Actionable insights are conclusions drawn from data that can be turned directly into an action or a response. The data informing the insights can be structured or unstructured, quantitative or qualitative. While actionable insights provide the impetus for an action, people or processes are needed to execute the actions.
Actionable insights arise from one or more people analyzing raw data. Actionable insights can be derived from big data when large amounts of structured and unstructured data are involved.
While some actionable insights can arise directly from data, analysis must be applied in some cases to inform the necessary actions. Data can paint a picture of what happened, but related data -- or analysis performed by a human being -- can be needed to explain why it happened. It's often the "why" (vs. the "what") that makes an insight actionable.
Anybody can use actionable insights. In a business setting, managers can use actionable insights to improve products, services, company culture, employee morale or employee retention. The data to inform these insights can come from employees, customers or partners. All departments in an organization can use actionable insights to make changes and improvements, including engineering, sales, marketing, finance, legal, operations and executive management.
Actionable insights must factor in the feasibility of the corresponding action. An actionable insight is useless if the proposed activity is impractical or impossible to perform. For example, an insight from customer interviews might be that 60% of customers want to visit the moon. This insight is not actionable because the company has no means to make it happen.
How to get actionable insights
To generate actionable insights, companies need an infrastructure to support data collection and data management. Companies also require the means to analyze the data to extract meaning.
Actionable insights are derived from data. Companies need internal IT support, or support from a third-party vendor, to collect, store, organize and analyze their data. The infrastructure provider, whether internal or external, must support a growing volume of data. For example, deriving insights from IoT data, social media data or multimedia assets may require terabytes of storage per day or per week.
Actionable insights follow the rule of "garbage in, garbage out." If the data being analyzed is incomplete, invalid or irrelevant, then the resulting insights will be flawed. Companies can institute quality assurance processes to test and validate the accuracy of data before it's imported into a repository or database. Routine audits can also be performed once the data is imported.
Data integration (e.g., breaking down silos)
Data silos can prevent companies from seeing a full picture. Data sitting in separate, closed-off databases prevents companies from a 360-degree view of their customers. A positive review for one service may mask a scathing review (from the same customer) for a different service -- one that is stored in a database disconnected from the first.
Actionable insights are most effective when all relevant data is brought together for analysis. For example, a bank might merge customer website interaction data with offline data of those same customers, such as branch visits and ATM transactions.
Publicize and evangelize
In larger organizations, managers and employees may not be aware that key data repositories or databases are available. Alternatively, they may be aware of their existence, but unaware of how to take advantage of the data. Companies can generate more actionable insights from their data by publicizing and evangelizing the available systems, tools and reports. Companies can hold company-wide training on the tools available or hold departmental or individual training sessions.
Actionable insights happen when humans analyze data to determine an action or a set of actions. The data, such as a set of reports or a dashboard, provide the "what," while insights are derived from the analysis of that data, or the "why." An insight that leads to an action is more valuable than one that simply answers a question. In addition, specific actions are more effective than generic actions. For example, the action of "hire two software engineers in the Midwest" is better than "hire more software engineers."
Actionable insights examples
Actionable insights can be generated from customer feedback. If customers of a software product state that they prefer a competitive offering because it's twice as fast, then an actionable insight, such as the need to improve software speed, can be directly drawn. In this scenario, the manufacturer might create a faster version of the product, then survey customers again to see if the response made to the actionable insight improved customer satisfaction.
Marketers use web analytics, such as Google Analytics, as a primary means for deriving actionable insights. A web analytics platform tracks hundreds or thousands of metrics from users' visits to a company's website. The metrics tracked includes page views, sessions, time on site, pages per session, bounce rate and conversions.
For example, a clothing manufacturer can segment its product pages by category, such as men's shirts, women's shirts, men's pants and women's pants, and analyze transactions and conversion rates per category. If sales of men's shirts are down 20% this week, the retailer might check their inventory data and notice that a popular shirt is out of stock. The actionable insight would be to accelerate the backorder of that shirt.
Medical diagnosis and AI
In hundreds of hospitals each day, physicians derive actionable insights from medical charts, X-rays, CT scans and MRIs. Physicians must process all the information in front of them. Then, they combine that with their own medical experience and the patient's medical history to determine a diagnosis.
In research studies, scientists have begun to use artificial intelligence (AI) to assist in deriving actionable medical insights. At the University of California, San Francisco, researchers combined neuroimaging (e.g., brain scans) with machine learning to predict whether a patient would develop Alzheimer's disease. When a radiologist reviews the scans, the development of Alzheimer's is difficult to detect with the naked eye.
Researchers trained the machine learning algorithm by feeding it images from a public data set of brain scans from patients with a wide range of conditions. The algorithm was able to learn which features are important for predicting the diagnosis of Alzheimer's disease and which are not.
The result was that the AI algorithm was able to diagnose early-stage Alzheimer's disease six years before a clinical diagnosis is made. This actionable insight gives doctors and patients much greater lead time to intervene with treatments.
Similar to web analytics, marketing attribution is a method for determining the channels and content offers that contributed to a sale. Marketing attribution attempts to track all customers' touchpoints with a company, especially during the research and evaluation period before the sale.
The most common marketing attribution models used are:
- First-touch attribution: the customer's first interaction with the brand receives credit for the sale
- Last-touch attribution: the customer's last interaction with the brand receives credit for the sale
- Multi-touch attribution: the credit is split up among the touchpoints. Several sub-categories of multi-touch attribution determine different ways to assign the credit
With marketing attribution, marketers seek to understand the advertising channel or content offers that return the highest ROI. The actionable insight becomes where (and how) to invest in future marketing programs.
Insights from business data using AI
In addition to diagnosing diseases from medical scans, AI can be used to process business data in the workplace. Employees can spend days or weeks combing through documents, spreadsheets and reports, looking for key answers hidden inside them.
This activity can be especially challenging when the data is unstructured, in formats such as email, video, audio or images. Some companies are using AI platforms and tools (e.g., Watson Explorer from IBM) to analyze raw data to derive actionable insights. The AI platforms can analyze structured and unstructured raw data, using natural language processing to understand the meaning of content and documents.
Toyota Financial Services used an AI platform to assist agents in their call centers. The AI platform processed data from both internal and external sources and provided insights about its four million customers that agents could act on. The AI platform provided near-instantaneous insights on customers' preferences or concerns, which helped agents handle the incoming calls accordingly.