Analytics
Quick definition: Analytics is the process of defining a method to transform data into action. It is determining the strategy a company will use to understand and take action on its data.
Key takeaways:
- Analytics is the strategy behind a company’s use of data
- The main types of analytics are descriptive, diagnostic, predictive, and cognitive. Each type develops into the next as a company’s use of analytics matures.
- Companies need to analyze and interpret data, not just collect it and let it sit.
The following information was provided during an interview with Nate Smith, group product marketing manager for Adobe Analytics.
What is analytics?
What is the difference between analytics and data analytics?
Why is analytics important?
What are the different types of analytics?
How do companies progress through the analytics phases?
What features are necessary for a good analytics solution?
How can companies successfully and effectively use analytics?
What are common mistakes companies make?
How will analytics change in the future?
What is analytics?
Analytics is the name given to the systematic, often automated analysis of data.
After data is gathered, it must be interpreted in order to be communicated intelligibly. That process is known as analytics.
What is the difference between analytics and data analytics?
Analytics is methodical. It’s strategic, and more focused on figuring out how to get answers. It’s the process of figuring out what data to collect, and then data analytics involves using technology to split things up and explore the information.
Analytics is also not just data collection — it’s the processing of that data to give it shape and meaning. And once you have shape and meaning, that's where data analytics comes in.
Why is analytics important?
At the end of the day, you need to have data-driven decision-making — that's the bottom line. Companies still have so many decisions that are based on gut or based on the intuitive nature of the business. But at the end of the day, if your decisions are based on data, you can’t argue with that.
Having your reality based on data is important. The importance of analytics is being able to have statistical or mathematical certainty in your decision-making.
What are the different types of analytics?
The main types of analytics are:
- Descriptive
- Diagnostic
- Predictive
- Prescriptive
- Cognitive
Descriptive analytics is pulling a report, getting information about user behavior.
The next phase is diagnostic, where you’re going in and manipulating the data to gain meaningful insights. You're trying to define or steer the direction of where that analysis should go to solve certain problems.
Descriptive and diagnostic analytics both give more of a rearview by looking at what happened in the past. Predictive analytics start to look forward and predict what is most likely to happen. Based on what has happened in the past, predictive analytics can predict certain types of behavior and engagement. For instance, you could see in the data that people from Los Angeles who come to your website through your Facebook page tend to buy a product. Then you can predict a lower cart abandonment rate from people from Los Angeles who come to your site through the Facebook page.
The point of predictive analytics is for organizations to be able to optimize their marketing dollars and activity. If you stick a dollar in the marketing machine, you want to have pretty high confidence that you're going to get five bucks back. With predictive analytics, your confidence increases because you know which machines are most likely to give you $5 back.
The next group is prescriptive analytics. With predictive, you’re essentially predicting something that is likely going to happen within a certain range. With prescriptive analytics, though, the system comes back and tells you what you should do to guarantee your result.
Let's say people are downloading your app. Prescriptive analytics is going to look at everything in play — the type of people, segments, traits, and behavior. Then it's going to come back and say, if a customer downloads your app, you should then send an email with a discount offer. Prescriptive analytics is prescribing, based on data, the next step for the marketer to take to move that customer to conversion.
Cognitive analytics takes that one step further. Instead of you defining a lot of the rules, the system is dynamic and self-educating. It will automatically know to send off an email with a discount, if it learns that's the next prescriptive step.
How do companies progress through the analytics phases?
If you've got a highly immature practice, you're going to be working with descriptive analytics at best. And it's going to be on what the industry calls vanity metrics. For instance, if you've got an executive that wants an app because everyone else has an app, that’s a pet project and they want to see how successful the app is. Seeing how many downloads you get, then, isn’t a real measure of success because there is no indication of how it impacts the business.
That's where this descriptive analytics or basic reporting, especially on vanity metrics, really comes into play. You see this a lot with organizations that are over-rotated on paid media for instance. All they care about is how many clicks or impressions they got from Google. Those are easy to report on, but at the end of the day, that's still a micro-conversion.
The better metric you should be looking at from a descriptive reporting standpoint is something along the lines of how many sales you got from the app, or how much revenue it’s generated.
From a diagnostic standpoint, that's where you break down descriptive analytics and figure out what target markets or audiences are of the highest value. Diagnostic analytics can help determine what type of customer from all channels has the highest average order value.
Then, you take it further as you start to break that down and go beyond the diagnostic. You’re trying to understand the elements that are contributing to revenue. That's where you get into predictive analytics, which is what most of the industry is trying to get into now.
Moving into predictive analytics, you have reasonable confidence that whatever action you take is going to have a net positive result for your business. So, if you break down the diagnostic analytics and see your top three segments that contribute to revenue and how you acquired them, then those segments are going to get more marketing dollars.
You predict that if you invest more in Google search, you’ll get this lift in average order value. That's the predictive department. You spend more money now and you expect to get that lift later with predictive analytics.
What features are necessary for a good analytics solution?
A good analytics solution is going to collect its own data. Being able to collect data from all the channels is important. So, if you've got offline channels as well as online channels that you need to look at together in a single environment, your analytics tool needs to offer that.
The second piece to a good analytics solution is that once you have the data together, it needs to be integrated and normalized. The analytics solution needs to take the data and process it, because when you have the integrated data store, everything else you do will be based on that integrated data store.
An analytics tool also needs to enable free-form analysis, which is where data analytics comes in. It should be a free-form environment where people can go in and can start to break down data, asking questions and interrogating the data.
Lastly, an analytics tool needs to provide great visuals, and it also needs to provide great export options, so that you can take action on whatever conclusions or outputs you have from your analytics tool.
Visualizations should be easy to understand for ease of communication, so you can clearly see the behavior of visitors and users, and it should also serve as an additional analysis technique. Many people view analytics as just breaking down tables of data, but analytics with a visual component allows you to use those visuals to help derive new insights. You can gain a contextual feel for the relationship of the data.
How can companies successfully and effectively use analytics?
There needs to be a strong focus on data governance and analytics guidance. You need to consider what guardrails and best practices you’re going to put in place in your organization or team.
Data governance is like a playbook for the company that determines rules for how each team will use data. If you don't have data governance in place, you may have one group that defines the variable one way, another group that defines the variable a different way, and the two won’t ever come together in reporting, let alone analytics.
Companies also need to focus on building a strategy first, not on collecting data first. A lot of organizations will say, "We've got a data lake, capture everything and dump it there. We'll find amazing insight into our customers and customer journey." You’ll never get to insights that will lead you to action that improves business results. Instead, if you say, "Hey, we need to increase online sales by 50 percent in 2020," all of a sudden, it's going to have a downstream effect. You’ll know what data to collect and analyze to get insights that improve business outcomes. So start with a strategy.
And then there are going to be probably three to five key business objectives or KBOs. The five levers that are going to influence that strategy.
You could almost think of it like a car. Your strategy is to drive from Anaheim to San Francisco. There are a few key levers in the car, like the gas pedal, and getting the car itself. Your strategy is to get in the car and go, and you've got a key business objective to stick your foot on the gas pedal and go.
And within a KBO, you have KPIs. What are the key performance indicators that your key business objective is influencing your strategy? Going back to the car example, when you hit the gas, you have a KPI gauge of how fast you’re going — the speedometer. And your miles per hour is a KPI that you can potentially track.
So you need a strategy, three to five KBOs, and then probably no more than three to five KPIs for each KBO. Once you have that in place, you're going to define the tactics that you need to take to influence every KPI. For instance, a retailer has determined that they needed to grow online revenue by 50 percent. One of their key business objectives would probably be to increase average order value by X percent. So you’re going to look at the KPI for average order value. And what are the tactics to take to accomplish that? Is it going to be cross-selling in the order cart process? Is it going to be discounts on promo codes you email out? Those are the tactics that you apply. So if you funnel all that down, you figure out what you're going to do. Then you collect all the data on all those tactics, all those engagements that you create, and all that funnels right back up to support the strategy.
What are common mistakes companies make?
Not putting data governance into place is a big mistake. Companies can also focus on the wrong questions and the wrong metrics, or assume that once they purchase an analytics tool the work is done. They may also assume they can collect all the data and figure it out later, without analyzing or integrating it.
There are a lot of organizations that are trying to centralize everything into a data lake. And then they have a centralized analyst team or data science team that's going to go in and find great insight because they have one source of truth. But you can't just have one little group deriving insight. You've got people that have jobs to do, and to do those jobs they need data and insight in real-time. If you've got a centralized team that's trying to manage this, and if you're the email manager sending in a request for a report, you're going to get that back in three weeks when the time has passed and you're too late to the game.
Also, going with hard numbers all the time at the expense of human understanding is as bad as ignoring data. If you have experience in an industry and you see hard numbers indicating something, but you know that goes against every instinct that you have, there's a qualitative factor that should be brought into play and you should try to understand if there’s an error in data collection or analysis.
How will analytics change in the future?
A big thing that analytics is going to have to address is privacy, especially with legislation, like the General Data Protection Regulation (GDPR) in Europe and the California privacy law. The next issue is the explosion of channels. Things that have never been digitized before are now being digitized. And those are going to be brand-new data sets and brand-new channels that you're going to have to integrate with existing data. So web and mobile are maturing channels, and you have other emerging channels right now, like voice assistants.
There will also be an increase in augmented analytics, where AI and machine learning come in to assist with analytics. There are novice users who can use advanced analytics because of the AI functionalities that are surfacing without the users even being aware of that help.