UPI Study › Courses › Business Essentials › Free lessons › Lesson 17.3 Free with a free account
Business Intelligence: Turning Data Into Strategic Insight
Business Essentials · Chapter 17: Information Systems and Digital BusinessSection 17.3 · lesson 79 of 10513 min · free with a free account
Free preview
Create a free account to keep watching
It takes a minute and costs nothing. Your free account unlocks all of Chapter 1 and 3 more lessons of Business Essentials.
Lesson complete
Earn the 3 college credits for Business Essentials
Enrolling adds a quiz after each lesson, the graded work and the official transcript. Recommended for credit by NCCRS.
Next lesson: Management Information Systems (MIS): Tools That Support Managerial Decisions →← Digital Communication and Internet Collaboration: Platforms and InfluenceNext lesson: Management Information Systems (MIS): Tools That Support Managerial Decisions →
In this lesson
What data can tell a manager, and what it can't.
- Define business intelligence and what it produces
- Trace the 5-stage data journey
- Match each analytical question to what it supports
- Assess a claim for correlation versus causation
Free study notes
Get this lesson's notes by email
The key points and 4 practice questions with answers, in your inbox. One email, free.
One email. Unsubscribe in one click.
Transcript
Everything said in the video, section by section (1,912 words). Read along, or search it.
Information Systems and Digital Business
Business Intelligence: Turning Data Into Strategic Insight
Every business leaves a trail of data behind it, simply by operating. Each sale, each shipment, each customer contact gets recorded somewhere. Business intelligence is the work of turning that trail into decisions. We'll follow data on its journey from capture to a manager's screen, and see where the real effort goes. We'll look at data quality, the precondition for everything else. Then four analytical questions, and what each one can and can't support. Then the most important limitation in business analytics. Correlation is not causation. We'll cover dashboards, self-service tools, and the skills managers need to use analysis well. And we'll end with an app that seemed to make customers spend more, but probably didn't.
Defined
Examining the Traces a Business Leaves
Start with a definition. Business intelligence is the systems and practices that turn operational data into information a business can act on. It's different from the operational systems we met in chapter eight. Those systems run the business. They record a sale, or log a shipment. And they generate data along the way, as a by-product of doing the work. Business intelligence doesn't run anything. It examines the traces the business leaves as it runs. Then it turns those traces into something a manager can use to make a decision. So one kind of system does the work. The other studies the record of the work.
The Data Journey
5 Stages From Record to Decision
Data travels through five stages before it becomes a decision. Capture happens in the operational systems, as transactions, interactions, and events are recorded in the ordinary course of business. Extraction and consolidation pull data out of those systems, and bring it into a common, comparable form. Storage keeps the consolidated data available over time, in a data warehouse or an equivalent facility. A data warehouse is a system organized to hold data for analysis, rather than for running daily transactions. Analysis examines the stored data to answer a question. And presentation puts the answer in front of the person who needs it, in a report, a chart, or a dashboard. So where does the effort go? Mostly into consolidation. The operational systems were often built separately, and they record similar things differently. One system records a customer by account number, another by name. Dates get stored in different formats. Reconciling all that, by hand or by carefully designed rules, is laborious, and it carries among the highest risk of failure. Yet it gets the least attention. Discussions focus on the visible stages, analysis and presentation. So a project that budgets by visibility underfunds consolidation, and is then surprised by the delay.
Data Quality
The Precondition for Everything Downstream
Everything downstream depends on data quality. An analysis built on poor data produces a poor answer, however sophisticated the technique applied to it. Quality has four dimensions. Accuracy asks whether a recorded value matches reality. Completeness asks whether everything that should be present is there, with no missing records or fields. Timeliness asks whether the data is current enough for the decision it's meant to inform. And consistency asks whether the same fact is recorded the same way across systems. One obstacle comes up again and again. Call it the definition problem. Two departments each report a figure called active customers. One counts anyone who has ever made a purchase. The other counts only those who bought in the last ninety days. The two numbers can't be compared until everyone agrees what each one measures. That's why definitions set at the point of capture decide what can be reliably consolidated later.
4 Analytical Questions
Not a Ladder, but 4 Questions
Business intelligence is often described as an analytics maturity progression, a ladder from basic techniques to advanced ones. It's more useful to think of it as four distinct questions. A technique's value lies in the question it answers, not in a ladder position you climb for its own sake. Descriptive analytics asks what happened. It summarizes past events, like last month's sales by region. Diagnostic analytics asks why it happened. It examines relationships in the data to explain an outcome, like why sales fell in one region. Predictive analytics asks what is likely to happen. It uses patterns in past data to estimate a future outcome, like which customers are likely to stop buying. And prescriptive analytics asks what should be done. It recommends an action, given the predicted outcomes and the business's objectives. Each question is legitimate. The mistake is asking one of them of data built to answer another.
Table 17.3.1
What Each Question Can and Cannot Support
This table sets out what each question requires, and the kind of conclusion it can support. Descriptive analytics needs reliable, consolidated historical data. It gives an accurate account of past performance. But it can't explain why anything happened. Diagnostic analytics needs data on the factors that might explain the outcome, and a method for comparing them. It supports a plausible explanation, consistent with the data. Not certainty that it's the only one, without further testing. Predictive analytics needs a pattern in past data that's assumed to continue, and a method for projecting it. It supports an estimate, with a stated degree of confidence. It can't guarantee the future will resemble the past. Prescriptive analytics needs a predictive model, plus a stated objective, and the costs of each option. It supports a recommended action consistent with that objective. It can't promise the action will succeed, unless it's later tested. Notice that each question needs more than the last. Ask a predictive question of descriptive data, or treat a diagnosis as certain, and you're asking more than the data can give.
The Crucial Limitation
Correlation Is Not Causation
Now the most important limitation in this section. Most business analytics is correlational. It observes that two things move together in the data. It doesn't establish that one causes the other. A correlation does support a prediction. If two things have moved together before, seeing one tells you something about the other. But it doesn't support a causal claim, about what an intervention would achieve. Here's the example. Customers who use a retailer's mobile app buy more than customers who don't. That doesn't show that making customers use the app would increase their spending. The app users may already have been the more engaged, higher-spending customers, before they ever downloaded it. The app may have changed nothing about their behavior at all. The gap is explained by confounding. A confounding factor influences both the supposed cause and the supposed effect, creating an association that isn't causal. Here, prior engagement is a plausible confounder. Engaged customers may be more likely both to adopt the app, and to spend. Push every customer onto the app, and the least engaged may use it without spending any more. That's because the app was never the cause of the original pattern.
The Design That Addresses It
The Experiment
So what addresses confounding? The experiment, which we met in chapter twelve, in marketing research. A business assigns customers to conditions, say app access, or no app access. The key is how. Assignment must be unrelated to the customers' prior characteristics. Then the business compares the outcomes of the groups. Because assignment had nothing to do with how engaged a customer already was, engagement can't explain the difference. So any difference in outcomes can be attributed to the app itself. It can't be put down to differences between the kinds of customers who chose each condition. An experiment isn't always feasible, or affordable. But it's the design that turns a correlational observation into a causal one. And a business that can't run one should hold its causal conclusions accordingly loosely.
Visualization
Dashboards, and the Errors That Mislead
A dashboard is a visual display of a small set of measures, updated on a regular cadence. It's meant to give an at-a-glance view of performance, and its value depends on restraint. A dashboard crowded with every measure a business could report gives the eye nowhere to rest. It defeats its own purpose. A small, well-chosen set, reviewed as often as the underlying reality actually changes, supports quick, confident reading. Three design errors mislead readers. A truncated axis, one that doesn't start at zero, stretches the visible range. A small change suddenly looks dramatic. A misleading scale, like uneven gridlines, or a distorted area where area is read as size, skews the comparison. And a chart chosen for appearance, like a three-dimensional design, looks impressive but makes values harder to compare. Here's the uncomfortable part. Each error can happen with no intent to mislead at all. It just takes an unexamined choice of format. So a reader should always ask whether the design is distorting the comparison being made.
Access, Skill, and Ethics
Who Uses the Analysis, and How
Self-service analytics gives managers and analysts tools to query and visualize data themselves. They no longer have to request a report from a specialized team. It trades access against consistency. Broader access lets more people answer their own questions quickly, without waiting in a queue. But people may define the same measure differently, apply different filters, or use outdated data. The result is numbers that disagree, even when they claim to describe the same thing. So self-service needs, at minimum, agreed definitions for the most important measures. Next, the skills question. There's a gap between producing an analysis and using one. A skilled analyst can build a sophisticated model. It creates no value until a manager can interpret it. That means knowing what it assumes, and recognizing the limits of what it can support. Neglect managers' interpretive skill, and sophisticated analysis may sit unused. Or worse, they're misapplied, with an estimate taken as certainty, or a correlation taken as cause. Finally, ethics. The ability to analyze data doesn't settle whether a particular use is appropriate. Chapter eighteen takes that up.
Assessing a Claim
3 Questions, Asked in Order
So how do you assess an analytical claim someone brings you? Ask three questions, in sequence. First, what data underlies the claim? And does it actually measure what it's presented as measuring? Remember the definition problem. Second, what comparison is being made, and is it fair? Is it comparing like with like, or groups that differ in some other important way? Third, what alternative explanation could produce the same pattern, without the causal story being offered? A confounding factor, affecting both the supposed cause and the supposed effect, is the classic example. A claim that survives all three is on firmer ground than one that was never asked them. And here's the point worth keeping. A manager who asks these questions routinely is better protected than one equipped only with more sophisticated software.
Case Vignette
The App That Didn't Cause the Spending
Here's a composite case, built for instruction. Sana is the marketing analyst at a regional home goods retailer. She finds that customers who've downloaded the retailer's app spend more per year, on average, than those who haven't. She recommends an aggressive campaign pushing every customer to download it, projecting a substantial rise in spending. Before approving the budget, her manager asks the three questions. The data? Reliable purchase records tied to app usage. That seems sound. The comparison? App users against non-users, with no adjustment for how the groups differed before either one encountered the app. The alternative? Frequent, engaged shoppers may have been more likely to download the app in the first place. So engagement could be a confounder, behind both app adoption and spending. The trade-off is acting fast on a promising pattern, or testing it first. And a full campaign is costly to reverse. On reflection, Sana proposes a small experiment instead. Offer the app to a randomly selected group of customers without it, and compare their spending with a similar group not offered it. If spending rises, the causal claim is supported at far lower cost. If it doesn't, no campaign gets built on a correlation.
About Business Essentials at UPI Study
Business Essentials (Business 100) is a self-paced online course of 105 video lessons in 24 chapters, from the business environment, economics and global markets to ownership, management, leadership, operations, people, marketing, accounting, finance and banking. It carries a NCCRS credit recommendation for 3 semester credits.
Every lesson transcript is free on pages like this one, and the videos are free with a free UPI Study account. Enrolling adds a practice quiz after each lesson, the graded work and final exam that earn the credit, lifetime access, and an official transcript on completion. UPI Study has helped over 130,000 students with college credits. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Enroll now → Course details Download the syllabus All Business courses Every course with PRO →
Why take it here
Why students take Business Essentials with UPI Study
Where the credit goes
Universities that work with NCCRS
Business Essentials carries a NCCRS credit recommendation. These universities work with NCCRS; some are UPI Study's signed partners. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Thomas Edison State UniversitySigned partner
University of the PeopleSigned partner
Charter Oak State CollegeSigned partner
Strayer UniversitySigned partner
Columbia Southern UniversitySigned partner
Indiana TechSigned partner
Penn StateWorks with NCCRS
Southern New Hampshire UniversityWorks with NCCRS
SUNYWorks with NCCRS
+many more cooperating universitiesFind my college →Find my college → Not listed? Ask Athena in the corner of this page, or book a 15-minute call.
FAQ
Questions students ask us
About Business Intelligence: Turning Data Into Strategic Insight
What is business intelligence?
What are the four types of data analytics?
Why doesn't correlation prove causation in business data?
How can a business dashboard be misleading?
About the Business Essentials course
How many college credits is Business Essentials?
How much does Business Essentials cost and what is included?
Which universities accept UPI Study credits?
How long does Business Essentials take?
How are the exams done?
How do my credits reach my college?
Ready for the full course? Earn 3 credits in Business Essentials.
105 video lessons, a quiz after each one, and the graded work that earns the credit. Recommended for credit by NCCRS.
3 credits · NCCRSEnroll →