Intel’s Chief Data Scientist, Bob Rogers, outlines the common misconceptions held about advanced analytics by business users and decision makers and explains why they are untrue.
I spend a lot of time speaking to Intel customers and others in the IT industry about analytics, and I’m often surprised about some of the misapprehensions out there. Here are six of the most common ‘myths’ I come across.
MYTH #1: “To deploy advanced analytics like machine learning (ML), we have to invest a lot in exotic new hardware and dedicated infrastructure.”
TRUTH: Your first advanced analytics project is actually a lot closer than you think.
A lot of companies already have much of the infrastructure in place for advanced analytics. For example, it’s common to have thousands of sensors on a factory floor, but often not all of them are used as the data they generate is so noisy. A good starting point then is to focus on ensuring you’re able to manage, store, and integrate all your data.
If you don’t have extensive existing hardware that could be used for analytics, don’t worry. A number of cloud service providers (CSPs) provide ready-made advanced analytics capabilities. Use these resources to run some trial projects and work out the most effective use cases for you – then you can build a roadmap for a long-term investment with a clearer idea of what the return will be.
If you choose to run things in-house, you can use your existing Intel® technology-based platforms to help you scale and flex to support your new projects securely and cost-effectively.
MYTH #2: “We don’t need advanced analytics like machine learning for our business.”
TRUTH: No matter its size or complexity, every business needs advanced analytics.
Often when I speak to a customer about analytics, they’ll say they’re already doing it. “Fantastic,” I respond, “tell me more!” They may then begin telling me about how they’re using spreadsheets or simple online systems to report on events that have already happened.
This is a great first step in the analytics journey. You need to know what data you have and have it organized well before you can analyze it. However, as analytics capabilities evolve, business intelligence that worked well a few years ago begins to look a bit dated.
To keep up with the competition, it’s important to understand real-time trends in your data and even use today’s data to predict and influence what will happen tomorrow. This is where advanced analytics comes in. With ML and other artificial intelligence (AI) techniques, you can understand more types and greater volumes of data more quickly. For example, AI can help you automatically include information from text and images in your business processes. That understanding can empower your key decision makers to do their jobs better. As these capabilities become more mainstream, those who don’t embrace them risk putting themselves at a competitive disadvantage.
MYTH #3: “We have a lot of data but analytics doesn’t work.”
TRUTH: It will, but clean data is essential.
A common complaint from organizations running their first advanced analytics projects is that despite having mountains of data, they’re not getting the insight they expected. Or they’re actually getting wrong or inaccurate results.
Unfortunately, in a lot of cases, it’s easier to write it off than address the root of the problem, which is frequently data quality. It’s essential to ensure you have an effective way of collecting all your data together, cleaning it to remove duplications and inaccuracies, and keeping it up-to-date moving forward. AI and ML algorithms can only work with the data they’re given, so low-quality data means low-quality results. However, get your data management policies right, and insights will come.
MYTH #4: “We will need to hire a bunch of data scientists.”
TRUTH: You’ve already got them.
As a data scientist myself, I perhaps shouldn’t be saying this, but not every company needs a resident Ph.D. data scientist. Understanding of the business needs and the nature of the data available to decision-makers are critical. It can be tricky to fill data science positions, especially in industries where specific product or subject matter expertise is also essential, so with some additional tools and training, you may already have the experts you need.
An alternative approach is to put the tools for getting analytics-driven insights into the hands of the people that run the business. Cloud-based services can clean and process your data, and provide the algorithms and dashboards you need to create and present insights. This enables you to transform your business experts into citizen data scientists.
MYTH #5: “Measuring everything is good.”
TRUTH: Measuring everything is impossible and unnecessary. Measuring what’s strategic is essential.
Once you start collecting and integrating your data, there’s a temptation to use all of it for everything. However, this approach can become unwieldy and expensive, and including a lot of irrelevant information can even have a detrimental effect on your results. It’s therefore important to continually ask yourself what you’re measuring or analyzing a certain data set for. If there’s no valid business reason to measure something, you can focus elsewhere.
Whenever one of our customers begins an advanced analytics initiative, we advise them to start with the business need. What strategic challenge are you addressing? Once you know this, you can set measurable objectives to use as your guide for what to measure and how to recognize success.
MYTH #6: “Advanced analytics means we need to hire additional IT staff.”
TRUTH: Your existing team is ready for the challenge.
If you’ve already got an IT team in place, you can make some good headway on your analytics journey. As with your IT hardware and your data science skills, initially, it’s more about making the most of what you already have. Chances are you’ve got talented IT experts who would relish the opportunity to lead the way in your organization’s analytics innovation.
Take stock of the skills and expertise you already have in-house, and work with external advisors like industry analysts or your trusted vendors to get their advice and best practices, and to identify any immediate needs for training. If you choose to develop algorithms in-house, you can make them available to others within the organization through a Function-as-a-Service (FaaS) model to help minimize the development resource you’ll need long-term too.
It's clear there are still a lot of questions around deploying advanced analytics, but they shouldn’t be a roadblock to finding value in your data. Discover more about the truth behind analytics myths in this video from Intel and Fero Labs.