How we underwrite a software acquisition for a 25-year hold.

Most of the questions we get from prospective investors are about the back of the deck — IRR ranges, distribution policy, fee structure. The more interesting question, and the one that actually determines outcomes, is the front of the deck: what does an acquisition have to look like before we will write a check at all?

Cobalt Glacier underwrites every acquisition to a twenty-five-year hold. Not a five-year hold with optional extensions. Not a seven-year hold targeting a strategic exit. A quarter-century. That single design choice — which we wrote about more abstractly in Permanent capital is the only sane structure for software — rewrites every line of the underwriting model. It also forces us to say no to a much larger share of opportunities than a traditional sponsor would, which is the part of the discipline that actually matters.

What changes when the hold period is twenty-five years

Three things change immediately. First, terminal value stops being a polite fiction at the bottom of the model and becomes the entire point of the model. Second, "synergies" stop being a credible underwriting line, because anything we'd squeeze out of the business in year two will be more than reversed by the harm done to the operating team over the following two decades. Third, the question of who runs the business in year fifteen becomes a diligence item on day one, not an afterthought at the end of the first hold period.

A twenty-five-year hold is not a longer five-year hold. It is a completely different exercise in what you have to be right about.

Practically, that means our underwriting model is shorter, not longer, than a standard LBO model. We have fewer assumptions, because we refuse to assume anything we wouldn't bet on for a full quarter-century. We have a smaller adjustment stack, because most one-time adjustments don't survive a decade of compounding. And we have a single number on the cover page that most sponsors don't put on theirs: the year in which we expect the brand's underlying category to begin its structural decline. If that year is inside our hold period, we don't do the deal.

The three filters every deal has to clear

We get sent more software businesses than we could ever evaluate seriously. The first pass is brutally simple — three filters, applied in order, before any modeling work begins.

1. The category has to exist in 2050

We are not in the business of betting on which categories will emerge over the next quarter-century. We are in the business of betting on which categories will still be around. Those are completely different bets. The categories we write checks for are ones where the underlying customer workflow is older than the software — fund accounting, advisor onboarding, renewal management, sales enablement. The software is the latest interface for a need that has existed in some form for fifty years and will almost certainly still exist in another fifty.

Categories that fail this filter are the ones where the software invented the workflow. Those can be enormous opportunities for a venture investor or a five-year sponsor. They are not opportunities for us, because we cannot underwrite a twenty-five-year hold on a category that didn't exist twenty-five years ago.

2. The customer base has to compound with its own end-market

The second filter is structural. We want customer bases whose spend on the brand grows mechanically with the success of their own business — more advisors hired, more renewals processed, more funds launched, more sales reps onboarded. That gives us NRR that compounds without heroic effort from the operating team. Without that, every year of the holding period is a fresh sales problem, and the long hold becomes a long grind rather than a long compound.

3. The founding team wants to keep operating

The third filter is the one most sponsors treat as a soft diligence item and we treat as a hard gate. If the founder wants out, we are not the right buyer. We have written about this from the operator side in Operator continuity is the real asset. From the underwriting side it is the cleanest possible rejection: a brand whose founder is checked out is a brand whose twenty-fifth year we cannot honestly model.

What the model actually looks like

The model itself is unglamorous. We take a defensible view of revenue growth for the next four years, derived bottoms-up from the customer base and not from a top-down market share story. We hold gross margin flat or improving — we will not underwrite to margin compression, because if we expect margin compression we should not be buying the business. We model NRR honestly, which in most cases means assuming the historical number degrades by two to four points before stabilizing.

We do not model price increases above the rate of underlying category inflation. We do not model cost-out below a defensible platform-services run rate. We do not model exit multiple expansion, because we are not exiting. The terminal value in our model is the present value of the next twenty years of free cash flow after the explicit forecast period, run at a discount rate that reflects our actual cost of capital and not a sponsor's return target. That discipline is what produces a check size we can actually defend if the business does nothing exciting for a decade.

The discipline is in what we walk away from

The output of any underwriting framework is the set of deals it causes you to decline. Ours causes us to decline most of what we see. We decline good five-year businesses where the category is clearly transient. We decline founder-led businesses where the founder is, sensibly, ready to do something else with their life. We decline businesses with structurally flat NRR even when the ARR growth chart looks attractive. And we decline anything where the only credible thesis is multiple expansion or consolidation-driven cost-out, because those are not theses that survive a quarter-century.

• Categories that will dissolve. Even great businesses inside dissolving categories are a bad fit for a twenty-five-year structure.
• Founders ready to leave. A great brand without its operator is half a brand by year three.
• Flat-NRR books of business. If the customer base does not compound, the holding period is not a moat.
• Theses that require an exit to work. Multiple expansion is a sponsor's product, not ours.

Why this is the right framework for an investor

Investors who back permanent-capital structures are not underwriting our IRR over the next five years. They are underwriting our judgment over the next twenty-five. The most useful thing we can show them is not a long pipeline of approvals; it is a longer record of disciplined declines, and a clear written framework for why each one was the right call. That record is what compounds into investor trust the same way the portfolio compounds into cash flow.

Every dollar Cobalt Glacier deploys is meant to look obvious in retrospect twenty-five years from now. That is a much higher bar than looking good at the next quarterly LP meeting, and it is the bar we hold ourselves to. Investors who want a deeper conversation about how that bar shows up in our pipeline can start one here. Founders weighing a permanent home for their business can do the same here.