Four decisions that cost rent, money or history, and why I made them

Choosing comparison flats a tenant could have chosen

Haselwerder sends about 52,000 rent-increase letters a year. Some 21,000 go to flats that no usable rent index covers, so the increase is justified, as German law allows, by naming at least three comparable flats. In 2018, as head of rent and portfolio analytics, I led the team that automated that choice. The system picked the three highest rents among eligible flats in our portfolio. By 2021, 19 per cent of tenants receiving such letters objected or refused consent, against 8 per cent where we cited an index, and we were bringing about 640 actions for consent a year. A regional tenants' association had shown the same few recently relet buildings appearing in thousands of our letters.

Rent growth drives our income and valuation, but so does the goodwill of towns and tenant groups, and two councils had raised the letters with our chief executive.

In September 2022 I proposed a published rule: the eligible flats nearest the tenant's own on stated characteristics, chosen closest to the median rent, with a one-page summary of the comparison range in every letter. Our chief operating officer, who answers for rental income, opposed it. The law did not require it, competitors would not follow, and it would forgo about €6.1 million a year of increases. Several analysts on my former team defended the original design as lawful. The management board approved the change for letters from April 2023, with a review after a year.

We retrained 140 letter-handling staff, and our central tenants' advisory council rewrote half the summary page. The average increase requested fell from 9.6 to 6.9 per cent. Consent without dispute reached 93 per cent by 2025, actions for consent fell to about 140 a year, external legal costs dropped by €1.1 million and increases take effect about four months sooner. Net of those gains, rent income is about €3.8 million a year lower than the old method would have produced.

The rule is now reviewed yearly with the tenants' association, which in 2024 showed that our construction-year bands treated buildings modernised in the 1990s as untouched.

I would do two things differently. I should have questioned my own design in 2019, when objections first rose. And I would have taken the tenants' council through the proposal before the board decided, not after.

Settling carbon costs where our floor-area records were weak

From January 2023 German law has divided the carbon price on heating fuels between landlord and tenant, in a ten-step model where the landlord's share rises as the building's emissions per square metre of living area rise. For our 8,900 centrally heated buildings we calculate the split in each annual heating statement. In the 2,700 buildings where 31,000 households buy their own gas, tenants can reclaim the landlord's share from us. Both calculations need two figures: the emissions from the fuel, which suppliers state on their invoices, and the building's heated living area. In spring 2024, preparing the first statements under the law, my team found that recorded heated area differed from the sum of leased areas by more than five per cent in 1,240 buildings.

The total carbon cost for 2023 was about €11 million. A wrong area places every flat in a building in the wrong step, and tenants may inspect the documents behind their statements.

Our controlling team proposed using the midpoint of the plausible range and issuing corrections once records were resolved. That was a defensible position. I proposed placing every unresolved building in the step more favourable to its tenants and never reclaiming the difference later. Our chief operating officer worried about a precedent in which every data gap was settled against the company. The chief financial officer accepted my proposal, and we also delayed statements for 310 buildings by six weeks to resolve them properly. The decision cost about €0.9 million beyond our best estimate.

Caretakers and surveyors re-measured attic conversions, cellars and common areas, and my building-records team reconciled them against floor plans and lease schedules. We wrote to the 31,000 gas-buying households with their building's step and how to claim. We processed 11,400 claims in 2024, averaging €31.

Unresolved buildings fell to about 190 by the 2025 statements, and objections to heating statements fell from 6.2 to 3.4 per cent. Heated area now has a named owner in our building records, and any attic conversion or change of use must update it before the next statement.

I would start the reconciliation in late 2022, when the law was passed, not when the first statements fell due. And I would print on each statement whether its building had been classified conservatively; we explained it only to tenants who asked.

Building a digital twin only where the risers were due

In 2022 our technical division proposed a digital model of all 11,600 buildings, every component mapped in three dimensions, at €41 million over five years with an external provider. Investor relations planned to show it at our capital-markets day. We were also preparing a €460 million programme to replace drinking-water and heating risers in 1,840 buildings between 2024 and 2031. Those buildings produced 1,930 water-damage claims in 2022, and a riser replacement takes a bathroom out of use for weeks, so the order of the programme mattered to thousands of households.

I argued against the portfolio model. Fewer than a fifth of our buildings had reliable as-built drawings, so most of it would be assumptions that looked complete. I proposed a condition model of the 1,840 riser buildings only, built from observed evidence, at €9.5 million over three years. Our chief technical officer disagreed. He wanted one model for energy-renovation planning everywhere, and he was right that we would need one eventually. We gave up the portfolio model and the showpiece, and 9,760 buildings stayed on construction-year assumptions. He agreed on condition that we set criteria for extending the model by 2025.

From 2023 we inspected at least one riser per stairwell by camera, 6,900 stairwells in all, and added pressure tests, mandatory legionella results, insurance claims and caretakers' leak reports, each with its date and method.

By mid-2025 all 1,840 buildings were covered for €9.1 million. We moved 212 buildings forward in the sequence and deferred 165 by three years or more because their risers were sound. Water-damage claims in those buildings fell to 1,420 in 2025. The evidence also reached our valuers, which changed how the board saw the programme.

The modelling rule has changed too. Since 2025 a building enters the model when a major intervention is due within five years; 1,100 more join in 2026.

I underpriced the caretakers' time: in the first year we gave them a form but no hours, and their reports stayed patchy until we paid for both. And I should have agreed extension criteria with the chief technical officer at the start; leaving them for three years let a reasonable disagreement harden.

Deleting what we still held about people who had left

In 2023 a data inventory found about 4.1 million documents about former tenants and rejected applicants, spread across shared drives, a retired letting system and scanned archives: copies of identity papers, pay slips, credit-agency reports and caretakers' free-text notes, some describing health or family circumstances. About 61,000 people apply for our flats each year and most are not offered one; some files dated from 2009.

The GDPR permits personal data to be kept only as long as a stated purpose requires. We kept it because nobody had decided to delete it. A breach of those archives would have exposed tens of thousands of people with no continuing relationship with us, and our average reply to an access request had reached 31 days, close to the one-month limit.

I proposed a retention schedule: applicants' documents deleted six months after a flat is let; former tenants' files deleted once the deposit is settled and claims have lapsed, apart from accounting records kept for the periods commercial and tax law require; and caretaker notes replaced by structured fields. Our receivables and legal teams wanted former tenants' files kept for ten years in case of disputes. My own analytics team wanted the applicant history for a model of letting demand and tenant turnover. I overruled my team and, after the data-protection officer's independent advice, the management board approved the schedule with a hold for open disputes.

The work took fourteen months, and by December 2024 we had deleted 3.4 million documents. Applicants now upload pay slips only when they reach the final shortlist of three for a flat.

Access requests are answered in nine days on average, and we store about 70 per cent less personal data about people who are not our tenants. The turnover model, retrained on three years of data, is somewhat less accurate. In 2025 we lost a €4,800 damage claim against a former tenant because a handover record had been deleted under a rule wrongly applied to a tenancy with an open dispute.

Deletion is now a scheduled process reported monthly to the data-protection officer, not a project.

I would build and test the dispute hold before the first deletion run, not during it. And I would have told the analytics team six months earlier that their history would go.