Homes passed is a construction figure, not a customer figure

Every cable and fibre operator I know opens its lender presentation with the same slide. It shows homes passed, the number of addresses the network runs past, and beside it the number of broadband subscribers, usually turned into a penetration rate. For years I prepared that slide without asking much about either figure. I now think the first is a construction statistic presented as a commercial one, and the second is often a billing statistic presented as a count of customers. Neither is dishonest. Both can be quietly wrong, and the error tends to run in a flattering direction.

What a passing really counts

A passing is recorded when the network is built along a road or into a subdivision. The engineering system marks each address as passed, and the total only ever rises. At Ridgeburn, when I arrived in March 2021, it stood at 1,080,000. When my team matched it against county property records and our own installation history, about 61,000 of those addresses could not be connected at our standard cost and within our standard interval. Some were lots in subdivisions where nothing had been built. Some were farmhouses set far back from the road, where a connection would have cost several thousand dollars that the household would have had to pay. Some were buildings destroyed by a tornado in 2019 and never rebuilt.

None of those addresses could produce revenue next quarter, yet all of them sat in the denominator of our penetration rate and in every valuation argument that relied on it. They made penetration look lower than it was, and so made the remaining opportunity look larger. A reader who accepted the slide would conclude that about 720,000 homes and businesses were left to win, when the realistic number was nearer 676,000.

The subscriber figure had a different problem. Our billing system counted one subscriber for each account with an active broadband product. That included 9,800 apartment units in buildings where the landlord paid us a bulk rate and the tenant had never chosen us, about 4,500 accounts on free service, mostly storm-relief credits and staff lines, and a few hundred business customers counted once for each site. We restated the year-end 2021 figure from 357,200 to 342,900. The adjustment was 4 per cent. It mattered more than its size suggests, because the excluded accounts behaved nothing like paying households. They never left, so they flattered our churn rate as well as our count.

Fixed wireless found the soft spots

In 2022 the large mobile carriers began selling fixed-wireless home internet across most of our towns, at a flat monthly price, with no contract and a receiver the customer installs. It is slower than our service and varies with the load on the cell, but for many households it is enough, and the price is easy to understand. Our net additions fell from 8,900 in 2021 to 1,200 in 2022, and in 2023 we lost 9,400 customers.

Had we still been reporting the old figures, those losses would have looked smaller and arrived later. Bulk units and free accounts do not leave for a mobile carrier, so they would have cushioned the decline, and thousands of accounts that could not churn would have diluted our churn rate. The board would have seen a softening where there was a break. Because the base had been cleaned a year earlier, the 2023 losses reached our reports at their true size, in the quarter they occurred, and the board judged the next fibre tranche with a realistic view of what it was defending.

Fixed wireless also taught us where to look. Our losses were concentrated among customers who had been with us for less than two years, on our slowest tiers, in towns where we had raised prices twice since 2020. Long-standing customers on faster tiers barely moved. A single churn rate hides that pattern. Churn by tenure and tier shows it within a quarter, and tells a board which customers its next capital dollar is meant to keep.

There is a further trap in the way fibre upgrades are reported. When an operator moves an existing customer from coaxial cable to fibre, some presentations count a fibre addition. It is nothing of the kind. It is a migration, valuable because it reduces churn and allows the old plant to be retired, but it adds no customer. We report fibre migrations and new fibre customers on separate lines, and only the second counts towards the take-up test that releases our next tranche of construction.

The measures I now put first

Our board pack and quarterly lender report now open with five measures, each with a written definition that our auditors have read.

The first is serviceable locations: addresses we can connect within our standard interval and cost, reconciled each quarter between the network inventory, the billing system and property records. Homes passed still appears, underneath, labelled as a construction figure. The second is paying customer relationships, counted once per household or business whatever they buy, with bulk units and non-paying accounts on separate lines. The third is penetration, calculated on serviceable locations and shown by town wherever we face a fibre or fixed-wireless competitor.

The fourth is churn by tenure band and speed tier, because an average rate tells a board almost nothing about which customers are at risk. The fifth is revenue per customer relationship, shown with and without equipment and late fees, since those fees rise when customers rent more equipment or pay late, and neither should be mistaken for pricing power.

These measures were not free. The restatement rebased our management incentive plan, including my own award, and Granmoor marked its holding down when its valuation team applied the same multiples to smaller figures. Our chief executive argued, fairly, that competitors would keep reporting the larger numbers and that we would look worse beside them. In some lender meetings we did.

They have nonetheless earned their place. When we asked lenders to finance fibre in 2024, the first question in every credit committee was whether our customer figures could be trusted, and we could hand over three years of reconciled history. For a director the test is simple. Ask management to define a customer and a passing in one sentence each, then ask who reconciles them, how often, and against what source outside the billing system. If the answer takes more than a minute, the first slide in the pack is measuring construction and accounting habits rather than demand.