First week launch report: plantrips.net
Read on to find more about how I validated my launch hypothesis with data & how we ranked on first page of Google with long tail keywords
Tl;dr: Conversion rate is great but no sign of product-market fit (yet).
Now if you are interested, keep reading on. But first, a recap on plantrips’ value proposition:
Plantrips.net helps travelers generate AI itineraries within 2 minutes instead of age googling and pulling resources together.
Every launch should have a different target.
While getting the top position on ProductHunt is desirable for many indie hackers because it usually means sustaining traffic sources and revenue, topping PH is not the goal for us. Because we operate as a free tool and take a cut from affiliate booking, our goal is simple: generate as much traffic as possible and see if people will convert to our affiliate sites. It’s better to hit a wide range of traffic sources for this goal.
Let’s start with our launch goal.
Hypothesis: People will find click on affiliate link generated by AI.
Measurement of success: 2% conversion rate on all unique visitors
Result: exceed expectation, conversion rate to affiliate partners are at >11%. That means out of 100 people visiting plantrips, 11 of them go to klook/agoda/hotellook.
After didau.co generated $2 on its first launch week, I was a lot more hype about the international launch of plantrips. Even though monetization is not a main goal, I was still excited to dig into the data. However, out of 102 clicks to our affiliate sites, we generated $0.
At this point, I have a few hypotheses:
Quality of traffic is low: we haven’t found the right group of people who need to travel at that point in time
Product failure: affiliate links generated by AI are low quality (404s, not linking to the correct hotel/activity)
So how do we validate these tests?
Product failure
About 35% of all the affiliate clicks are for generic booking like in the picture below:
The remaining 65% converts through links like:
When I did a random click test, 20% of those links are not found - a lack of inventory on partner sites. However, even for links with inventories, sometimes it doesn’t bring to the exact hotel/activity.
This is a problem. Why so? This means users have to spend more effort finding the hotel or activity they want on partner sites. Too much hassle, I wouldn’t blame them for dropping out at this point.
So is there a product quality problem? Highly Yes
Another bigger question to answer is: do users find value in this product? If so, they should be either recommending their friends to use it or simpler, returning to use it the next day.
While I do not have the NPS implementation in place, a quick retention analysis of unique users can be done on Posthog:
Only the cohort on Jun 2 sticks around. Most other group of users drop off after the first day.
Now that’s a big deal.
That tells me the user doesn’t find value to return to the site again.
How about a more specific retention analysis? Let’s look at the number of users creating a trip and returning the next day to create more trips:
A little bit better but not by much.
My saving hope right now? Maybe travel is an activity with low frequency, thus not fit for a daily return report. I could run weekly/monthly retention reports in the future. But not a good sign at all.
So do I conclude that the product is a failure? Let’s look at the fit between visitors & the product next.
Low Traffic Quality
In order to understand the quality of traffic, I run a simple breakdown analysis based on referrer domains:
Our traffic acquisition plan at launch was simple:
Hit as many AI directories as we can.
Promote on a private mailing list from my founder’s previous business (the group focuses on agriculture practices)
Promote on our own social circle (Twitter & Facebook) - very limited size (my Twitter has 70 followers and my co-founder's Facebook has ~180 friends)
Observations:
As expected, more than 60% of traffic comes from these AI directory sites with a clear winner futurepedia.io. This site generated 10x the traffic compared to all other AI sites combined. If you’re building an AI tool, get your app in there.
<10% comes from the private mailing list. And 0 conversion to affiliate partners.
~28% is direct traffic. I highly suspected this is a bug because, for a newly launched page, there’s no way our brand name is that strong. I did shout about the site to all climbers at my local gym Vertical Academy but it should not be that memorable. People need 7 repetitions before they remember your brand, the so call Rule of 7 from marketing. Will need to monitor this further.
Only 2% of traffic comes from organic Google searches. Given that our site only started to appear on Google Index 2 days ago, this is actually a win in my bag. This means some users with travel intent actually landed on the site. The programmatic SEO effort is paying off. If only we have more users to generate even more trips 🤣
What’s the other sign of low traffic quality? Well, high bounce rate at 76%. Here is how many people view a second page after landing on plantrips.net:
What does this tell me?
The message on my home page might not be clear enough. Do people understand the benefit of time-saving from AI travel tools? Here is the existing tagline:
To be honest, while it’s attention-grabbing, it neither wows nor triggers an emotional reaction in visitors.
Thus, my suspicion right now is: many people visit the tool to “try out” the hot new AI future, not because they want to/need to travel at the time. Even if they want to travel, they might not understand the value enough.
A small detour: How we got on the first page of Google for some long tail keywords
Before I built plantrips.net, I was curious about 1 experiment: Could you get AI content to rank on Google?
The result: yes, AI content can be ranked on the first page for certain long tail keywords.
A word of caution: this probably means nothing right now because we’ve yet to optimize for search volume to get meaningful traffic.
How did we do it for plantrips.net?
For each travel brief a user enter, we generated a friendly URL like /trip/<slug> and told AI to create an optimized meta description tag for it. E.g: if a user search for 5 days in Bali with my wife, we create a /trip/5-days-in-bali-with-my-wife URL
We then run a cron job against our database to fetch all these URLs and create a sitemap with those URLs
Submit this sitemap to Google Index
Make sure your server is properly scaled so Google doesn’t encounter 5xx or 4xx when they crawl your sitemap.
Wait for a few days & profit
So why do I think this might work?
As users create more niche trips on our site, more niche pages are created & indexed. This gives us more chances to get traffic through really really specific keywords. After all, while TripAdvisor dominates 5 best things to do in Jharsuguda, I don’t think they will bother creating content for 1 day in busan Korea, please include exciting activities that don't involve walking, hiking, seeing temples. I want activities that will knock me right off my shoes (this is a real travel brief on our site).
Next plan of attack
What the future looks like for us: traffic dwindles down as launch activity drops.
So does that mean I give up on plantrips.net? Hell no, it’s still too early. There are a few low-hanging fruits we can take:
Sit down and talk to 5 users. Now that we have a concrete product, getting feedback on its value prop should be much easier. How does the itinerary generated help you? Which part of the itinerary is most interesting to you? How else will you use this product? Those are questions I’m eager to learn.
Tweak the value prop on the home page to be more direct about time-saving.
Find out a way to reach out to people who click through to affiliate sites and ask why they didn’t book.
Promote to travel community on reddit/facebook/tripadvisor
Promote to travel community with busy parents
Test out a SaaS angle as a trip planning tool through a few simple buttons like “Upgrade to edit trip”, and “Upgrade to export to pdf”. I won’t build a full fledge feature. A coming soon message and an input to collect emails should be enough here.
If you are interested in figuring out what happens next, subscribe to the next episode of “Co struggling in public” here:
If you just want to help, please share this post with whoever you think will benefit.