are paid. If we speak of the most popular product analytics platforms, they include Amplitude, Mixpanel, Adobe Analytics, Pendo, Heap, and most of them are paid if you want to use the unabridged version, but free basic tools like Google Analytics also work.
Do I really need product analytics? If you have a SaaS product and a team of developers, you can try to hire the product analyst along with or instead of the platforms, but usually the cost of hiring a new employee and using the existing resources is even higher than the platforms. Do not forget that the platforms also need to be set up, and even the best support team can not implement analytics for you.
Mobile analytics is considered part of product analytics, but it has evolved to the point where several types of analytics platforms have been developed: App Marketing Analytics, Product Analytics, and Store Analytics. The latter is unique to the type of product.
Mobile analytics provides insights into user behavior and preferences, as well as user acquisition best practices, but the incredible rivalry of apps contributes to the fact that another vendor’s scenarios may not be good enough for your product and you’ll always have to use your intuition to make the right decisions. The fact that the apps may be cross-platform is also a pain: what works well for IOS does not necessarily work for Android, and vice versa.
Mobile analytics is an absolute must if you want to reach an impressive user base and make money with your app. As with product analytics, there are free tools like Google Firebase, but for marketing analytics, you need marketing-focused solutions like AppsFlyer, Adjust, Kocheava, etc., as well as the aforementioned Amplitude and Mixpanel. Tracking your competitors also costs something. This is where platforms like AppAnnie come in handy.
If you have website sales and mobile products as well as an offline presence, cross-channel platforms might be the way to go. Cross-platforms bring all your data sources together in one place (including offline sales, call center data, logistics, etc.), but we will cover that in our future articles.
Its goal is to make predictions and create hypotheses that can be tested against historical data that is as big as possible in your situation.
If we compare predictive analytics and business analytics, BA uses historical data and requires you to draw conclusions manually and with your own mind.
Predictive analytics use algorithms and machine learning to model the future.
To build an adequate predictive analytics model, you need two types of data, internal business data (the entire item assortment, your expenses and revenue, the number of existing and historical customers) and external data (like pandemic statistics, crypto market data, exchange rates, and even the weather). This somehow makes predictive analytics more powerful and efficient than business analytics, if done right, of course.
Predictive analytics does not always work well even in the oldest and most classic use cases, like predicting oil prices and currency rates and not accounting for force majeure, but they are slightly different from product analytics when the external factors are less influential.
You can build your own predictive analytics if you have a strong enough development team and it “fits the macros of your business”. However, it might be cheaper or more efficient to use purpose-built products with years of experience, like BigQuery, RapidMiner, or Oracle Big Data Preparation.
There are other types of enterprise analytics that we have not covered in this article. You can see the full picture in the image below.
You probably need at least one analytics platform to start with, whether it’s a free or an enterprise solution.
However, we suggest that as your product grows, you find points of that growth by continually implementing the various new analytics platforms and evaluating your business needs each time you reach a plateau or plan a massive expansion.
Businesses rarely settle for just one platform, as there are no platforms that can achieve all goals and solve all problems at once. We are confident that you will find your perfect product combination. And if you need even more insight, follow our blog: We will be covering many more types of analytics platforms and their mappings soon!