Dmitrii Khasanov: Metrics and algorithms of evaluating AI Startups

It’s time to admit a simple fact: We live in the era of AI technologies. They attract interest from businesses and investors, but, as elsewhere, only a few succeed. Every second founder of an AI startup promises almost a technological revolution, but how do you distinguish a really promising project from a dummy or a scam like an online casino or a pyramid scheme?

We talked about this and much more with Dmitrii Khasanov, the founder of Arrow Stars investment fund and a digital marketing guru.

A century-old artificial intelligence

It’s worth starting with the fact that there is no unified artificial intelligence. Of course, ever since the time of Alan Turing (the British mathematician, creator of the first computer in history), the dream of creating AGI (Artificial General Intelligence), a universal artificial intelligence that can think and act like a human, has been living. However, unfortunately or fortunately, the modern market is not able to offer us this technology. What is meant by the term “an artificial intelligence” today is a set of many different mathematical approaches that solve various business problems.

Each AI algorithm is useful for solving a specific set of tasks. These algorithms are not easy to develop, and they require extensive mathematical knowledge. There are very few real breakthroughs. The basic mathematical concepts underlying most well-known AI algorithms appeared many decades ago. However, at that time it was impossible to implement such concepts – at that time we did not have enough computing power. It is only in the last few years that computing power has reached the level where it is possible to implement AI in the form that was originally conceived.

Before start

Basically, the economics of an AI startup are not much different from those of other projects, with the exception of three points: paying for development, paying for data, and paying for specialists. However, before evaluating this, an investor should understand what the final product will be.

Therefore, the first question to ask a startup team is what problem they are trying to solve with AI. Suppose a company (a potential partner) plans to create a service based on a large language model.

The next question is: will the company use the existing model or develop it from scratch? The cost of the project strongly depends on this. Both options have certain advantages. If the company buys (or takes a subscription) an existing model, it will be cheaper and easier to install, but it is less flexible and may not be suitable for the initial task. In this case, it will cost money on a subscription or on a “bundled package” that technicians will install on the internal infrastructure. If a startup plans to produce its own model, the cost and duration of the project will increase significantly, but it can be customized as needed by the creators. In this variant, the startup team will have to pay for the work of developers or buy and install its own development infrastructure and develop it by themselves.

The next point is data payment. To train a model, the company needs data, and those that are publicly available will not be enough: the information will quickly become outdated, and the model will work inaccurately. Therefore, the owners will have to buy your own data from third-party companies, and they will have to do this regularly.

The cost of people assumes that in addition to the main staff, the project will employ an army of data markup specialists, testers, teachers for the model, and so on.

After receiving answers on these points, you will be able to roughly calculate the startup’s costs.

Main metrics to monitor AI Startup

After the investor has calculated his costs and decided to participate in the project anyway, he will need metrics in order to monitor the effectiveness of the startup’s development.

1. Customer Acquisition Cost (CAC): AS practice shows a highly effective marketing strategy. To calculate it, divide the total cost of attracting new customers by the number of customers who came to you over a certain period.
2. Monthly Recurring Revenue (MRR): This metric is important for startups with subscription-based models because it provides a clear picture of projected revenue. MPR is the total amount of regular income, systematized by the daily amount.
3. Churn rate: The churn rate shows how many customers have unsubscribed during a certain period. This metric shows customer satisfaction and product relevance.
4. Gross Margin: In simple terms, this is the difference between revenue and the cost of goods sold divided by revenue. This metric shows how much money you will earn.
5. Burn Rate: This indicator explains how quickly a company spends its capital before becoming financially independent.
6. Customer Lifetime Value (CLV): This is the total income that a company can reasonably expect from a single client throughout the entire period of cooperation. This indicator helps companies develop strategies to attract new customers and retain existing ones.

Dmitrii Khasanov says that his methodology is far from the only way to evaluate AI startups, but it has been tested on dozens of companies from his personal experience.

Have you read?
The World’s Best Medical Schools.
The World’s Best Universities.
The World’s Best International High Schools.
The World’s Best Business Schools.
The World’s Best Fashion Schools.
The World’s Best Hospitality And Hotel Management Schools.