The dependencies I want made visible

At the end of 2022, our twenty largest depositor groups represented 38% of customer deposits. The balances were associated with valuable commercial relationships, and many had been with Fennervale Commercial Bank for years. Treasury considered the concentration manageable under its existing scenarios. I was less comfortable with the way those scenarios separated accounts controlled by the same owners. The issue was not that we expected those customers to leave. It was that our confidence depended on knowing more about their behaviour than our records actually demonstrated.

I asked for a joint review of account ownership, withdrawal authority and the commercial events that could cause balances to move. Several relationship managers worried that the exercise would be interpreted as distrust. The finance team also showed that replacing some deposits with a broader mix would raise annual funding expense. We estimated an additional C$8.6 million at the planned balance level. I took that cost to the board with the concentration analysis instead of presenting diversification as a free improvement.

We introduced group-level monitoring, stopped matching every exceptional rate request and invested in transaction relationships with smaller commercial customers. Existing large depositors retained named contacts and clear explanations of our pricing decisions. The transition was deliberately gradual; a sudden rejection of balances would have created the very funding pressure we wanted to avoid.

By June 2026, the largest twenty groups represented 24% of C$13.2 billion in customer deposits. Some of the reduction came from overall deposit growth, and I distinguish that from deliberate balance replacement. The annualised funding cost increase was C$7.9 million against the original estimate. The bank gained a clearer view of connected accounts and more room to manage individual withdrawals. I would start customer explanations earlier if repeating the programme. Our internal analysis had advanced faster than the language relationship managers could use with confidence.

A portfolio review in 2023 identified C$1.35 billion of commercial exposure whose repayment depended, directly or indirectly, on the same regional construction ecosystem. The borrowers operated in different recorded sectors. Their common exposure emerged through supplier relationships, project timing and property collateral. The aggregate was larger than the executive team had understood from conventional sector reporting, although individual facilities had been approved within delegated authorities.

The immediate choice was uncomfortable. We could sell or rapidly reduce facilities and accept pricing losses, or continue renewing them while hoping that separate underwriting decisions would control the combined risk. I rejected both extremes. With the chief risk officer and commercial leadership, I proposed a thirty-month reduction using scheduled maturities, tighter conditions for incremental lending and active support for customers seeking other funding sources. That approach preserved relationships but kept concentration on our balance sheet for longer.

Some commercial executives argued that we were giving competitors our most knowledgeable customers. I asked them to identify which facilities had repayment sources independent of the shared construction cycle. Where the evidence supported that distinction, we retained capacity. Where it did not, we set explicit renewal boundaries. The board received a quarterly bridge showing repayments, new drawings and movements in connected exposure, including exceptions that had increased the total.

By June 2026, exposure had fallen to C$980 million without a forced sale. We recorded lower origination income during the transition and did not claim that the reduction eliminated credit risk. The more lasting change was a requirement to discuss economic connections before a facility reached the approval meeting. Looking back, I would invest sooner in maintaining the underlying relationship data. Our first review depended too heavily on a few experienced bankers whose knowledge was not yet captured in a form others could challenge.

When I became chief operating officer in 2016, Fennervale Commercial Bank could report the volume of service requests more reliably than the number still waiting for a satisfactory resolution. Experienced branch teams used personal contacts to accelerate difficult cases. That helped individual customers but made the underlying process hard to see. A request could be transferred several times without anyone recording that the customer still regarded it as open.

I sponsored a common case register across all 62 branches and the central operations teams. We defined completion from the customer's required outcome, with checks for regulatory and documentation obligations, rather than from the last departmental action. The contentious decision was to publish aged unresolved cases to the executive operating meeting. Some managers feared that a more honest record would make their performance appear worse. I agreed that the first three months would establish a baseline and would not be used to rank branches.

We then assigned one accountable owner to each case, even where several teams had work to perform. Frontline colleagues helped simplify the categories after the first version proved too elaborate. We removed two approval steps that added waiting time without changing the underlying control, while retaining the evidence those controls were meant to protect. Training concentrated on awkward cases, not only the straightforward examples used in launch presentations.

Over eighteen months, median resolution time fell from 4.8 to 2.1 business days. The share of cases older than ten business days declined from 17% to 6%. We separately tracked reopened cases so that premature closure could not improve the headline measure. The programme changed the discussion from who had processed a request to who remained responsible for the result. My main correction would be to involve smaller branches earlier; their limited staffing made some of our initial hand-offs impractical.

Our payment-recovery documentation described system restoration clearly but said less about customer decisions during the interruption. In 2024 I asked for an exercise centred on a payroll processing disruption affecting commercial customers on a Friday. The technical team preferred a narrower test of infrastructure recovery. I understood the appeal: the narrower exercise had measurable boundaries and would not interrupt relationship managers during a busy period. It would also leave our most consequential assumptions untested.

We allocated a half-day exercise involving operations, technology, treasury, legal, communications and a sample of branch colleagues. The scenario used synthetic customer records and did not touch live payments. Participants had to establish which files were affected, decide whether an alternative process was safe and explain what customers could reasonably expect. Independent observers recorded decision times and differences between what each function knew.

The first exercise took ninety-four minutes to produce a consistent customer message. Two teams believed that another team owned approval of the fallback process. We found a contact list that could not be accessed through the assumed recovery channel. These were not reasons to describe the exercise as a failure and move on. I required named corrections, completion dates and a repeat test before authorising the next discretionary service enhancement.

The repeat exercise reduced the time to a consistent message to thirty-one minutes and removed the ambiguous approval hand-off. That does not predict performance in a real incident, where facts and pressures will differ. It demonstrates that we corrected specific weaknesses under observed conditions. The investment decision also became clearer: some resilience spending belonged in access and operating procedures rather than additional infrastructure. Next time I would include a customer representative in the review of our messages; internal clarity does not necessarily produce useful external language.

At Tarnwick Credit Group in 2013, a proposed expansion of equipment finance promised attractive origination growth. Our sales pipeline supported the ambition, but servicing teams were already carrying a backlog of documentation exceptions. Most exceptions did not indicate an immediate loss. Together they meant that the operating record was less reliable than the credit approvals suggested. I was chief commercial officer, so slowing the proposal would reduce the growth for which I was directly accountable.

I asked the executive committee to defer entry into two additional regions for six months. The chief executive challenged whether I was using a control problem to avoid a commercial stretch. I brought a sample of complete files rather than an aggregate backlog slide. It showed how an apparently minor missing condition could delay a drawdown or make subsequent monitoring uncertain. We agreed to keep developing prospective relationships while postponing commitments that would add to unresolved work.

I moved eight experienced commercial colleagues into a temporary file-resolution team and paired them with servicing specialists. We classified exceptions by consequence, contacted customers for missing evidence and established a clear owner for each recurring documentation issue. I also changed sales reporting so that a facility counted as fully established only when the agreed operating conditions were complete. That reduced the incentive to celebrate approval while passing the difficult work downstream.

The exception backlog fell from 640 to 185 files over six months. The expansion resumed in one region first, with the second following after another quarter of stable servicing measures. We gave up an estimated C$3.4 million of near-term revenue compared with the original plan. The business gained a repeatable entry standard and fewer avoidable customer delays. I would now put more weight on the workload transferred to the temporary team; several colleagues carried their original responsibilities for too long before I corrected the staffing arrangement.