Decisions that cost revenue before they improved the buildings

Declining to sell a label estimator before the Dutch office deadline

In early 2022, with the Dutch requirement approaching that most offices hold an energy label of C or better from January 2023, our customers' asset managers wanted one answer from us: which of their buildings would pass. About 410 of the office buildings on our platform had labels below C or labels that had lapsed. A competitor had launched an estimator that predicted a building's label from its floor area, construction year and metered consumption, and our sales director wanted the same, priced at €1,200 a building. He expected €2.6 million of one-off revenue.

I opposed it. A Dutch energy label is calculated by a registered adviser from the building's fabric and installations under a prescribed method; it is not derived from the meters, and a building can consume little and still hold a poor label, or the reverse. An estimate on our screen would be quoted in investment committees as if it were a label, and some landlords would postpone an adviser's visit on the strength of it. Our chief executive sided with sales until we tested a prototype on twenty buildings with known labels and it misclassified seven.

Instead we built two things. The first was a measured-performance view that placed each building's actual consumption per square metre beside the figure its label assumed, so landlords could see where a building did better or worse than its paper. The second was an export that gave a registered adviser the drawings, equipment lists and meter history they would otherwise gather on site.

We gave up the estimator revenue, and two prospective customers with 140 office buildings between them chose the competitor. Advisers used our export for 1,140 buildings in 2022 and 2023. In buildings using the measured view with fault detection, consumption per square metre fell by an average of 11 per cent over two heating seasons, mostly through schedules and setpoints. The gap between measured and calculated performance became the central figure on our platform.

I would do two things differently. I would have involved advisers in designing the export from the start; the first version left out ventilation data they needed, and a third of the early exports came back with questions. And I would have told the two lost prospects plainly why we would not build an estimator, rather than letting the sales team describe it as a roadmap item.

Pricing by floor area instead of by data point

When I joined Lindevaart in 2021, we charged landlords €1.60 a month for every data point connected to the platform: each meter channel, temperature sensor, valve position and alarm. The model had been easy to explain to early customers. By 2022 it was working against them. Our usage data showed landlords connecting fewer points to keep bills down, often leaving out the sub-meters and room sensors that let us find faults. An office of 8,000 square metres might expose 3,000 points from its controllers; the median customer connected about 900. Bills also moved whenever an installer added equipment, which landlords' finance teams disliked.

I proposed pricing by lettable floor area, with a rate per square metre that varied by building type and module and a minimum fee of €2,400 a building. Our chief financial officer disagreed. Forty-one accounts with heavily instrumented buildings would pay less under the new model, removing about €2.3 million of annual recurring revenue at once, and she argued for a hybrid that kept per-point charges above a threshold. The operating partner from our investors shared her concern about that year's numbers. The chief executive backed the floor-area model on condition that we phased it in at contract renewals over eighteen months.

We published our rate card for the first time, trained the sales team to quote from it and gave every existing customer a comparison of both models for its own portfolio. Renewals moved onto the new model between January 2023 and June 2024.

The €2.3 million reduction arrived as forecast. Connected points per building more than doubled within eighteen months, and the share of buildings with active fault detection rose from 38 to 71 per cent. Net revenue retention, which had been 103 per cent, reached 111 per cent in 2025 as customers added buildings and modules. The minimum fee had a cost of its own: about sixty small logistics buildings left the platform, worth roughly €0.4 million a year, because their owners were paying for little more than a monthly meter reading.

I would test a minimum fee on one segment before applying it everywhere. I would also have offered owners of small logistics buildings a lighter product, which we built eighteen months later, instead of letting them leave first.

Giving up our own gateway for open protocols

Until 2024 every building on our platform was connected through a Lindevaart gateway: a small industrial computer that our field partners installed in the plant room to read the building automation system and meters. We sold about 650 a year, and the hardware line earned €4.4 million of revenue in 2023 at a gross margin of 31 per cent. It also made us slow. The median building took 19 days from signed contract to first data, mostly spent scheduling an installer and waiting for site access, and landlords were writing clauses into tenders requiring that a building's data be readable by any vendor.

I proposed that we stop selling the gateway, connect through the open building-automation protocols most controllers already support, using certified third-party gateways where a building needed one, and describe each customer's building in an open, community-maintained vocabulary that it could export with its full history at any time. Our chief technology officer supported the engineering case. Our chief financial officer and the head of our hardware business disagreed: we would lose a profitable revenue line, write off stock and make it easier for customers to leave. The investors' operating partner asked for a churn forecast before the supervisory board would decide.

The board approved the change in March 2024 on condition that we report gross revenue churn every quarter. We wrote off €1.7 million of gateway inventory, offered the hardware team's eleven people roles in integration and field support, and certified twenty-three installers on the new connection process.

The median time to first data fell to 6 days. We connected 1,050 new buildings in 2025, against 610 in 2023. Gross revenue churn fell from 4.1 to 3.6 per cent rather than rising, and we won nine of the fourteen tenders in 2025 that carried open-data clauses. The first six months were rough: certified installers made configuration errors in about one building in eight, and our support team absorbed the cost of correcting them.

I would build the installer certification and a configuration checklist before announcing the change, not after it. And I would have shown the hardware team the plan before the board saw it; two of them heard about it first from a customer.

Reversing an app-only entry system at Hallenbrook

In 2018 Hallenbrook Workspaces replaced key cards at its 38 locations with entry through the member app, using the phone's short-range radio to open doors. I supported the decision. It removed about €380,000 a year of card printing and reception work, and our chief executive described the company as app-first in every investor presentation.

By early 2019 the evidence was poor. In our access logs, 4.8 per cent of entry attempts failed at least once. Members with older phones, visitors without the app and people whose batteries had run down were queuing at reception. In interviews my researchers held with thirty members, a blind member described standing at the door while the app's unlabelled controls were announced to his screen reader as "button, button, button". An external accessibility audit against WCAG 2.1 level AA found 212 defects in the member app.

I proposed restoring a physical credential for anyone who wanted one and rebuilding the app to meet that standard before adding any new features. Our chief operating officer disagreed: cards would bring back cost and complicate an access system we had just simplified. Our chief executive was reluctant to retreat from a position he had taken in public. We compromised on offering cards or fobs free on request, rather than issuing them to everyone.

The change cost about €210,000 in its first year for reader firmware, fobs and reception time, and it delayed our planned room-booking release by a quarter. Failed entries fell to 0.6 per cent of attempts within four months. About 19 per cent of members chose a fob. A repeat audit found nine remaining defects, all minor, which we fixed in the following release. Complaints about entry, our largest single category in 2018, fell below room temperature and coffee by the end of 2019.

What changed was how we decided. Every member-facing change afterwards was tested with members who used assistive technology, and reception staff reviewed each release before members saw it.

I would never again approve removing an old way of doing something before the new way had been used by people unlike the team that designed it. I should have asked for those interviews in 2018, not after the complaints arrived.