Government revenue is sticky until the buyer changes how it buys

Software sold to public bodies is often described as the best revenue an acquirer can buy. Municipalities, courts and school boards rarely switch systems, they pay on time and they renew maintenance for decades. About 54 per cent of Sorrelgate's revenue comes from such customers, and I have spent more than twenty years on one side or another of their purchasing. My position is that the usual measure of this revenue, customer retention, is the wrong one for valuing it. The real risk sits in decisions taken above the customer about how it buys, and acquirers should price that risk explicitly.

Retention tables hide the procurement calendar

A seller of a public-sector software business will usually show a buyer a retention table. Gross revenue retention of 96 or 97 per cent a year is common, and it is usually accurate. What the table measures is whether the people who use the software choose to keep it. In a public body those people rarely make that choice alone. The decision sits with a procurement function, a shared-services agency, a provincial ministry or a state technology office, and those bodies work on their own calendar and for their own reasons.

When I priced bids for an Ottawa integrator in the early 2000s, I learnt that a contract vehicle, such as a list of pre-qualified suppliers or a multi-year standing arrangement, often matters more than any single customer relationship. Being on the vehicle lets dozens of departments buy from you with little effort; falling off it, or watching it be replaced, can end those relationships however happy the users are. Retention data cannot show this, because the event has not yet happened, and when it does happen it happens to many customers at once.

The consolidation tender

In 2019, while I ran our Civic group, a provincial government decided to consolidate the case-management systems used by its administrative tribunals. Eleven tribunals used software from two of our businesses, and their users were by every measure satisfied. The province held a single competition for a consolidated platform, awarded it to a large integrator offering a configurable product of its own and gave the tribunals three years to migrate. We bid as a subcontractor and then on our own, and lost both times. About C$11 million of annual recurring revenue left over the next three years.

Nothing in our acquisition files for those two businesses had flagged the risk. The retention tables were excellent. In hindsight the signals were visible: the province had published a digital strategy that mentioned consolidation, two tribunal leaders had moved to a central agency, and the contracts all expired within a two-year window. We had read those signals as background rather than as the most important fact about the businesses. The general managers had noticed some of them and assumed, reasonably, that satisfied users would protect them. Nobody had asked who would actually sign the next contract, and by the time we asked, the answer was an official in a ministry none of us had met.

The pattern is not peculiar to one province. Public bodies everywhere periodically decide that dozens of separate systems should become one, that a shared-services organisation should own purchasing, or that a cooperative purchasing arrangement should replace individual tenders. Sometimes consolidation fails and customers drift back. Often it succeeds, and the incumbent suppliers of the separate systems lose their customers in a single event. A portfolio of public-sector software businesses is therefore less diversified than its customer count suggests, because many customers answer to the same buyer.

Underwriting the buyer, not the customer

Since 2021 every acquisition Sorrelgate considers in a public-sector vertical carries a buyer map. It identifies, for each significant customer, who actually makes the purchasing decision, which contract vehicle the business sells through, when the relevant contracts and vehicles expire and whether any level of government above the customer has signalled an intention to consolidate. We then measure the share of revenue sitting behind any single purchasing authority. Where that share exceeds about a quarter, we pay less, ask the seller to defer part of the price against a defined event, or decline.

The test has cost us acquisitions. We have declined or lost six since 2021 because of it, including two businesses with outstanding customer satisfaction whose revenue rested almost entirely on one provincial vehicle. A competitor later bought one of them at a price I thought fair, and I accept that we may have been too cautious there. But none of the businesses we have bought under the test has yet lost more than 5 per cent of its revenue to consolidation, and our board can see the exposure for the whole portfolio in one table.

The buyer map also changes what we do after an acquisition. A general manager who knows that a state technology office is reviewing systems across its counties can join that conversation early, offer to be part of a shared solution or build the integration that makes consolidation unnecessary. Our US businesses now track the cooperative purchasing contracts through which many counties and school districts buy as carefully as they track their own renewals.

There is an uncomfortable implication for sellers, and I try to be candid about it. A founder whose business sells mostly through one provincial vehicle will receive a lower price from us than the retention table seems to justify, and some have told me so with feeling. I explain the buyer map to them, show them the tribunal story and offer a deferred payment that rewards them if the vehicle survives. About half accept that structure. The rest sell elsewhere, which is their right.

I would not argue that public-sector revenue is overvalued in general. It remains durable, and the customers are among the most loyal anywhere. My argument is narrower. When you buy a business that sells to governments, you are buying its relationship with a purchasing system as much as with its users, and that system answers to policy, budgets and politics rather than to satisfaction scores. Price the system. The customers will usually look after themselves.