How AI will change the way you use Analytics in 2024

Generative AI Use Disclosure:
This blog post was edited and optimized with AI, but the final output was generated by me, a human.
The primary {robot} image in this post was generated with Midjourney.ai, but the content surrounding the image and the final output was generated by a human.

Why AI in Analytics?

As analytics professionals, we're always seeking ways to enhance our understanding of data and to make more informed decisions. At Seer Interactive, we believe that the integration of AI into Analytics is not just a technological leap, but a transformative shift in our approach to data, analytics, and insights. 

But why should analytics professionals, seasoned in traditional methods, consider this AI-driven approach? The answer lies in three distinct areas where AI in Analytics will have the greatest impact.

• Enhanced Efficiency: AI drastically reduces the time required to process large datasets, increasing productivity.
• Deeper Insights: Uncovering previously hidden patterns and insights, AI enables more accurate predictions.
• Strategic Value: AI empowers analytics professionals with tools for enhanced customer segmentation, real-time decision-making, and predicting market trends, adding greater value to organizations.

AI in analytics isn’t just a technological upgrade; it's a paradigm shift. Whether it's through enhanced customer segmentation, real-time decision-making, or predictive market trends, AI empowers analytics professionals to deliver more value to their organizations.

In this blog post, we'll explore the compelling reasons why embracing AI in analytics is not just an option but a necessity for those looking to stay ahead in a data-driven world. So, let’s embark on this journey to understand how AI is reshaping the analytics field and why it's time for you to be a part of this transformation.

What Types of AI are used in Analytics?

Figure 1: Types of AI used in Analytics [Image created with Midjourney.ai]

Business Benefits of using AI in Analytics

In today's digital marketplace, the business benefits of using AI in analytics are profound, especially for marketers. AI-driven analytics enhances decision making, which has the potential to offer insights that are precise and actionable. But don’t expect these insights and actions to fall into your lap automagically. AI in Analytics is currently an imperfect data science that requires practice, refinement, and (above all else) validation by you, human, to make sure you get it right. 

Jim Sterne, widely recognized as the Godfather of Digital Analytics, recently spoke to my Analytics & Insights Team at Seer. When asked about the benefits of AI in Analytics, he put it this way:

“It's a creative brainstorming tool. It's not an answer machine. So put on your creativity hat. If you expect to type in text and get usable output, you're fooling yourself. What you'll get is somebody else's idea of the thing that you had in mind, which will give you new ideas, which you can then feed back into the machine and get better ideas, which you can then take out of there and make it your own. Use [AI in Analytics] as a creativity tool, as a device, as an assistant, rather than an answer system.” Jim Sterne

While I fully agree with Jim, I also believe that given proper instruction, AI-powered analytics can provide outputs that are usable (with some human validation of course). Consider these business benefits of AI in Analytics:

• Task Automation: AI can help automate routine tasks such as data cleaning, a critical aspect of data analysis. This process involves correcting or removing inaccurate records, treating missing values, and standardizing data formats. Traditionally, this requires extensive manual effort and is prone to errors. AI transforms this by automatically detecting inconsistencies, intelligently filling in missing values using techniques like regression or predictive modeling, and standardizing various data formats for analysis. 
• Trend Identification: Another significant benefit of AI in analytics is the ability to analyze customer behavior to identify trends. AI algorithms can analyze complex datasets, including browsing behavior, social media interactions, customer feedback, and purchase history, revealing subtle patterns and correlations. Identifying these patterns and discerning trends is historically a difficult task. Analysts using AI in their workflows can expedite findings and explore hypotheses faster and more effectively with AI support. 
• Marketing Intelligence: Consider the scenario of a digital marketer in an e-commerce company aiming to increase new customer acquisition rates. By using AI analytics, data analysts can create advanced customer segments based on behavioral patterns, such as browsing habits and interactions with previous marketing campaigns. Marketers can then customize campaigns for each segment, like targeting 'highly engaged but not purchasing' customers with special offers or personalized recommendations. These targeted campaigns are more likely to resonate, leading to higher conversion rates.

AI in analytics enables analytics pros and marketers to develop more effective, personalized, and timely strategies, significantly improving acquisition rates. By leveraging AI for nuanced customer segmentation, data analysts provide marketers with the insights needed for more impactful marketing strategies.

 Artificial Intelligence in GA4

It shouldn’t come as a surprise that the world’s most popular digital analytics platform, GA4 has artificial intelligence baked right into its functionality. In fact, Google has embedded Analytics Intelligence into GA4 as “Insights” that use advanced modeling techniques to help you better understand and act on your Google Analytics data. 

Figure 2:AI powered “Insights” are native to GA4 in the popup and search bar

Think of these Insights as thought starters to help you get to your data by asking natural language questions to get your data. 

• Wondering which days you get the most website traffic? Ask Insights. 
• Curious about how many users arrived via organic search? Ask Insights.
• Want to compare revenue from organic search vs paid? Ask Insights
• Did your boss ask you what your best selling online products are? Ask Insights.
• Which browsers do visitors use most to access your website? Ask Insights. 

Hopefully you get the idea. Insights is taking the chore of digging for data out of the hands of the Analyst and democratizing the data to make it accessible to all Marketers. And kudos to Google for allowing its users to leverage the power of AI without making them really learn AI. But these insights definitely won’t work for everyone. Seer Team members commented:

Thoughts: AI in Google Analytics 4 + ChatGPT

“GA4's AI features seem to be kind of weak and, by today's standards, outdated. It seems that they are primarily based on pattern recognition and there are no generative intelligence features that I know of at the moment. I don't believe I have caught any good predictive analytics features in comparison to other analytics tools like Amplitude.” ~ Sasha Helms, Sr. Associate, Analytics Development

“...while it takes longer, IMO you're better off just finding the data you want and then using that as your training data set and giving it to chatgpt with the right prompt to do forecasting / predictive modeling accordingly.” ~ Jonathan Wehausen, Client Lead, Digital Measurement Solutions

“When I think about AI in GA4, all it really does is point you to the reports you can use to get that info. Ideally what I'd want to see is a feature that allows you to ask questions and have a report generated around them. For Seer it would be questions like: ‘show me which pages declined month over month in the organic channel’, or ‘which pages contributed the most to conversions in December’, or ‘"Build me a report that shows pages which have been growing in traffic each month throughout the year’. This would allow an analyst to actually pull insights out quickly.” ~ Evan Cohen, Manager, Marketing Analytics & Enablement

I personally found the “Insights” feature to be marginally valuable. The outputs do report metrics, yet there are no interactive capabilities. Insights also only affect the right margin and main charts don’t change. This could be by design, but I still believe this has a long way to go. Needless to say, I don’t think that GA4’s “Insights” will be replacing Analysts any time soon. 

Anomaly Detection

Anomaly detection has long been a feature within Analytics tools. Adobe Analytics first introduced its anomaly detection in 2014. This feature was a part of Adobe Analytics' suite of predictive analytics tools. Anomaly detection in Adobe Analytics utilizes statistical modeling to automatically identify data points that deviate significantly from expected patterns or trends in a dataset. This capability was woefully underwhelming when first introduced. Undoubtedly, it has improved markedly over that past decade. But Google too has jumped aboard the anomaly train with its anomaly detection baked into Intelligence. According to Google’s website: 

Google’s GA4 Analytics Intelligence applies a Bayesian state-space time series model to the historic data to predict the value of the most recent datapoint in the time series. The model produces a prediction and a credible interval that we use to evaluate the observed metric. We use this feature regularly to alert us of traffic spikes, website crashes, and viral episodes. This type of machine learning applies to businesses of all types across industries. Imagine if you will, a financial institution that’s interested in measuring the new student online enrollment journey. A data analyst at the university is likely monitoring engagement levels on the institution's online course selection platform. The Analyst is alerted to a spike in page views of the university’s AI Learning programs. This information could inform the number of professors required and the size of classroom needed to accommodate enrolled students.   

Audience Segmentation

Any digital advertiser knows that finding your audience is critical to marketing success. AI is taking the guesswork out of segmentation by analyzing dimensions, metrics, and events to identify infinite combinations of users with similar preferences. Predefined audiences include all users and users that have made a purchase by default. As new data is added to audience members, the audiences are reevaluated to assure that each member of the audience meets the segment criteria thereby increasing likelihood of conversion. Check out my colleague Mike Sarnoski’s post on Understanding Your Users in GA4 for more info. 

To begin building audiences within GA4, Google does make it easy for you. Simply navigate to Property Settings > Data display > Audiences. There you will find a number of General Audiences ready for use. GA4 also provides the ability to build audiences by Internet & Telecom, via Templates, or using the Predictive AI functionality. 

• GA4 General Audiences:
  • Recently active users
  • Purchasers
  • 7-day inactive purchasers
  • Non-purchasers
  • 7-day inactive users
• Internet & Telecom Audiences: 
  • Registered Users
  • Billable Users
  • Tutorial finishers
  • Tutorial abandoners
• Template Audiences:
  • Demographics
  • Acquisition
  • Technology
• Predictive Audiences:
  • Likely 7-day purchasers
  • Likely 7-day churning users
  • Predicted 28-day top spenders
  • Likely first-time 7-day purchasers
  • Likely 7-day churning purchasers 

However, be aware that you will have to meet the eligibility requirements for volume of relevant data and model quality to use the Predictive Audiences. If you don’t, then you’ll experience errors like in the screenshot below. 

Figure 2: GA4 Audience Setup

Predictive Analytics in GA4

Similar to all AI models, you have to train your GA4 predictive analytics to operate effectively. Google provides the following requirements for predictive analytics accuracy. 

• 1,000 returning users triggered the relevant predictive condition (purchase or churn) 

• Model quality must be sustained over a period of time to be eligible.

• To be eligible for both the purchase probability and predicted revenue metrics, a property has to send the parameters for that event.

According to one member of the Seer Team, Predictive Audiences may be limited in its use cases to Analysts and Marketers that meet very specific criteria. And even those users may find limitations. 

“Predictive Audiences are really geared towards ecommerce clients (purchase probability, churn probability, and predicted revenue) If you aren't in the ecommerce space, purchase probably and predicted revenue will not work for you. Also if you have fewer than 1000 purchasers each month these aren't going to work so smaller companies and those were purchases may not be their main conversions are also out. From a churn probability this one I question how useful this will be. This takes the number of people that were active on the site in the last 7 days and predicts how many of them will not be active in the next 7 days. I don't know that there is a way to adjust the 7 day window and that may or may not work with everyone's sales cycle.” Jessica Propst, Sr. Manager Digital Measurement Solutions. 

Figure 3 below depicts an output, which produced a Prediction Summary that is configurable. 

Figure 3: GA4 Predictive Audiences

Google Analytics will very likely continue to improve its AI integrations and capabilities. Despite the intent, I’ve found the AI outputs to be somewhat clunky and full of limitations when utilizing the tool.

AI in Analytics Use Cases and Applications

If you’re not using Google Analytics for your AI inspiration, the plethora of alternative options is exploding as I type. Numerous AI in Analytics tools aimed at specific use cases are emerging by the minute. We’ve identified four distinct areas of analytics that have seen the greatest tool proliferation and provide this as a cursory list of emerging technologies. 

Generating Code & Debugging

AI significantly aids digital analysts in code generation and debugging, enhancing their work in several ways:

• Automated Code Writing: AI can write basic code structures, making the initial coding process faster and more efficient.
• Error Detection and Correction: AI algorithms quickly identify and suggest fixes for bugs, reducing debugging time.
• Learning from Patterns: Over time, AI learns from past coding errors, improving the precision and reliability of both code generation and debugging.

In a digital marketing context, this means analysts can devote more time to creative and strategic tasks, confident that AI is handling the technical heavy lifting with increasing accuracy and efficiency.

Vendors that offer AI coding and debugging assistance include: DataCamp Workspace AI, Anaconda Assistant, Jupyter AI, and GitHub Copilot. Microsoft 365 Copilot and MutableAI

Exploratory Analysis & Insights

AI supports exploratory analysis and insights in digital marketing by automating data sifting, identifying patterns, and providing predictive insights. This technology enables digital analysts to:

• Recognize Patterns: AI quickly identifies trends and patterns in vast data sets, saving time in exploratory analysis.
• Predict Behaviors: AI forecasts future trends based on historical data, allowing for proactive strategy development.
• Visualize Data: AI tools often include advanced visualization capabilities, making complex data more accessible and understandable.

These AI capabilities allow digital analysts to focus on strategic decision-making and creative problem-solving, backed by data-driven insights.

Vendors that offer AI Exploratory Analysis & Insights include: Tableau GPT, Microsoft Power BI, Polymer, Akkio, AnswerRocket, Equals, and Julius AI

Google Sheets AI integrations

It’s worth noting that Google Sheets (and Excel), arguably the world’s most common analysis tools, are also getting into the AI game via plugins and add ons. While Google Sheets itself doesn't have built-in AI tools for visualization, you can integrate third-party AI tools that enhance data analysis and visualization. Here are a few options:

Google Cloud AI Services: Google Cloud offers various AI services that you can integrate into Google Sheets, such as Google Cloud Vision API for image analysis or Google Cloud Natural Language API for text analysis. These won't directly create visualizations but can provide insights for your data.

• DataRobot: DataRobot's AI for Excel add-on allows you to build and deploy machine learning models directly within Excel or Google Sheets. While it's more focused on predictive modeling, it can assist in analyzing and interpreting your data.

• Supermetrics: Supermetrics is an add-on for Google Sheets that allows you to pull data from various sources, including AI and machine learning platforms. It's particularly useful if you want to combine AI-generated insights with your existing data.

• Tableau Extension: Tableau is a powerful data visualization tool. While it's a separate platform, you can use the Tableau Extension for Google Sheets to embed Tableau visualizations directly into your Google Sheets.

• AI/ML Add-ons: Explore Google Sheets add-ons that focus on AI or machine learning. While not as comprehensive as standalone tools, some can assist with data analysis. For example, "AISheety" is an add-on that brings machine learning capabilities to Google Sheets.

Creating Synthetic Data

AI supports the creation of synthetic data, a crucial aspect in digital marketing analytics. This capability is especially beneficial for testing marketing strategies or analyzing potential customer responses without relying on extensive real-world data collection. Here are a few use cases:

• Generating Realistic Data: AI can create data that simulates real-world customer behaviors and patterns, ideal for testing and analysis.
• Enhancing Privacy: Synthetic data generation helps maintain customer privacy by creating anonymized datasets.
• Filling Data Gaps: AI can generate data to supplement areas where real data is scarce, enabling more comprehensive analysis.

These training datasets can be generated and fed into machine-learning models. This can be done through either free tools like ChatGPT or paid tools like Mostly AI, Jentis or Gretel AI.

Visualizations, Dashboards and Reports

While many will argue that dashboards and visualizations are more art than science, AI can definitely play a role. AI enhances the creation and management of visualizations, dashboards, and reports for digital analysts in several key ways:

• Automated Data Visualization: AI tools can automatically generate visually appealing and informative graphics, making data interpretation faster and more intuitive.
• Customizable Dashboards: AI enables the creation of dynamic, user-specific dashboards that update in real-time, providing tailored insights.
• Intelligent Reporting: AI can analyze data to highlight key insights and trends, generating concise, focused reports.

I will admit that my first hand experience using AI to generate charts has been less than stellar. Yet, these tools are improving every day. AI-driven tools have massive potential to help digital analysts quickly grasp complex data, tailor presentations to specific audience needs, and make informed decisions based on up-to-date, accurate information.

Here are a few tools that will help you get there: use the Midjourney AI to generate some eye-catching ideas for dashboards; Luzmo’s AI-powered dashboard builder, Hex, or go old school and automate data visualization tasks with ChatGPT.

Can AI in Analytics be Trusted?

In the realm of digital marketing, the trustworthiness of AI in analytics is a complex and evolving topic. A study by MIT Sloan Management Review highlights the importance of understanding the limitations of AI in Analytics. Their take is that AI model failures may be difficult to spot, thereby increasing the importance of both validating AI-generated outputs as well as frequently revisiting models to assure accuracy and viability. According to the article, AI developers must gauge a model’s ability to work into the future and beyond the limits of its training data sets — a concept they call generalizability. (MIT Sloan Management Review, 2023). 

This underscores the need for caution and comprehension in AI deployment.

Key Concerns:

1. Bias in Data: AI can perpetuate biases present in the training data.

2. Lack of Transparency: The complexity of AI algorithms may obscure the reasoning behind their outputs.

3. Over-reliance on Automation: Excessive dependence on AI can lead to missing nuanced, human insights.

4. AI Hallucinations: AI can sometimes generate false or misleading information based on flawed data or algorithmic errors.

5. Data Security and Privacy: The use of AI in handling large datasets can raise concerns about data security and the protection of user privacy.

Practical Advice:

• Regularly audit and cleanse data to minimize bias.

• Balance AI insights with human judgment.

• Educate teams about AI capabilities and limitations.

• Validate AI-generated insights against real-world data to detect and correct "hallucinations".

• Implement robust data security measures and adhere to privacy regulations to protect sensitive information.

• Regularly update AI models and data handling practices to reflect the latest security standards and ethical guidelines.

As Harvard Business Review suggests, "...when it comes to AI, transparency is not only about informing people when they are interacting with an AI, but also communicating with relevant stakeholders about why an AI solution was chosen, how it was designed and developed, on what grounds it was deployed, how it’s monitored and updated, and the conditions under which it may be retired." (Harvard Business Review, 2022). This approach, combined with a balanced use of AI and human expertise, can harness the true potential of AI in analytics for digital marketing.

Are You Ready for AI in Analytics?

AI in Analytics offers transformative potential in digital marketing, but with nuanced challenges. As highlighted throughout this post, AI serves as a creative tool, augmenting human ingenuity rather than replacing it. 

From automating routine tasks to enabling sophisticated market intelligence, AI empowers marketers and analysts with actionable insights and enhanced efficiency. However, trust in AI requires balancing its strengths with vigilance over data biases, transparency, and security. Embracing AI in Analytics means engaging in a continuous process of learning, validating, and adapting, ensuring its integration enhances rather than overshadows human expertise. 

As we navigate this AI-augmented landscape, the collaboration between human creativity and AI capabilities will shape the future of digital marketing analytics.