8.4 Web Analytics

Improving an Organization’s Online Presence

Web analytics play an important role in helping organizations optimize their websites and improve their online presence. By analyzing the data obtained, organizations can identify areas for improvement and make data-driven decisions to enhance their websites. For example, through web analytics, organizations can identify pages with high bounce rates and low conversion rates, indicating potential issues in user experience or content. A bounce rate in web analytics refers to the percentage of visitors who navigate away from a website after viewing only one page, indicating a lack of engagement or interaction with additional content. A conversion rate in web analytics refers to the percentage of website visitors who complete a desired action, such as making a purchase, filling out a form, or signing up for a newsletter.

Based on the data, an organization can make targeted improvements, such as optimizing page load times, enhancing navigation, or refining the messaging on those pages. This knowledge enables organizations to tailor their online strategies, optimize marketing efforts, and create a seamless user experience, ultimately driving higher customer satisfaction and better online performance.

Optimizing Metrics

A metric is a quantifiable measure used to track and evaluate the performance, progress, or success of a particular aspect of a business, campaign, or activity. Metrics provide specific data points that can be analyzed to gain insights into how well a website or digital platform is performing. They can include a wide range of measurements, such as website traffic, conversion rates, bounce rates, session duration, and many others. A KPI is one type of metric, for example. These are typically high-level metrics that directly align with the organization’s goals and are critical for assessing performance and progress.

The ability to use these metrics to measure website performance and user behavior allows organizations to gauge the effectiveness of their online presence. Web analytics provide insights into KPIs such as the number of unique visitors, page views, average session duration, and conversion rates. These metrics allow organizations to track and measure their website’s success over time and compare it against predefined goals and benchmarks. They can also identify opportunities for optimization, refine marketing strategies, and create personalized user experiences based on user preferences and patterns.

Gaining Insights

By using the data from web analytics, decision-makers can gain a comprehensive understanding of their website’s performance, identify areas of improvement, and assess the impact of various marketing initiatives or website changes. For example, web analytics can help determine the effectiveness of different advertising campaigns by tracking referral sources, click-through rates, and conversion rates associated with each campaign.

Measured as a percentage, the click-through rate (CTR) tells the viewer how effective an ad is at attracting clicks. The CTR represents the total clicks an ad receives divided by the total impressions, or instances the ad is loaded on a page. A 2 to 5 percent CTR is generally accepted as being successful, but this varies by industry. So if an ad was viewed 10,000 times and was clicked on 500 times, that’s a 5 percent CTR. The CTR helps assess the effectiveness of digital marketing efforts. It allows decision-makers to allocate resources effectively and invest in strategies that generate the highest ROI.

Web Analytics Platforms

Web analytics platforms offer a wide range of features and functionalities to explore and analyze website data. Data scientists can access prebuilt dashboards, reports, and visualizations provided by these tools to gain insight into website performance, user behavior, and conversion metrics. They can segment data based on various dimensions, apply filters, and conduct in-depth analysis using available metrics and dimensions. Additionally, these platforms often offer advanced features like custom event tracking, goal tracking, and e-commerce tracking, enabling data scientists to perform detailed analyses tailored to specific business goals.

Data Extraction and Transformation

Data scientists can extract data from web sources using their APIs or data export tools. Extraction involves removing data, often in multiple forms, from an online source or repository. Frequently, the data file types vary and come in an unstructured form. Extraction makes it possible to filter, organize, and store the data in a common location. Analysts can programmatically retrieve raw data, such as page views, events, and user demographics. To put the data into usable context, it may be helpful to extract data from multiple sources, transform the data into a usable format, and load the data into a data warehouse for data analytics, a process called extract-transform-load (ETL) (Figure 8.13). These steps are as follows:

1. Information is extracted from one or more sources and is prepared for transformation.
2. Transformation may involve cleaning up missing or inconsistent data, creating new derived variables, and transforming data into a suitable format for analysis. The data are filtered and organized.
3. The transformed data are loaded into a centralized location.

Figure 8.13 ETL is part of the overall workflow for moving data from a database, transforming it, and loading it to a data warehouse for transmission and analysis. (attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

Statistical Analysis and Modeling

With data that have undergone the ETL process, data scientists can apply various statistical analysis techniques and modeling approaches to gain insights and make predictions. They can use tools like exploratory data analysis classification and customer segmentation.

Exploratory Data Analysis

Exploratory data analysis (EDA) is the process of conducting the initial review of a dataset to spot any patterns or trends early on. Data analysis at this stage looks for relationships between variables. When one feature increases, does another feature do the same? What are the correlations, if any? This initial overview helps data scientists formulate the right questions to ask. Visual outputs can help by providing clues to these relationships. Consider the scatterplot diagram in Figure 8.14. This data from an educational environment recorded student absences and cross-referenced them with the final grade point average (GPA).

Figure 8.14 Scatterplot diagrams can provide evidence of correlation between features such as absences and GPA. (attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

You can observe quickly that there is indeed a relationship between the variables of absences and GPA. The lower the absences (the independent variable), the higher the GPA. The trend goes downward and to the right, signifying a negative correlation.

Classification

A scatterplot like the one in the previous example is helpful not only for seeing trends but also for classifying data. One such approach involves clustering, which involves identifying the most relevant characteristics of the data and plotting them out, with the idea being that similar data points will appear to cluster together near the average value, or centroid. Using the iris dataset (refer to 8.1 The Business Analytics Process), you can plot data on petal width (Figure 8.15). For clarity, the apparent clusters are colored differently to communicate the distinctions better visually. For example, the Iris setosa data group together in the lower-left portion of the chart and are colored purple.

Figure 8.15 In plotting petal width of iris species, the graph reveals three very clear clusters of data. Purple circles = Iris setosa; green circles = Iris versicolor; blue circles = Iris virginica. (source: modification of work “Iris dataset scatterplot” by “Nicoguaro”/Wikimedia Commons, CC BY 4.0)

Custom Analysis and Visualization

Data scientists can also utilize programming languages such as Python or R to perform custom analysis on the cleaned data. They can leverage libraries and packages specific to web analytics, such as the Google Analytics Reporting API client libraries for Python or R, to access and analyze data programmatically. To gain greater flexibility in conducting advanced analysis, implementing specific algorithms, and creating tailored visualizations to communicate insights effectively, data scientists often write custom code.

Using the data from the customers in your study to help the marketing team, after you clean the data, your next goal might be to identify which characteristics have the most impact on the customer’s spending score. You could investigate this by trying to plot age versus spending score (Figure 8.16).

Figure 8.16 Plotting the relationship between spending score and age reveals some groups of customers that may need further analysis. (data source: Zubair Mustafa, "Shopping Mall Customer Segmentation Data," Kaggle, updated April 2024, https://www.kaggle.com/datasets/zubairmustafa/shopping-mallcustomer-segmentation-data; attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

Looking at the resulting scatterplot, you can observe that the data seem to be distributed evenly across the whole graph. However, looking more closely, you can see some groups of data, such as a spending score over eighty in people ages seventy to ninety years, a spending score between forty and sixty in people ages forty to sixty years, and a spending score between twenty and forty in people ages twenty to forty years. These distributions likely do not tell the whole story, which can lead you to examine the relationship between age and annual income to determine if that provides any insights (Figure 8.17).

Figure 8.17 Comparing annual income and age in the customer data again reveals several subtle clusters. (data source: Zubair Mustafa, "Shopping Mall Customer Segmentation Data," Kaggle, updated April 2024, https://www.kaggle.com/datasets/zubairmustafa/shopping-mallcustomer- segmentation-data; attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

Again, the data here is distributed somewhat evenly across the graph, but there are some clusters of data. The first shows one cluster of people between ages forty and sixty years with the highest annual incomes, between $150,000 and $200,000, and another cluster shows people between the ages of twenty and forty with incomes between $100,000 and $150,000.

With this data in mind, the next step is to examine the correlation between spending score and annual income (Figure 8.18).

Figure 8.18 Comparing the customers' annual income and spending score reveals additional information that may help to find the best groups to target with a marketing campaign. (data source: Kaggle: Zubair Mustafa; attribution: Copyright Rice University, OpenStax, under CC BY 4.0 license)

Here, there are several clusters of data that give the marketing team a better idea of who their target audience is. Spending score is highest across all income groups but especially for customers who earn $100,000 to $150,000 annually. There is also a cluster of customers ages twenty to forty in that income group, but the customers ages seventy to ninety show a cluster with a higher spending score than those younger customers. These data may be used to show shopping trends, but they do not provide data on what the customers are buying, so the marketing team will need to do some additional analysis on products to further refine their target audience.

This example shows how data scientists can employ web analytics tools and techniques to access, transform, and analyze website data. They can use the tools to uncover meaningful patterns, correlations, and trends, providing valuable insights into website performance, user behavior, and marketing effectiveness. The analysis conducted by data scientists helps organizations optimize their online presence, make informed decisions, and drive business success in the digital landscape.

Identifying Areas for Improvement

Data can be analyzed to uncover insights such as which pages have high bounce rates or low conversion rates or are underperforming. By examining user behavior, traffic sources, and engagement metrics, data scientists can pinpoint specific areas where improvements can be made. These improvements might involve, for example, optimizing landing pages, refining calls to action, or streamlining the checkout process. Web analytics tools provide visualizations, reports, and data segmentation capabilities to support this analysis.

A/B Testing

The method A/B testing, also referred to as split testing, is used to compare two versions (A and B) of a webpage, email, or advertisement to determine which one performs better in terms of user engagement or conversion rate. This process randomly splits website visitors into different groups and exposes them to different versions of a page that may have, for example, different headlines, layouts, or calls to action. A call to action is a prompt or directive placed within a website, an advertisement, or marketing material that encourages users to take a specific action, such as making a purchase, signing up for a newsletter, or requesting more information.

A/B testing helps identify which elements have a positive impact on user engagement, conversion rates, or other key metrics, and this enables organizations to make data-driven decisions about how to optimize their websites and improve overall performance. This testing can be accomplished by dividing pages into control and variant groups, and the changes can be implemented on either the server side or client side.

On the server side, SEO tests point toward the code itself. The advantage is a smoother and more stable experience for the user. The drawback is its higher complexity, which may require knowledgeable information technology staff to implement and monitor. Client-side testing is deployed with JavaScript coding. There’s a slight unsteadiness between the old and new versions of a page. The advantage here is that it is easier to implement. The process does not require hardwiring or specialized training.