The decisions a permanent owner of software businesses cannot delegate

Because we intend to own what we buy indefinitely, a mistake in price is never corrected by a later sale; it is carried for decades. I therefore treat our acquisition hurdle as a commitment to shareholders rather than a negotiating reference. Sorrelgate requires an 18 per cent after-tax internal rate of return on a conservative base case built from cash the business has already collected, not from the seller's plan. Group presidents can approve purchases up to C$20 million within that hurdle; I chair the review for those between C$20 million and C$75 million, and larger ones go to the board's investment committee. Every acquisition is revisited three and five years after closing in a cohort review comparing projected and realised returns. That review has led us to stop buying in four verticals since 2013, and it is the paper I would most want a new director to read.

About 54 per cent of our revenue comes from public bodies, so I spend much of my time on a risk that rarely appears in a seller's data room: the way governments buy. For each public-sector acquisition we build a buyer map showing who makes the purchasing decision above each significant customer, which contract vehicle the business sells through, when contracts and vehicles expire and whether a higher level of government has signalled consolidation. We measure the share of revenue behind any single purchasing authority and price, structure or decline accordingly. Inside the portfolio the same map tells general managers where to watch. I learnt this from both sides, pricing bids for an integrator and later losing C$11 million of recurring revenue when a province merged its tribunal systems. I describe procurement rules in general terms, because they differ by province, state and county and change often.

Our US revenue rose from 19 per cent of the total in 2020 to 36 per cent in 2025, almost entirely through small acquisitions in verticals we already understood in Canada: municipal permitting, court records, building-supply dealers, school administration and agricultural co-operatives. I argued for that route over a large platform purchase because our advantage lies in knowing a vertical's customers, not in running a big American company. The route has its own difficulties. County and state procurement differs from provincial practice, many US public bodies buy through cooperative purchasing contracts, and sales tax, payroll and benefits obligations vary from state to state. We built a deal team of seven in Minneapolis and, later than we should have, appointed US-based leaders inside each operating group. I am responsible for the US acquisition programme and sit on the board of our US holding subsidiary.

Sorrelgate is decentralised by design, and I think the design is right: general managers who know a vertical make better decisions about it than anyone in Ottawa. The hard question is when head office should require something anyway. I apply three tests. The requirement must protect customers or the company from a risk no single business can manage alone; it must be cheaper or better done once than many times; and each business must be able to opt out with a written reason that its group president and I accept. Canadian hosting for public-sector customers' records met all three, and we made it mandatory in 2022 with an opt-out route that five businesses used. A shared payroll system did not, and we left payroll alone. I keep the list of central requirements short and argue against additions, including my own.

Implementation services, especially converting a customer's old records, have long been a large and profitable part of vertical software revenue. AI-assisted mapping and cleansing tools are now removing a large share of those hours, and the commercial question is who keeps the saving. I led our Civic group's move from hourly estimates to fixed, published prices by scope, about a quarter below the previous average, with an hourly rate card retained only for changes in scope because many public bodies' purchasing rules require one. I am now on the harder sequel: how maintenance fees, often eighteen to twenty-two per cent of licence value on older installed products, should be explained as the cost of supporting software changes. My view is that customers will keep paying those fees for longer if we can show them plainly what the fees fund.