Suggestions
Or Arbel
Co-Founder And Chief Technology Officer at Anima App
Highlights
𝐒𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐢𝐧 𝐧𝐞𝐰𝐬𝐥𝐞𝐭𝐭𝐞𝐫𝐬 𝐢𝐬 𝐚 𝐭𝐡𝐢𝐧𝐠 𝐨𝐟 𝐭𝐡𝐞 𝐩𝐚𝐬𝐭
TIMES SQUARE, NEW YORK – 2050 — The protagonist is walking down the street.
Electronic billboards display hyper-personalized ads specifically tailored for him. —
We’ve seen this in the movies, but the technology is already here. Except instead of billboards these are newsletters.
𝐍𝐞𝐰𝐬𝐥𝐞𝐭𝐭𝐞𝐫𝐬 𝐚𝐫𝐞 𝐚𝐛𝐨𝐮𝐭 𝐭𝐨 𝐦𝐚𝐤𝐞 𝐚 𝐜𝐨𝐦𝐞𝐛𝐚𝐜𝐤
Newsletters are a common marketing practice. Those of us fortunate enough to curate a monthly newsletter have experienced a unique challenge.
It involves mixing and matching the best content to engage the entire audience.
This process is much like solving a sudoku puzzle.
In the end we usually pick 3 articles where each will speak to a different target audience.
The more skilled among us segment the audience in Mailchimp. We then send different content to each segment.
But segmenting takes time. It takes time to build the initial segments and to actively maintain them. And don't forget about the rules, ugh, so many rules.
In the end, instead of sending the same email to everyone, we send maybe 6 versions. Perhaps even 10 versions. And we call it personalization. How naive of us.
I recently used the UI for creating segments and rules. It occurred to me that this interface is a thing of the past. In fact, there should be no such UI at all.
What if, instead of creating 6-10 segments, we send a hyper-personalized newsletter to each subscriber?
No need to set up segments and rules. We construct a prompt with the following structure: 𝐹𝑜𝑟 𝑎 𝑛𝑒𝑤𝑠𝑙𝑒𝑡𝑡𝑒𝑟 𝑓𝑜𝑟 𝑜𝑢𝑟 𝑝𝑟𝑜𝑑𝑢𝑐𝑡 𝑦𝑜𝑢 𝑤𝑖𝑙𝑙 𝑏𝑒 𝑔𝑖𝑣𝑒𝑛 𝑎 𝑙𝑖𝑠𝑡 𝑜𝑓 𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠. 𝐼𝑛 𝑎𝑑𝑑𝑖𝑡𝑖𝑜𝑛 𝑦𝑜𝑢 𝑤𝑖𝑙𝑙 𝑏𝑒 𝑔𝑖𝑣𝑒𝑛 𝑡ℎ𝑒 𝑟𝑒𝑐𝑒𝑛𝑡 𝑒𝑣𝑒𝑛𝑡𝑠 𝑜𝑓 𝑎 𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝑢𝑠𝑒𝑟.
𝐵𝑎𝑠𝑒𝑑 𝑜𝑛 𝑡ℎ𝑒 𝑢𝑠𝑒𝑟 𝑒𝑣𝑒𝑛𝑡𝑠, 𝑟𝑒𝑝𝑙𝑦 𝑤𝑖𝑡ℎ 𝑡ℎ𝑒 𝑡𝑜𝑝 3 𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠 𝑡ℎ𝑎𝑡 𝑤𝑖𝑙𝑙 𝑏𝑒 𝑚𝑜𝑠𝑡 𝑟𝑒𝑙𝑒𝑣𝑎𝑛𝑡 𝑡𝑜 𝑡ℎ𝑎𝑡 𝑢𝑠𝑒𝑟.
𝐼𝑛 𝑎𝑑𝑑𝑖𝑡𝑖𝑜𝑛, 𝑝𝑟𝑜𝑣𝑖𝑑𝑒 𝑡ℎ𝑒 𝑐𝑜𝑛𝑡𝑒𝑛𝑡 𝑜𝑓 𝑡ℎ𝑒 𝑛𝑒𝑤𝑠𝑙𝑒𝑡𝑡𝑒𝑟 𝑡ℎ𝑎𝑡 𝑒𝑥𝑝𝑙𝑎𝑖𝑛𝑠 𝑡𝑜 𝑡ℎ𝑒 𝑢𝑠𝑒𝑟 𝑤ℎ𝑦 ℎ𝑒 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑑 𝑡ℎ𝑒𝑠𝑒 𝑟𝑒𝑐𝑜𝑚𝑚𝑒𝑛𝑑𝑎𝑡𝑖𝑜𝑛𝑠.
𝐿𝑖𝑠𝑡 𝑜𝑓 𝑎𝑟𝑡𝑖𝑐𝑙𝑒𝑠:
𝑈𝑠𝑒𝑟 𝑒𝑣𝑒𝑛𝑡𝑠:
We use this prompt for every subscriber and get back a hyper-personalized email that is unique for them.
𝐖𝐚𝐢𝐭 𝐚 𝐦𝐢𝐧𝐮𝐭𝐞! 𝐛𝐮𝐭 𝐡𝐨𝐰 𝐦𝐮𝐜𝐡 𝐝𝐨𝐞𝐬 𝐢𝐭 𝐜𝐨𝐬𝐭?
Great question. A prompt with a list of 100 articles is about 5000 tokens. With GPT-4o that comes out to 3 cents per OpenAI API call.
For an email list of 10,000 subscribers, sending a hyper-personalized email will cost $300.
Is it worth it? It depends.
First, we assume this cost will go down as prices of LLM inference is going down dramatically over time.
Second, we'll need to track the uptick in click-through rate and BOFU conversion rates.
So... will newsletters make a come back?
#AIMarketing #EmailMarketing
