"Synergy" is the word conglomerates use when they cannot explain why they own unrelated things. I try not to use it, because a portfolio has to earn its structure with specifics.
Here is what genuinely shares across our companies, and what does not.
What shares well
Engineering and design. This is the big one. The team that built a real-time card game can build a marketplace. The design system developed for one property makes the next one faster and better than it would have been alone. A new venture inside the portfolio starts with capability that would take an independent startup a year and a hiring budget to assemble.
This is most of the case for our structure. Xsatori Labs does client work and also builds for the portfolio, which means the capability is funded by revenue rather than carried as overhead.
Operating systems. Not software — the accumulated knowledge of how to run something. How we onboard. How we handle a customer complaint. What our contracts look like. What we have learned about collecting on time. Every company we start inherits this and skips several years of expensive lessons.
Infrastructure and vendors. Hosting, tooling, insurance, legal, accounting. Individually small. Across five companies, real money and much less administrative surface than five separate stacks.
Judgment. The least measurable and possibly the most valuable. Having watched several businesses hit the same wall, you recognize it earlier. That pattern recognition only accumulates if you keep things long enough to see how they turn out.
What does not share
Customers. This is where holding companies fool themselves. The instinct is that a staffing client will want the software product, or that game players will care about the marketplace. Almost never true. These are different people with different problems who happen to have the same owner.
We do not force it. Cross-selling unrelated products annoys customers and produces very little revenue.
Brand. Each company has to stand on its own. A staffing client does not care about the holding company. A player does not care. The T47 name matters to partners, to potential portfolio companies, and to people evaluating whether we are serious — and almost nowhere else. Trying to make the parent brand do work at the customer level is effort spent on the wrong audience.
Culture, entirely. A staffing operation and a software studio have genuinely different rhythms, and pretending otherwise makes one of them worse. Shared values, yes. Shared operating cadence, no.
Management attention. This is the real cost of a portfolio and there is no way to share it. Every company needs an owner who thinks about it constantly, and that cannot be me for five companies. It is why "who runs it on day two hundred" is the filter that kills the most ideas.
The overhead nobody mentions
Shared services are not free. There is a coordination cost: whose priority wins when two companies need the same engineer, how shared costs get allocated, who decides. Get this wrong and the shared team becomes a bottleneck that makes every company slower than it would be alone.
Our approach is that shared capability serves committed work first and opportunistic work with what remains. It is imperfect and it requires an actual decision rather than a policy, but it prevents the failure mode where everything is equally urgent and nothing ships.
The honest test
A portfolio is justified if each company is better off inside it than outside it.
Not "the group is bigger." Not "there are synergies." Specifically: would this company be growing faster, operating better, or making more money as an independent business with its own investors and its own engineering team?
If the answer is no for a given company, it should probably be sold or spun out, and the fact that it is not is a sign the structure is serving the owner rather than the business.
I ask this every year, company by company. So far the answer has been that each one is genuinely better inside — mostly because of engineering capability that none of them could afford alone. The day that stops being true for one of them, the right move is to let it go.