In the early days of the startup phase, you will undoubtedly be concerned with validating one of your key assumptions: whether there is a strong market demand for your product. This is commonly known as finding Product Market Fit (PMF), a term popularized by Andreessen Horowitz in this article.
During this phase, you are searching for a repeatable and scalable business model while faced with plenty of uncertainties. One of the key tools that you should rely on during this phase is experimentation and validated learning. The latter is a term proposed by Eric Ries in his seminal book — The Lean Startup.
One of the most effective ways I have found to navigate this early stage is to put your products in the hands of early adopters. You might also be familiar with the early adopters term from the graph below, which was popularized in Geoffery Moore’s “ Crossing the Chasm “ book
To be more specific, you want to find Earlyvangelists a term popularized by Steve Blank.
“Earlyvangelists are a special breed of customers willing to take a risk on your startup’s product or service because they can actually envision its potential to solve a critical and immediate problem-and they have the budget to purchase it”. Steve Blank — The Four Steps to the Epiphany
According to Blank, these Earlyvangelists need to have these set of characteristics
Source — Steve Blank
I’ll skip a lot of details, namely the interaction between customer discovery and validation, which Blank discussed at length in his book (hint: it’s a must read). I will also relax the 5th requirement, which I have found useful in the very first iterations of finding and working with these Earlyvangelists. I will be using the terms early adopter and Earlyvangelists interchangeably in this post.
In my experience finding an Earlyvangelists without a budget early on is better than not finding any. You will absolutely want to enforce this constraint as you progress along in your journey.
With that bit of definition, the next question is how do you find these Earlyvangelists? I will offer three ways I have found useful based on my experiences. There are undoubtedly a lot more.
The Open Source Software (OSS) model
Back in 2016, a few engineers from Dremio led by co-founder and CTO Jacques Nadeau along with other remembers of the OSS community started to work on Apache Arrow. One of the intended uses of Arrow was to replace legacy data APIs like ODBC and JDBC. Arrow is used extensively in Dremio as shown in this article.
“ In Dremio we make extensive use of Arrow. As Dremio reads data from different file formats (Parquet, JSON, CSV, Excel, etc) and different sources (RDBMS, Elasticsearch, MongoDB, HDFS, S3, etc), data is read into native Arrow buffers directly for all processing. Our vectorized Parquet reader makes reading into Arrow especially fast, and so we use Parquet to persist our Data Reflections for accelerating queries, then read them into memory as Arrow for processing.” Jacques Nadeau — Dremio Co-founder & CTO
It turns out that Arrow became hugely popular. To the tune that it is downloaded ~10M+ times a month. It’s also extremely useful and powerful. You should try it! And you should try Dremio (hint: I might be slightly biased!)
The rapid adoption of this project by developers, its association with Dremio and the company offering a free version of its product complemented with an online community has yielded lots of Earlyvangelists to Dremio.
This model will likely only work if your product is geared towards a software developer persona. It is fairly common with products that are associated with a popular OSS projects. Notable examples include Databricks/Apache Spark and Confluent/Apache Kafka. Notice how the products that these companies make are predominantly used by engineers.
This model can have a dramatic side effect if it works: your products will be deployed and used by 1000s of users, even enterprises. If that happens, then congrats are in order! Your next challenge will be to try and monetize this widespread (free) adoption.
The spear-fishing model
This is a targeted and deliberate model. Instead of casting a very wide net like the OSS model, you are focused on identifying your target early adopters and actively seeking them.
In the early days of Qumulo, the Product team spent the majority of their time conducting customer discovery interviews. These were targeted interviews conducted with users of storage systems, typically storage administrators.
“Before building our file system, we conducted thousands of interviews with enterprise data storage users. You told us your struggles, your concerns, the annoying things that ruin your day. And you shared your wishlist, desired features, and “what-if” ideas. What we learned inspired us to rethink what’s possible.” Source
One of the questions the product team would ask, if they found that there was a good fit between the interviewee’s pain-points and Qumulo was to give them access to Qumulo’s product. Initially access to this product was as a virtual machine, but it was sufficient to garner significant product feedback and validate numerous hypotheses. Later on, as the product evolved these early adopters could get Qumulo’s software running on hardware in their data-centers.
This model is will require dedicated resources. You have to put in the time to find, interview and to be actively engaged with adopters who start using your products. When it does work, it yields tremendous validation data and an opportunity to convert those early adopters into your first customers.
The collaboration model
This is a model that I have been recently exposed to at Kheiron. First, a bit of context on what Kheiron does. Kheiron’s flagship product — Mia — is based on a deep learning model for breast cancer detection. Like all deep learning models its performance is highly dependent on the amount of training data — the more data, the better the model’s performance.
Hospitals and medical research institutions have a lot of this data. Many of these are also eager to adopt AI based products with the ultimate goal of using them in their clinical workflow. However, no medical institution — rightly so — will start trialing a product like Mia’s and immediately put it in their clinical workflow — the risks are too high. True adoption of products like Mia require clinical trials, regulatory approvals and trust between the medical community and the company.
This is why a collaborative approach between the company and the medical community works best. Kheiron gets access to data, feedback and a very good understanding of clinical workflows, whilst the community gets to influence the development of healthcare AI products and be thought leaders in the adoption of AI within the healthcare industry, initially outside the patient or clinical workflow. These collaborative engagements can also result in joint research publications too.
This model is well suited for products that are used in heavily regulated industries, or ones where the adoption of these products requires a collaborative effort between the company and its customers. It has worked very well in the intersection of health-care and AI, which is an incredibly exciting space to be. That’s a topic for another day.
You’ll notice that I didn’t include a freemium model here. Putting a free version of your product on the internet requires market awareness. You’ll have to solve this problem first before you can get any meaningful early adopter engagement. Freemium models are a fantastic lead generation tool, once you have already established a PMF and the market is starting to know you and your products.
Another important distinction to make is the difference between a true early adopter or Earlyvangelists and a casual user of your product. A true early adopter has all the characteristics that Blank outlined above. I’d add that they will want to actively use your product and put it in a production workflow, versus simply kicking the tires with it. That is an important distinction to make, especially if you adopt an OSS like model. You’ll get lots of users who tinker with your products and few true early adopters. You need to separate the signal from the noise.
The table below highlights some of the main differences between these three models. They are not mutually inclusive. You can use one or more during your search for Earlyvangelists.
There are undoubtedly many more models, but these are the ones I have tried and am familiar with. I’d love to hear about others you have used.
Finally, I’ve referenced a few books throughout these article, which I believe are a must-have, especially in the startup phase. These are the following:
Originally published at https://karimfanous.substack.com.