How to build and launch a product in 6 days
or: the making of an AI Shopping Assistant
Hello everyone,
This is Lior, and in today’s issue of the newsletter, I’m going to cover my decision to build an AI Shopping Assistant for Amazon and launch it within six work days. Stay tuned for the inside story.
The previous launch
A couple of weeks ago, I launched DreamGift, an AI assistant that helps you find the perfect give for every person in your life.
The launch on Product Hunt did not go as well as I had hoped, as I didn’t make it to the Top 10 products that day. To my surprise, though, the actual outcomes were good:
DreamGift received 135 votes on Product Hunt on launch day and landed as the #15 product of the day. I aspire to get to the top 10 because every morning, Product Hunt emails the Top 10 Products to hundreds of thousands of subscribers, which provides a nice push in terms of exposure and traffic. Historically, whenever I reached the top 3 on Product Hunt, I witnessed thousands of visitors in one day, the equivalent of a TechCrunch PR.
I didn’t do any media outreach, but DreamGift was organically featured on Ben’s Bites (an AI newsletter with 50K followers) and got a feature article in SAPO in Portuguese. That was a delightful surprise.
On launch day and the following week - 990 users visited the DreamGift site, about 268 from Ben's Bites, 206 from Product Hunt, and 127 from SAPO.
There were 300 clicks to Amazon and two items ordered. Not surprising since users from Product Hunt come to check out the product and not necessarily with the intention of buying someone a gift.
Since any product launch usually involves a large spike in traffic which declines after a few days, I was looking for a sustainable distribution model. For DreamGift, I believe that might be SEO, so my #1 goal for the launch was getting 20 organic referring domains so the site can rank on long-tail uncompetitive keywords.
Again, to my surprise, the launch did deliver: the site grew from 5 referring domains before the launch, to 26 referring domains after. Not bad.As with any launch, I got additional feedback from users, which is always invaluable!
Do I wish it made it to the top 10? Of course I do. But at the same time, the launch accomplished my main goal of getting good organic backlinks, which should kickstart its growth. It’s important to note that it might take months until the site starts ranking on key terms on Google.
The eureka moment
I spent the week following the DreamGift launch in ideation mode, assessing some of my existing product ideas, and looking at new ones at well.
Since I already have several products that rely on SEO for distribution, and on Amazon Affiliates as a business model - I decided I want the next product to utilize a different distribution model and business model. Just like managing a portfolio of stocks, bonds, and other financial assets - I’d like to have a more diversified portfolio of products.
I was still debating between several ideas and was about to go into market research to determine the feasibility of these ideas.
But then, last Sunday, I had this Eureka moment.
Amazon Prime Day is coming up (July 11-12), and it’s a GREAT opportunity to launch a product related to shopping since many people are in a shopping mindset during Black Friday, Prime Day, etc.
Luckily - one of the ideas I was considering is an AI Shopping Assistant extension for Amazon!
Sometimes you have to say “fu** it” and just do it.
While analyzing and optimizing for the right decision is something that I love greatly, sometimes being decisive and seizing the right opportunity in front of you is the better move.
As an indie hacker, there’s something to be said for a propensity toward action and speed.
Who knows - this might have been the right decision, or the wrong one, but since I’m going to launch one product per month, there’s very limited downside to just going for it.
Working under extreme time constants
There was just one, tiny problem:
That Sunday was July 2nd, and Prime Day is July 11-12.
That’s six work days to design it, build it, test it, get it approved in Chrome Web Store, brand it, prepare marketing collateral, and launch a product on Product Hunt.
I decided to use that as a forcing function and take that chance - if I’m going to make it, then great! If not, at least I’ve built up great momentum, and I can launch it the following week.
The jury is still out, but today is day #3 and I’m optimistic - I’ll keep you posted next week, closer to Prime Day.
Where do I start?
It was clear that starting from scratch to launching a product in 6 days is very tight.
My #1 concern was the external risk: any Chrome extension must be uploaded to the Chrome Web Store and reviewed. This may take a few days, or sometimes even more, and the extension could get rejected, which means waiting for a few additional days, which I don’t have with this timeline.
So - I set a goal to upload SOMETHING to the Chrome Web Store within the first couple of days, so we can get the ball rolling. Once the extension is approved, it’s live, and I can optionally upload an updated version.
Also, I build the app in a way that most of the logic and UI for the Assistant is on my servers, and not inside the browse extension. The extension essentially loads the assistant in an isolated frame within the Amazon site and sends it only the relevant data. This approach has two benefits:
Keeping the extension code extremely simple and minimalist (one JavaScript file, one CSS file), which potentially means faster review times and fewer reasons to be rejected.
This allows me to push quick fixes and improvements before the launch, without needing to wait for another review and approval.
The first couple of days
On day #1, I was busy building the fundamentals:
A Chrome Extension that recognizes Amazon product pages, and is able to inject a frame into the product page, and relay any necessary data from the product page into the frame.
A web application that serves that frame, builds a GPT prompt, queries the OpenAI API, and shows the result.
As you can witness, day 1 version was extremely ugly, but it was a proof-of-concept:
On day #2, I made sure the frame loads immediately upon page load, and built a more reasonable look and feel. I had to improve the extension to relay all the relevant data and fix various bugs and issues (the result of building a product in one day).
That same day, for the first time, I set up a cloud environment on Vercel to host the application, and made sure the extension can work both locally and with the production (cloud) environment.
Finally, on the night of day #2, I got to the point where the extension and web app were reasonable enough, so I can upload the extension to the Chrome Web Store for review:
Now we play the waiting game…
Good morning!
Today, I woke up to the greatest email:
To my delight, the extension was reviewed and approved overnight!
It’s now LIVE on the Chrome Web Store, with zero users, and zero reviews!
Once you install it, you can visit any product page on Amazon.com and get an immediate summary of user reviews - as a concise list of pros and cons.
If you’re feeling adventurous today, here’s early access to the extension, a week before it’s officially launched:
Take the extension for a ride on your favorite products on Amazon.com, and let me know your thoughts.
What other AI insights would be useful for you on Amazon?
How could the extension be improved?
What would be a good name for the extension (SmartBuy sounds like something from the 90’s)
Reply to this email and let me know your thoughts 🙏
- Lior
Hi, I have been a long-time user of Keepa and am really excited to use this extension. However, it doesn't work on the IN domain. Are there plans to extend it to the other Amazon domains?