Unlocking email success: Using GenAI to analyze and enhance your campaigns

Email marketing remains a powerhouse in digital communication, but crafting the perfect email isn’t always a straightforward process. Conversion rates can fluctuate and leave us wondering what went right or wrong. In this two-part series, we’ll explore how generative AI, specifically ChatGPT, can help analyze and improve email campaigns. In the first part, we’ll analyze six of our email newsletters — three with stellar conversion rates and three that didn’t perform as well.

GenAI for in-depth email analysis

To gain comprehensive insights into our emails, we decided to use ChatGPT-4o for the analysis. This advanced GenAI model can read links and images, which allows it to assess not only the text but also the design elements of emails. During the analysis, we uploaded PNG images of the emails into the chat and provided direct links to them. ChatGPT-4o was able to perform a detailed examination of every aspect of the emails’ content, including the layout, color schemes, and visual appeal.

The importance of crafting effective prompts

When working with generative AI, such as ChatGPT, the quality of the output is heavily influenced by the prompts we provide. Thus, crafting effective prompts was essential to guide GenAI in producing the most useful and relevant analysis of our emails.

How to compose effective prompts:

• be specific: Clearly state what you want the GenAI to do. Provide detailed requests instead of vague instructions;
• include directives: Use action-oriented language that tells the GenAI exactly what you need;
• encourage detailed responses: Ask the GenAI to think thoroughly or conduct a comprehensive analysis;
• set the context: Provide background information or specify the role you want the GenAI to assume — for example, “I want you to act as a professional email marketing specialist at Stripo.”

An insight on prompt phrasing:

Interestingly, research indicates that the way we phrase our prompts can significantly impact the quality of the AI’s response. For example, starting a prompt with “Deep breath” can lead to more detailed and thoughtful answers compared to phrases like “Let’s think step by step.”

To test this claim, an experiment was run with ChatGPT:

1. Prompt: Tell me about ____.
2. Prompt: Deep breath. Tell me about ____.
3. Prompt: Tell me in detail ____.

The responses to Prompts 1 and 3 were shorter and similar in structure. However, Prompt 2 elicited a more expansive and differently structured reply, which we found to be subjectively better.

(Source: Large Language Models as Optimizers)

Analyzing the top-3 high-converting emails

To uncover what made our top-performing emails so successful, we conducted a thorough analysis using ChatGPT. With this tool, we could examine the content, design, and overall strategy of each email. We focused on identifying the strengths and potential weaknesses to understand the key factors driving high conversion rates.

Our first digest with good conversion:

(Source: Stripo’s weekly digest)

ChatGPT analysis of our email digest:

Design

• visual appeal: The header features a friendly opening line, “Welcome to our Weekly Digest.” It is accompanied by a mailbox icon, which adds relevance to the newsletter theme. The clean, modern layout includes ample white space and a color palette of greens and blues, which creates a professional yet inviting feel. Subtle icons and images, such as the mailbox and the International Women’s Day illustration, reinforce the subject matter without overwhelming the reader. Consistent design elements from the header to the CTA buttons enhance the cohesion and visual flow.

Text/copywriting

• header: While clear, the header could be made more engaging with personalization (e.g., “Here’s what’s new this week in your inbox!”) to evoke excitement and connection.
• subheadline and content: The concise introduction explains the focus of the email but could use more dynamic language to evoke emotion — for example, “Celebrating the inspiring achievements of women in email marketing and beyond.” The section on “Women influencers on gender equality and challenges” is well presented with key points that are easy to digest, but it could benefit from the use of more empowering language to motivate readers;
• CTA buttons: To encourage higher engagement, the action-oriented CTAs, like “Start Exploring” and “Read Article,” could be strengthened with more personalized or urgent phrasing, such as “Explore Women’s Impact Now” or “Read Powerful Insights. "

Strengths

• clean and modern design: The email’s visual layout is professional and easy to navigate, and the effective use of color, space, and simple icons enhances readability.
• clear structure: The email is organized into easily digestible sections, with headlines, bullet points, and short paragraphs offering a seamless reading experience;
• relevant content: The focus on women in marketing and gender equality is timely and resonates with audiences who are interested in topics of diversity and empowerment.

Weaknesses

• tone and engagement: The tone could be more personalized and engaging. Adding excitement or urgency could inspire immediate action;
• CTAs: The generic CTA buttons could be improved by adding urgency or personalization to boost engagement;
• visual impact: While professional, the visuals are minimalistic. Incorporating bolder, more impactful imagery could capture more attention, especially on important topics like gender equality.

From our analysis of the first email, we gained valuable insights into what resonated with our audience and drove high engagement.

We repeated the same process with the next two emails (email 2 and email 3). The results were presented consistently, with ChatGPT providing feedback in the same format, sequence, and structure for each email. The GenAI offered detailed advice about strengths and weaknesses, covering aspects such as subject lines, personalization, content quality, visual elements, CTAs, and mobile optimization. The consistency of the results allowed us to compare the emails directly and identify the common factors contributing to their high conversion rates.

Analyzing 3 lower-converting emails

After gaining valuable insights into our top-performing emails, we applied the same analytical approach to assess three emails that didn't fare as well in terms of conversion rates. Using ChatGPT, we meticulously examined each aspect of these emails to identify the areas where they fell short.

Our first digest with a weak conversion rate:

(Source: Stripo’s inspiring digest)

Analysis of our email digest (with lower conversion):

Design

• visual appeal: While the email maintains consistent branding with the green mailbox icon at the top, the overall design feels flat compared to the previous high-performing emails. The visual elements are less dynamic, especially in the Chelsea FC and Yakaboo sections;
• lack of visual impact: The design is overly simplistic. The Chelsea FC section could have used team-related imagery to increase engagement, and the Yakaboo section could have featured book-themed images to better represent the brand;
• overuse of white space: The excessive white space makes the design feel sparse and lacks vibrancy and energy, especially given the exciting subject of success stories.

Text/copywriting

• header: The header “Inspiring Digest” is generic and lacks urgency or a clear value proposition. A more intriguing headline, such as “Uncover How the Pros Inspire with Every Email,” could spark curiosity and action;
• introduction: The introduction is factual but not compelling, and it lacks an emotional hook to entice readers. Reworking phrases to better emphasize the value of the content could create more excitement and urgency;
• Chelsea FC section: The copy doesn’t capture the energy or prestige of the Chelsea brand. More engaging language, such as “Discover how Chelsea FC’s game-changing email strategy won over millions,” could attract more attention;
• Yakaboo section: As in the Chelsea section, the content here is informative but lacks flair. Using language that evokes excitement about reading, such as “Yakaboo turns email into a best-seller experience,” could enhance the emotional appeal;
• CTAs: CTAs like “Read the Case Study” are serviceable but bland. More compelling phrases, like “See How Chelsea Wins with Email” or “Unlock Yakaboo’s Secret to Success” could provide more incentive for readers to click.

Strengths

• clear structure: The email is well organized, with dedicated sections for each case study, which makes it easy to follow;
• consistent branding: The email maintains Stripo’s established color scheme and design structure, which contributes to brand recognition;
• valuable content: The content includes case studies of reputable brands, like Chelsea FC and Yakaboo, which offer strong value to readers who seek inspiration.

Weaknesses

• flat visual design: Because the design lacks dynamic or engaging visuals that capture attention, it fails to evoke excitement about the case studies;
• uninspiring copy: The text is straightforward but lacks emotional appeal, excitement, or urgency, thus missing opportunities to inspire readers;
• generic CTAs: The CTAs are overly generic and do not create a sense of excitement or urgency, so they do not generate the emotional pull needed to encourage clicks;
• excessive white space: The overuse of unused space makes the email feel empty and less engaging.

We observed that the format and structure of the ChatGPT-4o analysis was consistent with the previous evaluations. However, this time, the GenAI focused more intently on identifying the shortcomings and the areas requiring improvement. It delivered detailed critiques on aspects such as design flaws, unengaging content, weak CTAs, and other factors that could contribute to poor conversion rates.

The analysis results for the other two low-converting emails were analogous. ChatGPT-4o consistently highlighted similar issues across these emails, which reinforced the patterns we noticed in the first analysis. The GenAI pointed out problems such as cluttered layouts, lack of personalization, vague messaging, and ineffective CTAs. This consistency in the feedback emphasized the key areas we need to address to enhance our email campaigns.

The difference in email conversion

Now, let’s explore why the first three emails achieved good conversion and views and why the last three were less successful. The same format was used for all six emails. However, the difference in their performance boils down to how effectively each email executed key elements that resonated with the audience.

Here’s a detailed explanation from ChatGPT-4o of why the first three emails outperformed the last three:

Why the first 3 emails achieved high conversion:

Superior visual hierarchy and layout

• clear structure: The clean, well-organized layout with a clear visual hierarchy made the content easy to scan and engage with;
• balanced imagery and text: The effective use of images and text (icons, logos, buttons) enhanced the visual appeal without overwhelming the reader;
• strong header: Attention-grabbing headers, such as “Weekly Highlight,” guided the reader’s focus and emphasized important elements, like CTAs.

Engaging and actionable copy

• direct, benefit-driven language: The clear and engaging language emphasized reader benefits (e.g., “Explore Women’s Impact Now,” “Discover More”), which made the content feel conversational;
• consistent tone and relevance: The professional yet inspiring tone aligned with the email’s purpose and resonated with the target audience;
• use of urgency or intrigue: Phrases like “2024’s can’t-miss email marketing conferences” added a sense of urgency and exclusivity that motivated readers to engage.

Compelling CTAs

• strong CTAs: The clear, action-driven CTAs encouraged immediate clicks;
• strategic placement: The CTAs were strategically placed throughout the emails, which ensured that actionable next steps were always available to the reader.

Why the last 3 emails achieved low conversion:

Flat visual design

• lack of visual impact: The emails were text-heavy with excessive white space, which caused them to appear flat and generic;
• insufficient visual support: Topics like gamification lacked compelling visuals that could enhance engagement.

Weak copy and tone

• casual and bland introductions: The introductions lacked intrigue or urgency and failed to hook readers;
• less focus on benefits: The copy was more descriptive than actionable and lacked motivational language to compel action;
• insufficient storytelling: There was an overreliance on long paragraphs without engaging narratives.

Weak and generic CTAs

• uncompelling CTAs: Basic phrases, such as “Read the Guide,” lacked urgency and clear incentives;
• disjointed CTAs: The CTAs were not fully integrated into the content, which reduced their effectiveness.

Dense text and overcrowded layouts

• text-heavy sections: Dense blocks of text without visual aids made the content feel overwhelming;
• cluttered event promotions: The lack of spacing and visual cues obscured key details, like event dates and discounts.

Lack of emotional appeal or storytelling

• missing storytelling elements: Information was presented mechanically without guiding the reader through a narrative;
• missed personalization opportunities: The copy didn’t speak directly to the recipient, which limited emotional connection.

Disconnected content flow

• unrelated interactive elements: Features like “click to reveal” predictions felt disconnected from the main theme;
• abrupt transitions: Sections didn’t flow smoothly, which made the content feel disjointed.

Summary:

• successful emails: High-performing emails featured a strong visual hierarchy, balanced content, engaging copy, and action-driven CTAs with a focus on value to the reader;
• less successful emails: Low-performing emails suffered from a weak visual design, dense text, generic CTAs, and a lack of engaging, benefit-driven language, which led to lower engagement and conversion.

Wrapping up

In this exploration of using GenAI like ChatGPT-4o to analyze and enhance email campaigns, we’ve delved into the potential benefits and limitations of this approach. While GenAI offered some interesting suggestions, we remain uncertain about the overall success of this endeavor and still have doubts about much of what ChatGPT proposed. Often, its recommendations focused on basic structural improvements that weren’t particularly innovative or unique. Nonetheless, it’s worth spending more time on understanding the process and refining our methods to achieve more acceptable results.

It is important to note that the GenAI analysis is subjective and requires validation. We plan to verify its suggestions through A/B testing in the near future and will share the results in the second part of this article.