Why we own the insurance stack instead of partnering

Nobody sets out to build six companies. We kept hitting a layer that was breaking our numbers, and kept deciding to own it. That sounds like a story about ambition, and it is really a story about arithmetic: in this industry the layers between the premium and the risk do not merely take a cut each, they take a cut of a number the previous layer already degraded. Owning them together is not six advantages. It is one advantage that only exists when the whole set is held at once.

Where the money in a policy actually goes

Ask a consumer what they think their premium pays for and they will describe risk: the chance that something happens and the promise that money arrives when it does. That is what an insurance product is for, and it is a smaller share of the bill than most people expect. The rest covers the cost of finding them, talking to them, evaluating them, placing the business, paying the people who placed it, and servicing the policy for as long as it stays in force.

Follow a single policy backward and you can name the parties. Someone generated the inquiry, usually a performance marketer buying traffic. Someone aggregated and resold it, often more than once. Someone provided the dialer the agent used, and someone else provided the minutes underneath the dialer. Someone provided the quoting tool. An agency contracted the agent, and a marketing organization above that agency held the carrier contract. The carrier underwrote and issued. A reinsurer took part of the risk. Every one of those parties is a real business with real costs, and each one is entitled to a margin.

The problem is not that any of them is greedy. It is that nine parties means nine sets of overhead, nine account managers, nine contracts, and nine incentives that are aligned with each other only loosely. The policyholder pays for the whole assembly, and none of the nine is accountable for whether the policy is still in force in three years.

What each handoff costs

A handoff between two companies in this chain has two costs. The obvious one is margin: the receiving company pays more than the sending company spent. The less obvious one is information, and it compounds in a way margin does not.

The lead handoff

A lead bought from an aggregator arrives without the thing you most want to know, which is how it was generated and what the person was actually shown. You get fields. You do not get the disclosure language, the page, or the timestamp, unless you insist on it and often not then. And because the aggregator has no reason to sell the record once, the same person may be fielding calls from four buyers within the hour. Contact rate falls, the person gets annoyed, and the cost per conversation rises for everyone including the buyer who called first.

When the demand is generated in house, the record is sold once because there is no upstream vendor with an incentive to resell it. The consent record travels with it because we wrote it. That is Solved Marketing, and it is first in the chain for a reason: a leak at the top is the only leak that scales with everything downstream.

The telephony and dialer handoff

A dialer that resells minutes from a layer above it cannot change a route. When a destination degrades on a Tuesday afternoon, the escalation walks up the chain one hop at a time, and every hop adds a queue and a business day. Meanwhile the agents are sitting through failed calls that nobody can explain, and the cost of that is not in anybody's rate card.

Holding the interconnects means route selection, quality reporting, and escalation end with us. It also means the media path can be built for machine listening rather than merely for people, which is what an AI voice agent needs. Solved Telephony and AgentTech Dialer are both in production with paying customers inside and outside the group, which matters for a reason we will come back to.

The quoting handoff

This is the handoff that loses the most information and gets the least attention. An agent takes a fact-find, works one carrier, gets a decline, and starts over. The second carrier conversation begins with the applicant repeating answers they gave twenty minutes ago, and the reason for the first decline is nowhere in the record. When a case is eventually placed, nobody can reconstruct why that carrier rather than another.

Collecting the fact-find once, into a versioned profile on a client record, is what makes the same answers serve a life case, a Medicare conversation, and an ancillary attachment. That is Solved Enroll, and it is in private beta with a public rollout planned for 2027, so this is a claim about a system running with cohort agencies rather than with everyone.

The contracting handoff

An agent who is not appointed cannot write the case, and appointment state lapses quietly. When contracting sits with a third party, the quoting platform either assumes appointment state or asks for it by email, and both of those produce the same outcome eventually: a submitted application that generates a commission dispute instead of a commission.

Owning contracting through Solved Solutions means appointment and hierarchy state is a record the other platforms read at submission time rather than a belief they hold. It also means the field force is not a channel we rent.

The product handoff

Here is the one that decides whether any of the rest matters. If you have lowered acquisition cost and tightened risk selection but somebody else owns the product, your efficiency becomes their margin. You can renegotiate; you cannot price. Owning the product is what turns lower cost into better pricing rather than into a wider spread for a counterparty.

Which is also why we are careful about this one. Solved Insurance, with Solved Re, is in development and pending state approval. Nothing is on sale. The argument above is the reason the product company exists; it is not a claim that it is operating.

Three levers, and why they only work together

Read those three as a loop rather than a list. Cheap acquisition is what makes disciplined declining affordable. Disciplined declining is what lets a proprietary product be priced honestly rather than defensively. Honest pricing is what makes the next lead cheaper to convert, which lowers acquisition cost again. Each turn of the loop makes the next turn cheaper.

Now break one link. Partner out acquisition and selective underwriting becomes unaffordable, because you cannot throw away leads you bought at retail. Partner out underwriting and you are pricing against somebody else's risk appetite. Partner out the product and the loop ends in a counterparty's income statement. Any single partnership does not cost you a third of the advantage. It costs you the loop.

The loop nobody talks about

There is a seventh flow that only exists when both ends are owned, and it runs backward. Placement rates, decline reasons, and persistency on business that was actually written get fed back to the acquisition company at the level of the program and the filter that produced it.

In a rented chain that signal either does not exist or arrives a quarter late and aggregated into uselessness. A lead vendor will tell you your contact rate. It cannot tell you which of its sub-sources produced business that was still in force a year later, because it never learns the outcome. So lead quality gets measured by whether the phone was answered, which is a proxy for the thing you care about only in the loosest sense.

When the same group owns both ends, the feedback is specific and it is fast: this program, this filter set, this state, this age band produced business that stayed. That is the difference between optimizing a funnel and optimizing a book, and it is the single strongest operational argument for integration that we have found.

The honest cost of vertical integration

Everything above is the case for. Here is the case against, which we would rather write ourselves than have written for us.

Capital

Six companies is six sets of fixed costs before any of them is at scale. Building a telephony network, a dialer, and a quoting platform is years of engineering that a single license agreement would have bought in a week. Insurance products move at the speed of regulators, so the last layer in the chain is also the slowest and the most capital-hungry, and it cannot be accelerated with effort. Anyone who tells you vertical integration is cheaper in the short run is selling something.

Our response to this is boring on purpose: own the layers where the margin actually leaks, which is acquisition, underwriting, and distribution, and buy the commodity inputs. Cloud, payments, and data vendors are bought, not built. Every company is expected to carry its own weight on its own customers before the next one starts.

Execution

Focus is a real advantage and integration spends it. A point-solution company gets to be excellent at one thing and have a single number that says whether it is working. We have six businesses with six operating rhythms, and the failure mode is not dramatic: it is six things that are each almost good enough. A partner who is the best in the market at one layer is a genuinely hard thing to beat with an internal team that has four other priorities.

The only honest mitigation is to be the first customer of everything and refuse to accept from ourselves what we would not accept from a vendor. Every platform in the group started as an internal fix for a problem we were living with, which helps, but it is a discipline rather than a guarantee.

The temptation to subsidize a weak layer

This is the one that actually kills integrated companies, and it is subtle. When you own a layer, its revenue is internal, so its weakness is invisible. A dialer that would lose every competitive evaluation in the market keeps its customer because its customer is you. Slowly, the group is paying above market for a capability it could have bought, and the number never appears anywhere because it nets out in consolidation.

Our structural answer is that the technology companies sell outside the group. AgentTech Dialer and Solved Telephony have paying customers with no other relationship to us, which means they are priced and judged by people who can leave. That is uncomfortable in exactly the right way. A layer that could only survive on internal revenue would be a subsidy, not an advantage, and we would rather find that out from a lost deal than from a slow decade.

How we decide whether a layer deserves to be owned

• Does margin leak here? If the layer is a commodity with thin margins and many suppliers, owning it buys nothing. Acquisition, underwriting, and distribution are not that.
• Does information die at this handoff? If the thing you most need to know is what the other party has no reason to tell you, the handoff is the problem rather than the price.
• Can it survive on external customers? If the answer is no, you are building a cost center and calling it a strategy.
• Does owning it make the next layer cheaper? The test is not whether the layer is profitable alone. It is whether it changes what is affordable downstream.
• Are we willing to be the worst customer? If we would not hold our own team to a vendor standard, we should buy from a vendor.

Where partnering is still the right answer

Most of the time, for most companies. If you are an agency, buying a dialer, buying leads, and contracting through an established marketing organization is the correct decision, and the cost of the layers is the price of not having to build them. If you are an insurtech with one genuinely differentiated capability, focus is worth more than breadth and a smaller capital requirement is worth more than control. If you are a carrier with an in-force book, decades of rating history, and capital, you already hold the layer that matters most, and buying distribution is rational.

Integration is the right answer under a narrow condition: when the advantages you are chasing are the kind that only compound if you control them together, and when you are prepared to pay for that in capital, time, and focus. We think final expense at the top of the senior market is exactly that situation, which is the subject of the next article.

If you want the version of this with the operating detail in it, the strategy page has the four pillars and the documentation index describes what actually crosses each boundary. Or talk with the team and ask the hard version of the question.