CitizenShipper
just words AI lifts 22% email open rates for CitizenShipper
CitizenShipper used just words AI to personalize emails dynamically for each of their users, bringing a whopping 22% increase in open rates in 3 weeks.
Introduction
Introduction
Introduction
Context
Context
Context
CitizenShipper is a two-sided marketplace for hard-to-ship items, mainly pet shipments. It is one of the most precious transportation, which is why offering trust, safety, and professionalism is top of mind for CitizenShipper. One of the many ways CitizenShipper brings comfort and peace of mind to its customers is through on-point, personalized, and ongoing communications via email and SMS throughout the shipment journey.
Introduction to just words AI and how it helped CitizenShipper
just words enables consumer businesses to place dynamic content in front of their users through continuous, self-learning experimentation cycles. just words analyzed CitizenShipper's branding & tone of voice, and customer profile, blended it with industry best practices, and fine-tuned a pool of copies for high open rates.
CitizenShipper is a two-sided marketplace for hard-to-ship items, mainly pet shipments. It is one of the most precious transportation, which is why offering trust, safety, and professionalism is top of mind for CitizenShipper. One of the many ways CitizenShipper brings comfort and peace of mind to its customers is through on-point, personalized, and ongoing communications via email and SMS throughout the shipment journey.
Introduction to just words AI and how it helped CitizenShipper
just words enables consumer businesses to place dynamic content in front of their users through continuous, self-learning experimentation cycles. just words analyzed CitizenShipper's branding & tone of voice, and customer profile, blended it with industry best practices, and fine-tuned a pool of copies for high open rates.
CitizenShipper is a two-sided marketplace for hard-to-ship items, mainly pet shipments. It is one of the most precious transportation, which is why offering trust, safety, and professionalism is top of mind for CitizenShipper. One of the many ways CitizenShipper brings comfort and peace of mind to its customers is through on-point, personalized, and ongoing communications via email and SMS throughout the shipment journey.
Introduction to just words AI and how it helped CitizenShipper
just words enables consumer businesses to place dynamic content in front of their users through continuous, self-learning experimentation cycles. just words analyzed CitizenShipper's branding & tone of voice, and customer profile, blended it with industry best practices, and fine-tuned a pool of copies for high open rates.
Challenge
Challenge
Challenge
The Problem
The Problem
The Problem
a. Previous email efforts: CitizenShipper has a robust setup in place for running campaigns for different types of emails (eg: welcome, and discount emails). All praises for their AI Officer, Cagri Sarigoz, (1) for setting up robust integrations with customer.io, and (2) for building personalization in messaging. Overall, CitizenShipper had a strong infrastructure for a robust data-driven marketing strategy.
b. Need for Innovation: While CitizenShipper had a solid experimentation and campaign strategy, there was limited content intelligence baked into the product. Specifically, this is where various other AI tools fell short:
1. Content evaluation and curation: As a pet shipment company that depends on users' trust, it was non-negotiable for them to produce messaging that is not uniquely tailored to each customer. AI-generated content was generic at best. There were no objective evaluation criteria to decide what's a good email vs not, which made it challenging to use AI for plugging in new content in their emails.
2. Dynamic personalization: Different users expect a different flavor of messaging, depending on their circumstances. While it was easy to create personalized templates in advance, creating personalized messages in real time wasn't easy.
3. Performance-based presentation: Another big challenge was to select the right copy based on how it has performed in the past. In other words, it wasn't easy to scientifically test the performance of content and auto-optimize the subsequent copy.
a. Previous email efforts: CitizenShipper has a robust setup in place for running campaigns for different types of emails (eg: welcome, and discount emails). All praises for their AI Officer, Cagri Sarigoz, (1) for setting up robust integrations with customer.io, and (2) for building personalization in messaging. Overall, CitizenShipper had a strong infrastructure for a robust data-driven marketing strategy.
b. Need for Innovation: While CitizenShipper had a solid experimentation and campaign strategy, there was limited content intelligence baked into the product. Specifically, this is where various other AI tools fell short:
1. Content evaluation and curation: As a pet shipment company that depends on users' trust, it was non-negotiable for them to produce messaging that is not uniquely tailored to each customer. AI-generated content was generic at best. There were no objective evaluation criteria to decide what's a good email vs not, which made it challenging to use AI for plugging in new content in their emails.
2. Dynamic personalization: Different users expect a different flavor of messaging, depending on their circumstances. While it was easy to create personalized templates in advance, creating personalized messages in real time wasn't easy.
3. Performance-based presentation: Another big challenge was to select the right copy based on how it has performed in the past. In other words, it wasn't easy to scientifically test the performance of content and auto-optimize the subsequent copy.
a. Previous email efforts: CitizenShipper has a robust setup in place for running campaigns for different types of emails (eg: welcome, and discount emails). All praises for their AI Officer, Cagri Sarigoz, (1) for setting up robust integrations with customer.io, and (2) for building personalization in messaging. Overall, CitizenShipper had a strong infrastructure for a robust data-driven marketing strategy.
b. Need for Innovation: While CitizenShipper had a solid experimentation and campaign strategy, there was limited content intelligence baked into the product. Specifically, this is where various other AI tools fell short:
1. Content evaluation and curation: As a pet shipment company that depends on users' trust, it was non-negotiable for them to produce messaging that is not uniquely tailored to each customer. AI-generated content was generic at best. There were no objective evaluation criteria to decide what's a good email vs not, which made it challenging to use AI for plugging in new content in their emails.
2. Dynamic personalization: Different users expect a different flavor of messaging, depending on their circumstances. While it was easy to create personalized templates in advance, creating personalized messages in real time wasn't easy.
3. Performance-based presentation: Another big challenge was to select the right copy based on how it has performed in the past. In other words, it wasn't easy to scientifically test the performance of content and auto-optimize the subsequent copy.
Approach
Approach
Approach
Hyper-personalization with just words
Hyper-personalization with just words
Hyper-personalization with just words
Copy Pool Generation
Our models generated a pool of 20 high-performing copies to replace the original subject lines and pre-headers. To do this, they combined CitizenShipper's branding, customer profiles, and personalization attributes, with industry best practices. Then, they evaluated the copies against performance factors such as open rates.
Copy Pool Generation
Our models generated a pool of 20 high-performing copies to replace the original subject lines and pre-headers. To do this, they combined CitizenShipper's branding, customer profiles, and personalization attributes, with industry best practices. Then, they evaluated the copies against performance factors such as open rates.
Copy Pool Generation
Our models generated a pool of 20 high-performing copies to replace the original subject lines and pre-headers. To do this, they combined CitizenShipper's branding, customer profiles, and personalization attributes, with industry best practices. Then, they evaluated the copies against performance factors such as open rates.
On-point Personalization
just words then stitched various personalized attributes (eg: name, animal_name, pickup_city, discount_amount) with the static copy.
Optimal use of personalization, i.e., deciding how many and which personalized attributes should be ranked higher and their placement in the copy. Placement became a huge consideration as the subject line truncates differently on different device sizes.
just words analyzed thousands of industry-wide emails and their open propensity, spitting out best practices for email hygiene. Surprisingly, many were counter-intuitive.
On-point Personalization
just words then stitched various personalized attributes (eg: name, animal_name, pickup_city, discount_amount) with the static copy.
Optimal use of personalization, i.e., deciding how many and which personalized attributes should be ranked higher and their placement in the copy. Placement became a huge consideration as the subject line truncates differently on different device sizes.
just words analyzed thousands of industry-wide emails and their open propensity, spitting out best practices for email hygiene. Surprisingly, many were counter-intuitive.
On-point Personalization
just words then stitched various personalized attributes (eg: name, animal_name, pickup_city, discount_amount) with the static copy.
Optimal use of personalization, i.e., deciding how many and which personalized attributes should be ranked higher and their placement in the copy. Placement became a huge consideration as the subject line truncates differently on different device sizes.
just words analyzed thousands of industry-wide emails and their open propensity, spitting out best practices for email hygiene. Surprisingly, many were counter-intuitive.
Smart copy selection
We selected the best copy from the pool of 20 selected copies in real-time uniquely for each user.
To optimize copy selection, we modeled various problem statements such as "Does the same copy repeat for the same user, or is it better to rotate different flavors over time for the same user to prevent fatigue? Can we infer more metadata based on the value of the attribute? Eg: if the drop-off city is Miami, can we speak to the sunny weather in the copy? How do we prevent fatigue from showing the same copy?
Smart copy selection
We selected the best copy from the pool of 20 selected copies in real-time uniquely for each user.
To optimize copy selection, we modeled various problem statements such as "Does the same copy repeat for the same user, or is it better to rotate different flavors over time for the same user to prevent fatigue? Can we infer more metadata based on the value of the attribute? Eg: if the drop-off city is Miami, can we speak to the sunny weather in the copy? How do we prevent fatigue from showing the same copy?
Smart copy selection
We selected the best copy from the pool of 20 selected copies in real-time uniquely for each user.
To optimize copy selection, we modeled various problem statements such as "Does the same copy repeat for the same user, or is it better to rotate different flavors over time for the same user to prevent fatigue? Can we infer more metadata based on the value of the attribute? Eg: if the drop-off city is Miami, can we speak to the sunny weather in the copy? How do we prevent fatigue from showing the same copy?
The results
A 22% Lift in Email Open Rates
Higher Open Rates
22% lift in email open rates
We observed a 22% increase in the email open rates. This lift in engagement demonstrates the power of continuously optimizing and refreshing the message pool.
Messaging insights
Copies that overperformed had distinct observable patterns.
1) Throwing the discount amount upfront annoyed users. Keeping it in pre-headers worked great.
2) Users opened the email more when the tone was more conversational and less sales-like. "Hey Neha, we noticed your post" outperformed "Save $50 on Buddy’s Adventure?"
Balanced personalization
Personalization, along with context wins the race
Too many personalization attributes did not work. Not all of them worked. There was a fine balance of the type and number of elements that did the trick.
The results
A 22% Lift in Email Open Rates
Higher Open Rates
22% lift in email open rates
We observed a 22% increase in the email open rates. This lift in engagement demonstrates the power of continuously optimizing and refreshing the message pool.
Messaging insights
Copies that overperformed had distinct observable patterns.
1) Throwing the discount amount upfront annoyed users. Keeping it in pre-headers worked great.
2) Users opened the email more when the tone was more conversational and less sales-like. "Hey Neha, we noticed your post" outperformed "Save $50 on Buddy’s Adventure?"
Balanced personalization
Personalization, along with context wins the race
Too many personalization attributes did not work. Not all of them worked. There was a fine balance of the type and number of elements that did the trick.
The results
A 22% Lift in Email Open Rates
Higher Open Rates
22% lift in email open rates
We observed a 22% increase in the email open rates. This lift in engagement demonstrates the power of continuously optimizing and refreshing the message pool.
Messaging insights
Copies that overperformed had distinct observable patterns.
1) Throwing the discount amount upfront annoyed users. Keeping it in pre-headers worked great.
2) Users opened the email more when the tone was more conversational and less sales-like. "Hey Neha, we noticed your post" outperformed "Save $50 on Buddy’s Adventure?"
Balanced personalization
Personalization, along with context wins the race
Too many personalization attributes did not work. Not all of them worked. There was a fine balance of the type and number of elements that did the trick.
Conclusion
Conclusion
Conclusion
Hyper-personalization speaks for itself
Hyper-personalization speaks for itself
Hyper-personalization speaks for itself
By placing a heavily personalized and scientifically optimized real-time copy selection algorithm, just words helped CitizenShipper with a 21.9% increase in its email open rates.
By placing a heavily personalized and scientifically optimized real-time copy selection algorithm, just words helped CitizenShipper with a 21.9% increase in its email open rates.
By placing a heavily personalized and scientifically optimized real-time copy selection algorithm, just words helped CitizenShipper with a 21.9% increase in its email open rates.
Just Words
Product
CUSTOMERS
Just Words
Product
CUSTOMERS
Just Words
Product
CUSTOMERS