Analytics is not rocket science. And business analytics, which uses relatively simple methodologies on past data, is a problem-solving tool that everyone in an organization can use to effect positive change on a day-today basis.
Business analytics can help you solve 80 percent of your business problems at a fraction of the cost of complex analytics. You can use them any time you are facing a management decision that involves data. The specific business question you are wrestling with will dictate which methodology you adopt. As you’ll see, the example we often use is a small Oregon winery, but the winery can function as a stand-in for other small businesses, including publishing companies.
The following chart shows the most commonly used analytics methodologies, and the text discusses them in terms of what they might do for you. Note that the first four get the most use and can be done in Excel by non analysts who understand the business contexts.
Aggregate analysis is the simplest and most commonly used analytics methodology. It is also the first step in most other methodologies.
It is used to describe a population (that is, a collection of things or people you want to understand, perhaps customers or prospects and perhaps products such as items produced in a given month or sold in a particular category on your website). It can also be used to describe a segment or to compare two segments.
This methodology can help you answer descriptive or comparative questions for your business, such as:
- Who are my customers?
- How are my customers different in one geographical region versus another?
- Do younger people access our digital product through tablets more so than older people?
- What worked and what didn’t work in the previous marketing campaign?
Gable Wines hosts weddings in addition to producing and selling its products. In order to customize communication, the CEO wants to understand who is booking weddings. After a look at bookings for the past three years (say, 300 customers), aggregate analysis on age, gender, and location reveals that 85 percent of the customers are women, with a mean age of 33, and that 60 percent of them are in Oregon.
This information is useful not only to set the tone of communications but also to target marketing efforts by asking questions such as:
Where can we find women in their thirties?
What publications do they read?
Do they eat in certain restaurants? Do certain stores cater to this demographic?
Can we advertise at locations that do?
Another business problem for Gable Wines is increasing conversion of people who fill out an online form for wedding reservations. Comparing results from a short form and a standard form can show which gets better conversion.
Correlation looks for the relationship between two or more things with the goal of being able to explain or drive one with the other. It is used to answer business questions like:
- Why is last quarter’s revenue below expectation? (By finding the segments that correlate with revenue gap)
- Why is the foot traffic down in a particular location?
(By finding correlators of foot traffic)
Looking for good leads, Gable Wines uses inquiry form submission and then narrows initial hypotheses down to the four most plausible ones:
- Certain wine guides and Google-paid searches produce better leads.
- Mobile users are better leads.
- People seeing the pricing page result in bad leads.
- Certain locations produce better leads, such as the local Oregon area.
Correlation analysis on each of these hypotheses—conversion by traffic sources, mobile versus not mobile, people seeing pricing pages versus those who do not, and location— quickly disproves two of the hypotheses, leaving two to explore further: that different traffic sources indeed have different conversion and that geography is a big factor as well.
The chart to the right provides an example of what the conversion data and correlation analysis for the conversion across various sources looks like.
By reallocating the budget from bad lead sources to good lead sources, Gable Wines drives an incremental gain. First, it cuts investment on weak lead producers (pay per click on Bing, Facebook page, and certain wine guides). Second, it increases spending on the efficient high producers of good leads (Google pay per click and certain wedding guides).
Through this simple correlation analysis, the company generates a 12 percent ($120,000) incremental revenue with the same $27,000 marketing budget it has had. Of course every channel saturates and conversion rates will start to drop after a certain limit.
Trends analysis is aggregate or correlation analysis over time; that is, analysis of trends over a particular period. It is most often used to examine sales performance or revenue growth. The goal is to identify breaks in a trend and pinpoint the impacted segments and drivers over a period of time.
This methodology helps with questions like:
- Has our customer base been shifting to a younger demographic (by trending age over time)?
- Why is growth trending down (by looking at growth over time and breaking it down by different segments to find correlators where the growth is slowing down)?
- Why are sales of the such-and-such a product slowing down (by looking at its sales over last few months and identifying internal and external correlators to sales)?
Sizing and Estimation
Sizing and estimation is a structured approach to making a nearly accurate guesstimate in the absence of historical data. It is technically not an analytics methodology because it does not use historical data. But since it uses a structured approach based on assumptions and limited external and internal data points and since it’s widely used in business to drive decisions, we are including it here.
It is typically used to make a business case for going into a new market, to understand the potential marketable universe for a product yet to be launched, and to quickly size up the impact of a decision or change.
Sizing and estimation can be used to answer questions like:
- How many networking routers are sold in the United
- States yearly, and what percentage of those are sold to consumers?
- How many wedding gowns are sold in Los Angeles annually, and what percentage of that market can we capture?
Predictive analytics looks at both current and historical data to make predictions about future events. Weather forecasts are common examples.
The methodology exploits fundamental correlations between a metric of interest in a certain future time and other correlated metrics at a current or historical time. The future state of a metric can be predicted with some accuracy by observing the current and historical correlators.
Questions addressed by predictive analytics look very similar to those addressed by correlation analysis. Examples include:
- What are the drivers of customer churn?
- What drives customer engagement?
- Why is conversion going down, and what is driving it?
Often, predictive analytics gets complex enough to border on rocket science.
Segmentation analytics group customers or products into meaningful segments, usually to enable better targeting in order to drive higher value through customization.
- You know you need segmentation when your goal is to answer customization questions like:
- How do we customize our offering (to whom do we offer which product)?
- What does our product portfolio look like (more profitable product SKUs versus the rest)?
Looking at your customer base as current customers and prospects is a simple segmentation. You can segment these groups further based on products they own (apparel shoppers and book shoppers for Amazon, for example).
As you can imagine, there are many ways of slicing the pie, depending on why you are segmenting. Simple segmentation methods include RFM (recency, frequency, monetary).
In its standard form, these three variables at three levels each—low, medium, high—divide the population into 27 segments.
Here’s a segmentation example using product versioning.
If you were running a marketing campaign to sell the justreleased version 9.0 of your consumer software, then you could segment the consumer base as prospects, trial downloaders, and users of versions 1.0 to 8.0, to see who the best adopters are.
Findings can be surprising. For instance, you might discover that version skippers (say, those who are on version 6 or 7) have higher adoption rates than people using version 8.0.
Customer Life Cycle (CLC) Analysis
CLC analysis, which is often done in conjunction with segmentation, looks at the different stages of the purchase process to determine what stage a group of customers is at and to decide how to move them up to the next stage.
This methodology answers questions like,
“What is our sales conversion funnel?” That is, of the people who show up as leads, how many qualify, how many become opportunities, how many try the product, and how many become customers? CLC can also answer other questions, such as, “How do customers progress through our products?”
Cohort analysis is a special type of CLC analysis in which customers are analyzed in relation to their start date or active date. For example, the July cohort may need 15 days to habituate to your product, but the December cohort may need 30 days. Sales funnel analysis is another application of CLC.
Applications At A Glance
The following charts summarize some common applications of these methodologies to business problems.
- In most cases, the data you will need for simple business analytics is already at hand; you just have to dig it out and look at it.
- Eighty percent of business problems are solved using the four most common analytics methodologies, which can be performed using Excel. Aggregate analysis is used to describe or compare. Correlation analysis is used to find relationships between things and use that to move one with the hope of moving the other. Trend analysis is used to analyze trends over time. Sizing and estimation is a methodology used to generate accurate estimates from limited internal data.
Piyanka Jain is CEO of Aryng, a management consulting company that focuses on analytics.
She has been a keynote speaker at major business and analytics conferences. Puneet Sharma is VP of Analytics and User Research at Move, Inc., having previously worked at PayPal, Capital One, and HSBC. This article is derived from their new book Behind Every Good Decision: How Anyone Can Use Business Analytics to Turn Data into Profitable Insight, which includes step-by-step guidance. © 2014 Piyanka Jain and Puneet Sharma. Published by AMACOM Books, a division of the American Management Association. All rights reserved. To learn more: amacombooks.org.