How to efficiently find a product that sells
To find product-market fit efficiently, test two very different MVPs at the same time that are of high enough quality
New to Ariyh? Join 3,685 evidence-based marketers for 3min tips 💡 based on research 🎓 to grow your business 📈
You may be familiar with the Lean Startup method by Steve Blank (Stanford University) to create successful products by trial and error:
Create a minimal viable product (MVP)
Try to sell it to your target customers
If people buy it, proceed and build the full product
If it doesn’t sell, update it or create a different MVP and try again. Repeat until it sells
This method is useful for companies of any size, not only startups.
Today’s research from University College London and Boston College uses a (Bayesian) statistical model to determine how to accurately find an MVP that sells, using the least amount of resources to get there.
P.S.: What you are reading right now is (sort of) an MVP of Ariyh.
Initially, I was tempted to build a large database of tips and release them all at once as a searchable platform.
Instead, this newsletter-first approach has allowed me to test at a relatively low cost whether there is a market for science-based marketing tips (it sure seems like it! So this is just the start 🚀).
And thanks to your feedback I can continuously improve Ariyh by learning what tips you find most useful. At the bottom of each tip, you can easily rate it (in 2 clicks). Thank you! 🙏
Test intermediate quality MVPs against each other. Disproving an MVP is just as valuable as proving one.
Impacted metrics: Customer acquisition
For: Both B2C and B2B
Tip type: New research (March 2021)
Previous tip: Which ads work best in emerging vs developed markets (All tips here)
To find product-market fit using different minimal viable products (MVPs) as efficiently as possible (accurately, economically, and fast), you should:
Test two (or more) significantly different MVPs at the same time. Avoid common features. This maximizes learning (e.g. what people liked of each) and allows you to compare sales (otherwise, you’re making assumptions of what are ‘good’ vs ‘bad’ sales).
Use MVPs of sufficiently high quality. If quality is too low, nobody will buy it, this gives us little or - worse - wrong information. If quality is too high, you’ve invested too many resources.
Use higher quality MVPs for product versions you think are more likely to work, and lower quality versions for products you want to test but are less likely to succeed (failure to sell is just as informative as succeeding).
An MVP is a version of the product that communicates the unique selling point of the product but without the “bells and whistles” (e.g. a website that has 10% of the functionality that you envision).
For new products (e.g. Clubhouse, the first iPhone), it’s often a challenge to understand customer demand without launching the product. Testing different MVPs until one works offers a way around that.
The method in the Recommendation section maximizes learning (of what works vs doesn’t) while minimizing the risk of choosing the wrong product.
Higher levels of learning from MVP testing are related to higher profits of the final product. The more you learn, the more you earn.
New here? Subscribe for the latest marketing research from top business schools in 3-min practical tips, twice per week.
Why it works
Mathematically, this method maximizes the probability of learning with the least number of different MVPs without sacrificing the accuracy of which product is best for the market.
This research is a theoretical analysis of the lean startup method. Since this was not its purpose, it lacks lab experiments or real-world observations of MVPs to observe what happens when they follow this theoretically optimal method.
In a perfect world, numbers don’t lie. But the world isn’t perfect (e.g. the researchers could have failed to take something into account in their model), so be cautious when applying this.
Companies using this
Startups often make the mistake of launching an MVP as a ‘fast way to launch’, and then build on it, rather than a way to test core product assumptions. If they do test different MVPs, they usually differ in design but are too similar in terms of functionality. This is suboptimal and limits learning.
Steps to implement
Finding product-market fit through MVPs is
You’re uncertain if a need exists (e.g. an app that plays soft music for your plants). Not if you’re quite certain it is (e.g. there’s a need for more fuel-efficient cars)
People in your target group (e.g. B2B marketers) have similar needs, so a few successful sales are a signal the whole group would buy. Otherwise, you risk misinterpreting sales in a subset of customers (e.g. B2B marketers in manufacturing) as demand for the whole group
Product changes are relatively easy and can be easily upgraded (e.g. software)
Limited when a product needs to have a high level of quality (e.g. luxury products) or completeness (e.g. a cargo ship) in order to sell
Build MVPs with different functionality assumptions and test them against each other. For example, a B2B accounting software could test two different core functionality assumptions with two MVPs:
Product A’s unique selling point is seamless integration with bank account data
Product B’s main value is extensive spending analytics instead
If you have many functionalities you want to test with different MVPs, you could start with testing options that you think will be discarded using low-quality MVPs (but not too low quality, otherwise nobody will buy and you won’t learn). Then, for the most promising one(s), you can create a higher quality MVP so you will have a headstart in product development if it sticks.
Testing just one time (two different MVPs) can be enough if it leads to sufficient learning (e.g. you see a large difference in sales).
Theoretical model (of optimal decision-making and captured market demand, using Bayesian learning)
Yoo, O. S., Huang, T., & Arifoğlu, K. (March 2021). A Theoretical Analysis of the Lean Start-up Method. Marketing Science.
UCL School of Management, University College London and Carroll School of Management, Boston College. United Kingdom and United States
Remember: Because of the groundbreaking nature of this paper, it could be disproven in the future (although this is rare). It also may not be generalizable to your situation. If it’s a risky change, always test it on a small scale before rolling it out widely.
Rate today’s tip to help me make Ariyh's next tips 🎓 even more useful 📈
Want to sponsor Ariyh or ask a question? -> Reach out at firstname.lastname@example.org
New to Ariyh? -> Subscribe below or read other 3min marketing tips here