Buying an AI model: Understanding the Factors

Sachin Raghavendran
Volant AI
Published in
4 min readMay 13, 2021

As we continue building out the platform at Volant AI as well as growing our waitlist, it’s important to take a step back and analyze the pull (and push!) factors for a buyer to get a model from our marketplace. Though we’re aiming to help not only SMBs, but also hackers and Kaggle competitors, the psychology framework we suggest can be applied to each potential use-case.

As a seller on our marketplace, this approach can be invaluable in helping you succeed. Let’s dive in.

There seem to be some underlying factors that influence a buyer’s need to obtain a model from our marketplace. This isn’t a comprehensive list and it is likely that as Volant AI continues to grow, we may find that some are wrong. This is a good start though, since you can image that many of these factors would be in play in a traditional marketplace (like Amazon). Here’s our list:

  1. Cost/Convenience
  2. Seller’s Expertise
  3. Objective Model Performance
  4. Task Importance
  5. Social Proof

Cost/Convenience

Cheaper models (which can be conveniently used on our platform) are pretty low-risk.

Seller’s Expertise

Seller reputation, whether from external work or in our marketplace, makes the model more inviting.

Objective Model Performance

The “Bread-and-Butter” of the model. The ground source of truth of how good it is when used on a dataset. The higher it performs, the better.

Task Importance

Models for business-critical tasks means higher stakes and less margin for error.

Social Proof

You can be influenced by others having bought the model and having good things to say about it.

Our Proposed Framework

Different people, depending on their role and usage, will probably have different factors influencing them. It’s a tall order to try to tailor your model (or marketplace!) to satisfy your consumers. A good framework to visualize this would be a venn diagram constructed like so:

Generic Factor Venn

We know that different factors are dependent on each other and this diagram captures that. The overlap between all five above will should (naturally) be no more than say, any two. However, it is important to note that any two factors may not be equally co-dependent. For example, if a model is built for a generally important task, it could be costlier but may or may not perform well. In this case there would be a higher overlap between Task Importance and Cost/Convenience and probably less so between those and Objective Model Performance.

Maybe it even looks like something below, where we see the Factor Venn with higher emphasis on Cost/Convenience and Social Proof but not so much on Seller Expertise. We could even see the two emphasized factors are likely dependent in this scenario while the other three may not be so interlinked.

Another Factor Venn. Described above.

Using the diagram above, we could now focus on catering towards this audience of buyers by aiming to make the model cheaper and trying to get as many people to use it as possible.

We’ll definitely explore this more in future articles.

How you can use this framework

Please note that this is NOT a quantitative framework. The venn diagrams could be deeply perused and argued about, but the important things to unpack from any diagram are almost superficial. To use this framework (which we will be constantly developing):

  1. Take your factors and reason about which ones are most tightly locked. Put those factors adjacent to each other.
  2. Now, figure out which ones are most strongly affecting your customer. Make the corresponding venn circles larger relative to other ones.

A question that I have (conveniently!) avoided is what factors to consider. The ones above seem like a good starting point for us at Volant AI, but determining those factors boils down to doing something simple: understand your customer:)

(It’s pretty hard to do that!)

As a potential seller on our marketplace, it will help you best if you figure out what entices your potential buyers (especially for the type of model, domain, etc.) to buy a model. Speak with interested parties and parse what insights you can. Constructing a venn like above can help give you another way to understand how to evangelize your models and make the money you deserve.

As always, feel free to share your thoughts by dropping a comment!

See our previous articles introducing our platform and discussing sellers.

If you are interested in a selling a model you have built in the past, then you should join our waitlist as we are starting to onboard sellers into our marketplace.

Join Us.

--

--