DTC+: AI & CRO

Hey everyone,

Product descriptions.

I’ve seen very few brands consciously and strategically optimize their product descriptions.

A vital real estate on any product page that besides explaining what the product is (ideally clearly) can really do a lot of selling for you.

There’s a big opportunity for that area to be continuously optimized based on customer feedback.

Being on the cutting edge of AI (who isn’t these days) we ran an experiment that yielded huge results for a client, which I want to share with you today, so you can use it as inspiration to make your product description SELL, not only explain.

We wanted to test whether an AI written product description based on client reviews would outperform the human written one we had on there for the top selling product.

These were the exact steps in order:

1. Export customer reviews

Thankfully, our client had loads of reviews (about 8K of them), so we had a lot of data to work with, but I believe results can be achieved with much less.

The key insight we are looking for is the language your customers are using to talk positively about your product and then leverage those exact words to improve your product description.

Resonate to dominate.

2. Import customer reviews into ChatGPT

Here’s the exact prompt I started the chat with:

It then gave me a list of 10 aspects customers appreciate about our product.

I followed that response with:

“Based on the 10 positive points, write a one paragraph product page description for my product, give me 4 variations“.

3. Run a 4 variation test vs your current product description

The rest was easy, we span up a 4 variation test to see whether an AI written product description (based on real world data though) would beat the neglected copy we had in there.

*Remember to name your variations clearly, in case you get a winner, so you know exactly what to deploy. We named them based on the first three words from the variation copy.

Drum roll and results after running approximately 800 orders through each variation:

The test resulted in a projected 15% lift to revenue for the top selling product.

And this is only just by literally copy-pasting what ChatGPT gave us from a chat that took no longer than few minutes.

There could possibly be further wins if a good copywriter goes through the suggestions/data and writes an even better description (good thing to test), but what the above shows us is that a product description based on real-world customer data/language will vastly outperform one that is written only to explain a product.

Give it a try and let me know how it goes 🤖✍️

Have a great week everyone, see you next Monday AM.

Julian

PS - What it feels like the first time using AI to write product descriptions

Share this newsletter with someone who will find it valuable/interesting.
This was sent to you? Sign up here.