I was reminded by a recent Twitter thread about “what constitutes an MVP” that
- I’m confident launching just a landing page and seeing if there’s demand long before I build a product and
- I did just that for CrowdTamers AI,
- I promised a follow up to our initial CrowdTamers AI testing post from, uh, July and
- I’ve yet to deliver that.
So a few things to rectify there.
Let’s start with the TL;DR and then we can look into the why and what’s next
No one wants CrowdTamers AI.
But why? Tell me more!
Let’s start by refreshing on the goal for the test:
Based on that goal, let’s see what we were able to achieve in round 1:
Results weren’t good, but we learned a lot, which is valuable. Mostly, it seemed that our key persona was the marketer at a company, rather than an overworked founder, so we decided to try out a number of new ways to message that marketer.
And then we tried round 2. How’d that go?
|Round 1 Results
|Round 2 Results
|Click through rate %age
Well, worse. Selecting for the audience that did best (maybe?) and trying to double down on more messaging tests to see what may work and what may not resulted in a notably worse outcome.
Ad creatives like these and others just, well, didn’t do the job.
I’ve recently started to use a pretty cool tool called Hippoc.ai to analyze some of my marketing ad images. It simulates eye-tracking and human interest via AI and gives you almost immediate feedback on the memorability and impression that the graphics of your display ads will convey.
Looking at the report it generates, it suggests that these ads aren’t doing a great job of capturing interest. which seems more or less borne out by the ad results themselves.
Worth paying particular attention to is the heatmap of what people will see immediately:
The faces draw attention, the headlines draw a little bit of attention, and the CTAs are basically invisible. So when you combine the fact that the value prop is not compelling enough with the image & color choices I ran with, and these ads just aren’t any good at their job.
I mention often enough that I’m wrong more than 66% of the time, and here’s a decent example of that: I was sure that these ads would do better than the last set and thought I’d validate my idea for CrowdTamers AI.
As it turns out, I didn’t.
It’s crucial here that you remain honest with yourself and with your team in these situations so that you’re flexible enough to learn from the data instead of keeping on charging ahead because of the momentum you’ve already got to execute on this idea.
With 2 iterations of a test where nothing showed a good result, it proved enough to my satisfaction that there was something fundamentally wrong with our approach. It seems likely that
- We didn’t have a clear enough value prop.
- Our audience didn’t really want what we were offering anyway, and
- Once they got to our landing page, they didn’t believe we could deliver what they expected.
That’s pretty “kiss of death” territory in my book.
But to have learned all this about a product that existed only as a landing page, and for a total ad spend of $400? I can’t tell you how happy I am to have learned this for so cheap.
When I talk about Launch Today, this is why I am so determined that companies need to launch more, launch faster, and learn before they build.
So what’s next?
Despite the shortcomings of GPT-3 for generating long-form content of any quality, I still believe in the potential of AI to help business owners create assets and copy for their own marketing. I’m continuing to think about what good AI tools might look like and I’ll be testing several smaller ones over the next 6 months.
Right now I’m developing some landing pages for 3 products:
- Instantly get recommendations for Facebook Ads / LinkedIn Ads targeting
- Generate new ad headlines, drop into our free templates, and add in photos with instant background removal.
- Recommendations on how to fix ad campaigns based on articles I’ve written in the past.
Anything you think an AI could help you with? Let me know @trevorlongino on Twitter and I’ll see about building and testing some landing pages for it. 🙂