This may feel a bit random of a post, as Business and Economics are not my areas of expertise and I usually do my best not talk about stuff I don’t know, but I have seen the complete disregard for this concept lately and I thought it would be a good starting point to define here, before I talk about it, what a “breadwinner product” is, from my point of view.
The term breadwinner is used generally to refer to the primary income-earner in a household. While I have not seen this very often extended to products and services in companies, I think it should be fairly obvious how the extension would work.
In a world of startups there are still plenty of companies that have a real “breadwinner product”, even when acting as startups. This is the case for instance of the company I used to contract out for, in Los Angeles: they have been in activity for a number of years with a different, barely related product, and I was contracting out for their new project.
I think it’s important to think of this term, because without having this concept in mind, it’s hard to understand a lot of business decisions of many companies, why startups such as Revolut are “sweeping up the market” and so on.
This is something that came up on Twitter a time or two: a significant amount of geeks appear to wilfully ignore the needs of a business, and the marketing concepts as words of the devil, and will refuse to try considering whether decisions made business sense, and instead they will either try to judge decisions based purely on technical merits, or even just on their own direct interests. Now it is true that technical merits can make good business sense, but sometimes there are very good long-term vision reasons that people don’t appreciate on the pure technical point of view.
In particular, sometimes it’s hard to understand why a service by a company that may appear as a startup is based on “old” technology, but it may just be the case that it is actually a “traditional” company trying to pivot into a different market or a different brand or level of service. And when that happens, there’s at least some gravity pull to keep the stack in line with the previous technology. Particularly if the new service can piggyback on the old one for a while, both in term of revenue, technology and staff.
So in the case of the company I referred to above, when I started contracting out they were already providing a separate service that was built on a real legacy technology, ran on a stack based on bare metal servers with RedHat 5. Since the new service had two components, one of them ended up being based on the same stack and the other one (which I was setting up) ended up based on Gentoo Linux with containers instead. The same way as the Tinderbox used to be run. If you wonder why one would run two stacks this separate, the answer is that messing with the breadwinner product, most of the time, is a risky endeavour and unless you have a very good reason to do so, you don’t.
So even though I was effectively building a new architecture from scratch, and was setting up new servers, with more proper monitoring (based on Munin and Icinga), and Puppet for configuration management, I was not allowed to touch the old service. And rightly so, as it was definitely brittle and it would have lead to actually losing money, as that service was running in production, while the new one was not ready yet, and the few users of it would be able to be told about maintenance windows in advance.
There is often a tipping point though, when the cost of running a legacy service is higher than the revenue the service is bringing in. For that company that happened right as I was leaving it to start working at my current place of work. The owner though was more business savvy than many other people I met before and since, and was actually already planning how to cut some expenses. Indeed the last thing I helped that company with was setting up a single¹ baremetal server with multiple containers to virtualise their former fully bare metal hardware, and bring it physically to a new location (Fremont, CA) to cut on the hosting costs.
The more the breadwinner service is making money, and the less the company is experimenting with alternative approaches to cut the costs in the future or to build up new services or open new market opportunities, the more working for those companies become hard. Of all the possible things I can complain about my boss at the time, ability to deal with business details was not one of those. Actually, I think that despite leaving me in quite the bad predicament afterwards, he did end up teaching me quite a bit of the nitty-gritty details of doing business, particularly US-style — and I may not entirely like it either.
But all in all, I think this is something lots more people in tech should learn about. Because I still maintain that Free Software can only be marketed by businesses and to be able to have your project cater to business users without selling its soul, you need to be able to tell what they need and how they need it provided.
¹ Okay, actually a bit more than one: a single machine ran the production environment for the legacy servers, and acted as warm-backup for the new service; another machine ran the production environment for the new service, and acted as warm-backup for the new service. A pair of the older baremetal servers acted as database backends for both systems.