Creating an outstanding customer experience is a critical part of every sales strategy.
More than 70% of customers cite customer experience as a key factor in influencing their decision to purchase a product. (Gottlieb, 2019)
Companies that create exceptional customer experiences bring in 5.7 times more revenue than their competitors. (Morgan, 2019)
To stay competitive, companies must meet the rising expectations of their customers. Historical product adoption trends can be analyzed to provide the insight necessary to drive consistent, personalized customer interactions.
Do historical product adoption patterns follow a normal distribution?
If so, which customers fall into each adopter category?
How do the adopter categories compare between various products?
How can these findings be used to influence an effective sales strategy?
The product studied was initially launched in 2001, with an add-on module introduced in 2009.
Data was extracted from the Accounts table using SQL. In total, there are seven product configurations used across more than six hundred accounts, both active and inactive. The central limit theorem states that sufficiently large data samples will be normally distributed regardless of the population distribution. (Kwak & Kim, 2017) Sample sizes greater than thirty are generally considered to be sufficiently large for analysis. Therefore, a population size over six hundred is adequate for producing meaningful results.
A diffusion model represents the rate at which new products are adopted across a population of prospective adopters as a function of the time that has passed since the product was introduced. (Mahajan & Muller, 1979) Rogers’ work provides the framework for building a diffusion model which will provide insight into customer behaviors based on how they adopted products in the past. The adoption of new products is a social process and therefore is expected to follow a normal distribution.
Rogers’ diffusion model is based on a normal distribution; thus, we must begin the analysis by confirming that the historical product adoption data is, in fact, normally distributed. There are several ways to evaluate a data set to determine if it is normally distributed. A histogram is a chart that shows the frequency of occurrence per value in the data set. Histograms are a common data visualization method used to assess the distribution of a variable. Data that is normally distributed will have a bell-shaped histogram that is symmetrical about the mean. The mean, median, and mode will all equal in a normally distributed data set.
The Shapiro-Wilk test was developed specifically for the normal distribution and is a powerful statistical test to detect if a data set is not normally distributed. The test returns a p-value less than or equal to 0.05 when the sample is not normally distributed. P-values greater than 0.05 mean it is reasonable to assume the data is normally distributed. (Data Science Team, 2021)
Standard normal distributions are a special normal distribution where the mean equals zero and the standard deviation equals one. A value on a standard normal distribution, known as a z-score, represents the number of standard deviations an observation is above or below the mean. Any point in a normal distribution can be converted to a standard z-score. If the sample is normally distributed, a customer's adopter category can be determined using the z-score.
Cross-tabulation is a comparison method designed to quantitatively examine the relationship between two or more categorical variables. (Atlan, 2016) A cross-tabulation table shows the correlation between adopter categories for different products and may uncover relationships that are not readily apparent in the data.
Git repository of code used for analysis files.
Power BI was used to create a histogram showing the frequency of product adoption over time, including a to dynamically change the chart so that several different samples within the population could be evaluated. The chart shows that the frequency of the base product adoption across the entire population is not normally distributed; however, the histogram does show a distinct bimodal pattern. The frequency of the base product adoption dips just before 2016 and appears to pick up again after 2016. Product adoption before 2016 appears to be normally distributed while adoption after 2016 appears to begin another normally distributed pattern.
The bimodal pattern is a significant finding in the data and is related to the release of the add-on product in 2009. The frequency of base product adoption and the frequency of the adoption of the add-on product are related. This is logical as the add-on product is an add-on to the base product. Thus, the adoption of both products follow a similar pattern.
The population can be split into two unimodal samples: product adoption through 2016 and product adoption after 2016. This sample is close to a normal distribution, although slightly skewed. Further, the observations in 2004 and 2005 appear to be outlier values. These anomalies in the data do not have a significant impact on this analysis.
R was used to perform Shapiro-Wilk test to statistically confirm that it is reasonable to consider the sample to be normally distributed. Product adoption data from 2006 through 2016 was exported from Power BI and imported into R. The Shapiro-Wilk test returned a p-value of 0.1238 (Figure 16) which is greater than .05, indicating that it is reasonable to consider sample is normally distributed. Thus, the base product adoption from 2006 through 2016 represents a good baseline sample to which to apply Rogers’ product adoption framework.
With the understanding that the adoption of the base product from 2006 through 2016 is normally distributed, Rogers’ framework for product adoption can be applied using the special properties of a normal distribution. The mean and standard deviation of the sample were used to calculate the z-score. The z-score was then used to identify the date range for each adopter category.
Customers who adopted base product data before July 2007 are innovators or early adopters, while those who adopted the product from July 2007 through June 2010 are the early majority. Those who adopted the product from July 2010 through December 2012 fall into the late majority category, leaving those who adopted base product after 2012 as laggards.
When Rogers’ product adoption curve is overlaid with the S-curve of cumulative market saturation, we see that 50% market saturation marks the separation between the early majority and the late majority adopter categories. Nearly 60% of customers have adopted the add-on product, therefore it is reasonable to consider it to have at least 50% market saturation. Therefore, customers who have adopted already adopted the add-on product are most likely in the innovator, early adopter, or early majority category.
Now that customers have been segmented into adopter categories based on the adoption of both products, the adopter categories can be compared. The comparison of adopter categories between products shows that overall, most customers tend to stay in the same adopter category when considering another product. Approximately 44% of customers who were early adopters of base product were also early adopters of the add-on product; 59% of customers who were the early majority for adopting the first product were also the early majority for adopting the add-on product. Over 60% of customers who were laggards in adopting the first product have yet to adopt the add-on product which is consistent with Rogers’ framework. One interesting finding is that nearly half of customers who were in the late majority for adopting the first product became the early majority when adopting the add-on product.
Creating exceptional customer experience is a necessary part of every sales strategy. Understanding customer needs at each step of the customer journey provides the opportunity to solve challenges before they arise, which in turn creates a better customer experience. Historical product adoption data can be used to understand customers buying characteristics, and therefore create positive customer experiences.
Develop Profiles for Each Customer Category using quantitative and qualitative analysis.
Understand the customer journey.
Introduce innovators and early adopters to products during development phase.