9 July 2019
Advanced Analytics is a next-generation set of analytical methods that use sophisticated modelling techniques and mathematical algorithms to uncover deeper insights, help organisations identify patterns, predict trends and drive change with fact-based solutions. It gives enterprises the ability to extract refined, detailed answers to such complex questions as “why is this happening?”, “what will happen if this trend continues?”, “what’s next?” or “what if…”.
While advanced analytics is about optimising and predicting the best next action, business intelligence is focused on metrics-driven reporting, dashboarding and ad-hoc querying (typically classified as descriptive analytics). It uses historical and current data to help organisations access more effective strategic, tactical and operational insights and decision-making. The main difference between business intelligence and advanced analytics is that the former is focused on the past (what happened?) and the latter strives to predict the future (what’s next?).
The applications of advanced analytics are wide and varied, ranging from classic sales and demand forecasts to predictive policing and customer success optimisation. To give you a quick example, here are a few common business intelligence vs. advanced analytics scenarios.
Companies that decide to invest in and try out advanced analytics often aim to improve their internal processes, such as risk or customer relationship management, develop new products or services, better target their offering to their customers or improve their existing products.
The opportunity to tap into advanced analytics gives organisations the ability to make more predictive and real-time decisions, often resulting in improved, more cost-effective performance. According to research, 58% of data-driven organisations that implement predictive analytics “see a correlation between analytics initiatives and a ‘significant improvement’ in their competitive positions”.
If this makes you wonder what’s stopping data-driven enterprises from unleashing the power of advanced analytics, the answer is in the numbers – 59% of companies are not using predictive models or advanced analytics “due to lack of analytics capabilities and skills”, a Forbes study reveals.
So, despite the increasing dependency on data and data-driven solutions, enterprises still lack skilled in-house workers to make sense of all that data. On top of that, many organisations cite lack of budget as another major blocker to their analytics efforts. And even those organisations that are making strides in capitalising on business data and are successfully deploying analytics strategies, don’t always have the in-house capacity to wield data science.
Qlik also reports that only 24% of decision-makers consider themselves truly data-literate, leaving a lot of room for growth.
But as markets evolve at breakneck speed and various technological disruptions continue to impact almost all industries, having a competitive advantage provided by advanced analytics is what determines industry winners and losers.
Is there a solution in sight? Enter Vizlib’s advanced analytics for Qlik Sense that combine the power of Qlik’s associative engine with sophisticated data science techniques in a few clicks.
As we start rolling out advanced analytics capabilities for our Qlik Sense extensions, the focus is entirely on engineering the ease of use. If you’ve tried advanced analytics with R and Python in the past, you know how complex and hacky the experience can be. It’s definitely not for everyone. Our goal, on the other hand, is to unleash the power of advanced analytics by making such capabilities super accessible and simple to use, for anyone. This means that, from now on, the end user will be able to reap the benefits of sophisticated algorithms simply by dragging and dropping our extensions into their Qlik Sense apps to leverage advanced analytics on the fly. (You can tell that we’re gloating like proud parents, right?)
Now, let’s talk about how.
Our built-in forecasting capability for the Vizlib Line Chart will enable you to forecast trends and make predictions in time series based on historical and near-real-time data. It decomposes the time series into 4 constituent parts: level, trend seasonality and noise. Then, a triple exponential smoothing algorithm (Holt-Winters approach) is used to calculate the forecast on the given time series components and parameters.
The provided algorithm uses a model-autofitting to optimize the forecast by minimizing the prediction error (the difference between actual and predicted values). As a user, you can choose the level of optimization (from economic to bazooka), which leads to a more accurate result by increasing the calculation effort. Additionally, it calculates the prediction interval using a bootstrap method.
Both, the predicted values and the prediction interval are added to the Line Chart at the end, extending your time series into the future.
With Linear regression forecasting leaders can predict future business objectively by analysing variables and forecasting the potential outcomes, or ‘best guess’, to guide decision-making.
Predicting a ‘best guess’ case with Linear regression forecasting in Vizlib Line Chart
Scenario analysis enables businesses to evaluate multiple future scenarios to analyse potential risks, and manage and plan for any unexpected change. The best solution to support big decisions using different assumptions, especially post-pandemic when ML provides lesser quality data.
Scenario Analysis with Vizlib’s Advanced Analytics for more accurate planning and decision-making
Next is our newly added clustering capability in the Vizlib Scatter Chart. But first, let’s establish what we mean by clustering (get your geek on).
Clustering is a task that involves grouping data points into categories by analysing their similarities. Simply put, the aim of clustering techniques, like K-means that we’re using in our Scatter Chart, is to distinguish groups with similar features and assign them into relevant clusters. Once the algorithm finds groups in the data, it assigns each cluster a representative variable K. Working iteratively, the algorithm continues to assign each data point to its nearest K variable, also known as the “centre”, until all clusters are established.
Discover more on Clustering in our Advanced Analytics: Clustering Guide.
Google News uses clustering to group news articles into different categories based on their title and content. And among many other applications, we could use the clustering technique to group search results or customers and detect anomalies. Deploying the K-means algorithm on Uber’s trip data based on the longitude and latitude would allow us to identify the pickup hotspots in any location. That’s pretty useful, eh?
Since K-means (a method of vector quantization) doesn’t fit all use cases, we have further added the following density-based clustering methods which group points that are closely packed together: DBSCAN and OPTICS.
Once clustering is calculated, scatter plot points are coloured by cluster and you can use the context menu to select clusters. This gives you a unique interactive analytics experience.
You can learn more about using advanced analytics, such as forecasting to manage unexpected change.
Keen to give these new capabilities a good spin? Download all our extensions from the User Portal.
Vizlib’s feature-rich data visualisation products will help you get started on your data storytelling journey.
Got a technical question? Our Help Centre has heaps of helpful guides, articles, FAQs and more. Access Help Centre