I help BU leaders of medium sized companies who are frustrated in their effort to leverage data into actionable knowledge.
I enable decision makers to extract insights from data, quickly setting up powerful bespoke dashboards.
I help companies effectively deploy advanced self-service business intelligence and predictive modelling tools.
I have over 25 years of business domain experience and advanced technical analytical skills.
This translates into a fast and results-driven approach, strongly rooted into each customer’s business logic.
I am part of the Kinetic Consulting team. Kinetic is a Strategy and Organizational consulting group with a consistent track record in delivering market place results.
Line managers often struggle to take advantage of the recent advances of business intelligence and predictive modelling tools.
One problem is the difficulty of specifying the business problem in a language actionable by IT.
The process is iterative: when IT eventually provides the answer, this often triggers further information requests by the business users, straining the capacity of IT to respond effectively.
The typical solution – to provide business users with self service business intelligence tools and secure access to the relevant data – only partially solves the problem: data needs to be prepared before being fed into the self service business intelligence tool.
This data preparation work requires technical skills and depends on the specifics of the business problem. It is iterative and often performed at the business user level. Typically it is done on Excel, with significant investment of time and risk of loss in data quality.
A similar issue arises with predictive modelling. While the power of this technology is rapidly improving, the challenge is to how to translate a business problem into predictive modelling terms and figure out how to match what the business wants with what can be delivered given the available data.
I help answer a specific business questions by quickly delivering advanced analytical dashboards. I leverage on my business experience and technical skills to bridge the business-IT gap.
These end-to-end dashboards incorporate the data preparation, data aggregation and data visualization workflow. They do not just deliver ‘one shot’ answers to specific problems, but can be used as ongoing data discovery tools.
I help decision makers incorporate unstructured data into the business analysis, using low cost, off the shelf, tagging services.
I also help implement an effective self-service business intelligence approach across the organization, by making sure the right questions are asked, the right requests are made to IT, and the correct approach to data preparation, aggregation and visualization is enacted.
Finally, armed with a robust understanding of the business, the issues and the data, I help the decision maker understand how and if to leverage predictive modelling tools to further support the decision making processes.
I am vendor independent. I use leading edge open source and proprietary tools such as the Python data analysis stack, D3, Tableau, and Trifacta.
Businesses that build network capacity into their core will outcompete their competition over the next 10-15 years.
Businesses that build data comprehension into their core will outcompete their competition over the next 10-15 years.
Whether you believe the hype or not is practically irrelevant: the shift is already happening.
The fear of losing competitive edge is pulling every organization into exploring big data technologies.
Getting advanced analytics right is not only about mastering the right tools.
It is about asking the right questions and identifying the decision making processes that can benefit from this technology.
The Googles, Facebooks, Twitters, Expedias of this world deal with really big data.
Most problems of most other businesses can be handled using relatively simple hw and sw tools.
Today's hw and sw tools make self-service business intelligence possible, often with little need to pre-aggregate data.
Effective self-service BI workflows increasingly need to accomodate self-service data preparation and the capability of making sense of feature-rich datasets.
Until recently, marketers had to rely in consolidators both to collect EPoS data from retailers and to pre-aggregate it.
In its granular form, retail data was just too big.
Today, marketers often can work with granular data, which can yield more powerful insights.
These datasets, huge in comparison with what could be processed in the past, often can be processed on single machines.
Yesterday's analytics were mainly about crunching internal data.
Today's analytics is increasingly about linking together external and internal data.
This means that data sources are also getting more complex: large number of columns, complex hierarchical structures, different datatypes, missing values that have to be managed in some way,...
A growing part, some say up to 80%, of data analysts's time is spent on "self-service" data preparation activities.
Robust and effective self-service data preparation is one of the keys for robust and effective self-service business intelligence.
Some solutions claim to be capable of delivering actionable insights out of the box.
This is rarely true: as Peter Thiel, the founder of PayPal, puts it:
Business intelligence tools will never have an 'improve sales button'.Image credit: Shyam Sankar
However, by minimizing friction at the interface between the decision maker and the data, they can help companies more easily use data to realize new business outcomes.
Best practice-driven data preparation is key.
While data sources are getting more complex, data analysis and visualization tools work with simple data structures.
In a self-service, exploratory, business intelligence setting, it is difficult to solve the data preparation issues centrally at the data warehouse level.
Doing it at the user level can be very time intensive without the right approach.
Excel is a great product, certainly the most popular data analysis platform in the world.
However, Excel is often used the wrong way, especially in data preparation tasks. Here the Excel workflow becomes prone to mishaps that typically can be traced to the following issues:
Migrating these tasks to a more effective and programatic workflow can help save time and money.
A key step in data preparation is understanding the structure of the data, so-called data discovery.
Data discovery wants to answer questions such as: Are there outliers? Are there null values? What is the "profile" of the data in each column? Does the data make sense?
One way of answering these questions is to build charts of the data distribution of each column.
Business intelligence has evolved immensely, in terms of volumes of data, level of interactivity and variety of visualizations.
Proprietary business intelligence tools such as Tableau and open source libraries such as D3 offer a stunning number of simple and esoteric chart types.
Different charts can be connected together, with relative ease, into powerful dashboards.
Data manipulation tools have become much more powerful, and allow for more agile development workflows.
Business intelligence projects still fail, mainly because users and data analysts do not manage to communicate effectively: users often lack technical knowledge and data scientists often do not understand the business requirements.
Examples of interactive dashboards and visualizations are shown in the slides below.
Hover over titles and menus to open info tooltips.
Click on the images below to see more business intelligence examples
Data mining is the practice of examining data to generate new information.
A data mining algorithm can be as simple as a linear classifier or as complex as a neural network.
Data mining (also known as machine learning) is being increasingly democratized, thanks to easy-to-use APIs and powerful libraries and end-user tools.
Data mining is the engine that drives Google's search engine. But it can also help any company improve the accuracy of its market predictions or better understand its customers.
The hype is about how the Amazons of this world automate an enourmous number of small decisions.
Less attention has been given to using data analytics to facilitate the less frequent but individually important decisions that lie at the core of most companies' business.
Natural language processing tools have improved significantly over the last 24 months.
They have become remarkably effective in identifying negative and positive sentiment, or at identifying the structure of a phrase in terms of keywords, verbs, subject and object.
This makes it possible to easily build simple, customized, applications able to process large volumes of text data such as emails, reviews, comments.
These applications can help the analyst shift through large volumes of data, identify trends and possible causes.
Examples of dashboards analysis customer revieas are shown in the slides below.
Hover over titles and menus to open info tooltips.
Click on the images below to see more text analysis examples
A key for the successful adoption of advanced analytics is the capacity of bridging the gap between users and technical staff, through a mix of business domain knowledge and technical expertise
It is important to cover all the steps of the worflow, from the careful definition of the user requirements, to the approach to data preparation, to the design and deployment of the dashboards.
The rapid development of data mining tools offers an amazing promise, conditional to the capacity of telling the hype apart from the concrete potential.
Python is an open source language, used by the likes of Google, Facebook, P&G and JPMorgan. It has a powerful ecosystem of data preparation, analysis and data mining libraries.
Tableau is a fast growing proprietary visualization package that allows for rapid prototyping.
D3 is a powerful open source visualization library developed at Stanford University and used by the New York Times.
Launched in 2015 by Joe Hellerstein of Berkeley and Jeffrey Heer of Stanford University, Trifacta is a data preparation service that helps transform data into a usable form for analysis.