We're a tiny boutique consultancy but we build end to end machine learning products for very large organizations. The CEO is our sales guy. We get referrals and he leverages his network to find projects. He also teaches a course in university and writes in a newspaper, and we get inbound requests.
Recently, however, we built our own machine learning platform (MLOps, for the buzzword) to be able to execute these projects and build these products faster and more systematically. We did that because they took a long time before and required certain profiles that are rare, and the way we were going about it had too many bottlenecks and risks.
We also refined the process of doing these projects, especially with scoping, understanding the jobs to be done and non-consumption, and involving the actual people we were building for, not just talking with the execs/management. We had to insist. This lead to "repeat business" and reduced the need to always be prospecting. In other words: make your clients happy and you'll work together again, and the next projects will be more important and you'll amortize the time it took to build the trust because you're practically colleagues. I.e: compounding.
You need some or all of:
- Finding new clients
- Charging more or finding clients who can pay more.
- Doing more projects with the same clients
- Improving your processes so projects take less time
- Re-selling the same product your built for a client to other clients, and amortize a codebase
- Building a product for a sector/vertical/role
- Improve your tooling
- Productize yourself so you can sell items or a bundle, as opposed or in addition to selling your time.