Three announcements that told customers the cost

Putting the twelve-month condition into the savings headline

In September 2023 competition for Belgian savings was unusually intense. A one-year government savings issue had drawn large sums out of bank accounts, and several banks were advertising combined rates on their regulated savings accounts. Treasury approved a base rate of 1.40% and a fidelity premium of 1.00% for our main account, and the agency's draft led with '2.40%' in large type.

The figure was accurate for money left twelve consecutive months. It was not accurate for a saver who moved the money sooner, and our contact centre had logged roughly 900 calls after the spring campaign from customers asking why a balance had earned less than advertised. I proposed '1.40% from the first day, plus 1.00% on money that stays twelve months', with a worked example: €10,000 kept for a year earns about €240 before tax; moved after six months, about €70.

The director of savings and the treasurer disagreed. They expected competitors to lead with combined rates, and the agency predicted a response 20 to 25% lower. The chief executive accepted my proposal provided we tracked inflows and retention together, and the €600 million plan stayed unchanged.

The ten-week campaign brought in €468 million, short of plan. Twelve months later 84% remained, against 72% from the spring campaign, although rates and competitors had both changed in between. Calls about earnings below expectation fell to about 210 over the following year.

The campaign also exposed a fairness problem. About 164,000 customers still held an older savings account paying a combined 0.50%. We wrote to them in January 2024, explaining the newer account and what moving would mean for any premium already accruing. By June, 58,000 had moved €1.9 billion, adding roughly €13 million a year to our interest expense.

I would write to those existing customers before the campaign, not three months after it. Advertising a better account to strangers while long-standing members waited was a sequence I should have refused.

Closing fourteen branches with the distances in the letter

By late 2023 the fourteen smallest branches in our network averaged 160 customer visits a week, against about 520 in 2016. The network and finance teams recommended closing them in 2024, and the executive committee agreed. About 52,000 customers had one of those branches as their home branch, including 7,400 people over seventy-five who had visited in the previous six months.

Our previous closures, in 2020, had been announced four weeks ahead in a letter that spoke of 'serving you differently' and gave no reasons. Complaints exceeded 360, and several mayors learnt of the closures from residents. I did not dispute this round's decision. I disputed how we proposed to announce it.

I asked for twelve weeks' notice and a letter specific to each branch: the visit numbers behind the decision, the nearest branch and its distance, the bus connection, and whether a cash machine would remain within five kilometres. In six of the fourteen villages none would, and we said so. The network director wanted six weeks and no visit figures, arguing that the numbers would invite criticism in the local press. The finance director questioned €520,000 a year for home visits by appointment and weekly service hours in six town halls and community centres. I met eleven mayors before the letters went out, and the executive committee approved the longer notice and one initial year of service hours.

In the six months after the closures, deposits held by affected customers fell 2.4%, against 4.8% after the 2020 round, though the rate environment was very different. Complaints numbered 174. Service hours averaged nineteen visitors a week; two locations ended after nine months with fewer than six, which we announced eight weeks ahead. Colleagues made 1,150 home visits in the first year.

We counted contacts and complaints, but not whether older customers could still manage their money. Next time I would involve municipal social services and debt advisers from the start, and follow a sample of affected customers for a full year.

Withdrawing a carbon claim nine days before launch

In early 2025 we prepared a renovation loan for homeowners improving insulation, heating or glazing. The agency's campaign led with 'Save up to three tonnes of CO₂ a year' and called it 'the loan that makes your home sustainable'. The tonnage came from the agency's modelling of an average improvement in Flemish energy-performance certificates. We had never collected a certificate after a renovation, so we had no evidence that our borrowers achieved anything like it.

I reviewed the final materials in March, nine days before launch. The question was not whether the modelling was reasonable but whether we could defend the sentence to a customer, a supervisor or a consumer organisation. We could not. The head of sustainability argued that the estimate was conservative and that competitors made similar claims. The head of lending worried that any evidence requirement would deter applicants. Stopping meant writing off €260,000 of production and delaying the launch by five weeks.

The executive committee supported the withdrawal, partly because I brought a replacement rather than only an objection. The product team agreed a discount of 0.30 percentage points on the loan rate when the renovation plan showed an expected improvement of at least one certificate class, with the new certificate due within twenty-four months. The advertising said exactly that. It carried no carbon figure and no general claim about sustainability.

In its first twelve months the loan financed 1,640 renovations, averaging €31,500, against a plan of 2,100. By September 2026, 610 borrowers had sent a new certificate, and 540 of those showed an improvement of at least one class. Those are the first renovation results the bank can actually state, and they now appear in our annual report with their limitations described.

The evidence should have been designed into the product before marketing was briefed, not rescued at the final review. I would also test whether customers understood what a certificate class meant before building a discount around it.