FINANCIAL SERVICES · NATIONAL BANKING NETWORK CASE
“How do we announce the phased removal of 1,200 ATMs in rural areas without triggering a reputational crisis?”
Completed case for a national European banking network.
THE CONTEXT
A rational decision that could destroy ten years of brand equity.
A national European banking network had observed, for three years, the steady erosion of traffic at its ATMs in low-density areas. The average usage rate of machines located in towns of fewer than 3,500 inhabitants was falling by 4.2% per quarter, with an average annual maintenance cost of €38,000 per machine. The finance department had built a multi-year plan to phase out 1,200 machines over 3 years, about 22% of the group's rural fleet. Projected savings exceeded €42 million per year at full run rate, with a compensating investment estimated at €8 million in the first year.
The decision was technically sound. It ran into a question dreaded across the leadership teams: how to announce it without triggering a media and political crisis that would cost more than the intended savings? Two precedents in the sector had left a lasting impression. A first network had announced a comparable removal three years earlier, triggering a mobilization of retirees' associations relayed by the regional press, then the national press. The network had to back down, restore half of the removed machines, and absorb a measurable reputational impact for 18 months. A second network had tried a more careful line, explaining low usage and digital modernity, and had triggered a “condescending abandonment of rural France” reading in the editorials, with national political consequences.
The executive committee had several execution options: announce everything at once and absorb the crisis, phase over three years by sub-region, phase in anonymous territorial waves without a national announcement, or couple the removal with a visible compensation scheme. Each scenario implied a different narrative, a different calendar, a different handling of the press and local stakeholders. The question before the committee was not whether the removal would happen, but how to frame it to maximize acceptance without dragging the brand into a long crisis.
That is the moment in the decision when the network mobilized our system. The brief was precise: simulate the reception of each envisaged scenario, across the rural and urban populations likely to react, with a projection over the 12 weeks following the announcement and a model of the associated media cascades. The result was expected within three weeks, ahead of the plan's presentation to the group's executive committee.
THE INQUIRY
Six insights that overturned the initial plan.
The rejection is not economic, it is symbolic.
Rural segments do not refuse the closure because they actively use the ATM. 73% of them have not used it for 18 months, having switched to supermarket withdrawals or fully digital banking. They refuse the closure because it symbolizes the abandonment of rural France by an institution that claimed to serve them. The rational message about low usage makes the rejection worse: it validates the perception of indifference. The 47 identified motivations converge on this structure: it is not the ATM's function that is at stake, it is the political signal of its disappearance.
Urban seniors are as affected as rural customers.
Counter-intuitively, senior urban customers, even in large cities where no ATM would be removed, identify massively with the fate of rural populations. The shared feeling of fragility transcends geography: “today it is them, tomorrow it is us”. The crisis thus spreads beyond the directly affected geographic perimeter. Communicating only to the territories actually concerned underestimates the real reputational impact, which reaches 34% of the customer base well beyond the targeted areas.
The three-part sequence turns the rejection around.
Announcing simultaneously (a) the phased closure, (b) the rollout of a mobile service designed specifically for these territories, and (c) an operational partnership with a national postal operator to maintain cash access through its branches: cuts the rejection from 64% to 31%. Media coverage shifts from critical to descriptive. The triptych is indivisible: each element taken alone produces an effect inferior to the three combined.
The postal partnership is perceived as a gesture of dignity.
Our system identified that the most powerful lever in the triptych is neither the phased announcement nor the mobile service: it is the operational partnership with the postal operator. The scheme is perceived as a public recognition of everyone's right to access cash, regardless of commercial logic. Mentioning it at the top of the announcement, rather than at the end, completely reverses how the decision is read. A scenario where the partnership is announced last reaches 45% acceptance. The same scenario with the partnership announced first reaches 67%.
Local elected officials are the priority media channel.
A national announcement through the daily press obtains an initial acceptance of 41%. A cascade announcement, first the prefects of the affected departments, then the mayors of the affected towns, then the regional press, then the national press, over six weeks, obtains 71%. The difference comes down to this: local elected officials relay the compensation scheme as a negotiated territorial agreement, not a decision imposed by a Paris headquarters. That reading fundamentally changes the symbolic charge.
The critical window closes at 8 weeks.
Our system measured the opinion trajectory over 12 weeks of projection in each scenario. The tipping point is identifiable: beyond 8 weeks without visible compensation, the rejection becomes structural and no longer recedes, even if a compensation scheme is deployed afterwards. The postal partnership must be operational within 8 weeks of the announcement, or it loses its neutralizing power. This temporal constraint was the single most decisive parameter in choosing the implementation calendar.
THE METHOD
How we built the inquiry.
Our system rebuilt a synthetic population of 1.2 million retail customers of the banking network, calibrated on public data (INSEE, sector demographic panels) and the network's proprietary segments. This population was structured into 12 locally coherent demographic segments: combining age, place of residence, territorial political sensitivity, banking usage, relationship to rurality and media exposure. No personal records entered the system. The coherence of the segments was validated by cross-reference with public sector studies from the Banque de France, Crédit Mutuel Alliance Fédérale, and surveys from the French research center for the study and observation of living conditions.
On this base population, our system individually interviewed 2,400 synthetic customers distributed proportionally across the 12 segments. Each customer was exposed to five communication scenarios, a blunt announcement, an announcement phased by sub-region, an announcement coupled with a mobile service, an announcement coupled with a postal partnership, and a prefects-mayors-press cascade announcement, in randomized order to neutralize learning effects. The dynamic agents followed up with each customer on friction points identified in real time, with an average of 7 follow-up questions per person, surfacing motivations the first round of answers had not revealed.
Our system then projected each scenario's media trajectory over 12 weeks, modeling the likely relays: regional press, retirees' associations, local elected officials, shopkeepers' unions, local social networks, national press. The cascade modeling produced 5 distinct media coverage trajectories, each associated with a projected reputational impact. It is this combination, individual reception plus media trajectory, that identified the prefects-mayors-press cascade scenario as dominating the others on 21 of the 22 analyzed typologies.
THE DEPLOYMENT
What was decided, what happened.
The executive committee retained the dominant scenario identified by our system: a cascade announcement, prefects first, then mayors, regional press, national press, over six weeks. The postal partnership, negotiated in parallel and operational by the sixth week, was placed at the top of the story in every public statement. The mobile banking service, sized to cover the 1,200 affected towns with a weekly rotation, was presented as the second pillar of the compensation scheme. The ATM removal was phased over 30 months, with a faster pace in the territories where the postal partnership was operational soonest.
Deployment began six months after the committee's decision, once the postal partnership was negotiated and the mobile service calibrated. Over the first 24 months, no national media crisis was triggered. Three local episodes of protest were identified, in towns where the postal partnership was slow to become operational beyond the 8-week threshold measured by our system, and were resolved by a targeted acceleration of the compensation scheme. National press coverage remained descriptive: 2 major articles in the daily press, 4 segments on regional television news, no structuring negative editorial.
At 24 months into the deployment, the network's internal measurements validated the robustness of the scheme: the attrition rate of rural customers over 65 remained equivalent to the network's national attrition rate, with no negative outperformance. Customer satisfaction in the affected areas even rose by 2 points, carried by the positive perception of the mobile service. The projected annual savings were achieved at 94%, with a 6-month lag due to the territory-by-territory phasing. On the reputational front, the brand emerged from the deployment period with brand equity up 3 points on territorial-proximity indicators, as measured by the usual sector studies.
- INITIAL VS POST-PIVOT ACCEPTANCE
- 36/100 → 67/100
- NATIONAL MEDIA CRISIS RISK
- 78% → 0%projected vs observed
- ANNUAL SAVINGS ACHIEVED AT 24 MONTHS
- €39.5M94% of projection
- ATTRITION, RURAL CUSTOMERS 65+
- equivalent to nationalno negative outperformance
- TERRITORIAL-PROXIMITY BRAND EQUITY
- +3 ptsat 24 months
- SIMULATION COST VS AVERTED CRISIS
- 1 : 87
THE LESSONS
Three principles transposable to other sectors.
The rejection of a rational decision is rarely rational itself.
In this case as in most of the high-stakes decisions we have tested, the first lesson is that resistance to well-founded technical decisions is not fought on the terrain of figures: it is fought on the terrain of symbols. The rational message explaining the soundness of the decision worsens the rejection because it implicitly validates the grievance. Any sensitive decision must first be framed in the register in which it will be received, before being framed in the register in which it was decided.
The media channel is inseparable from the content.
The channel/content distinction, common in communication strategies, proved artificial. The same compensation scheme announced by a local prefect or by a headquarters press release produces opposite qualitative effects on opinion. The cascade, local elected officials first, press second, is a narrative device as much as a media one. It turns an imposed decision into a negotiated territorial agreement.
Measurable critical time windows exist.
Our system identified in this case an 8-week threshold beyond which the rejection becomes structural. These critical windows exist in almost every case we have modeled, but their duration varies by sector. In crisis communications, they are counted in hours. In industrial transformation, in months. In public policy, in half-years. Anticipating these windows and calibrating execution accordingly is often more decisive than the content of the scheme itself.
GET STARTED
Preparing a decision of this kind?
This case illustrates a recurring class of decisions in the banking sector, but also in territorialized public services, energy infrastructure, and planning policy. Every decision is singular: populations, territorial symbols, calendar room for maneuver, reputational constraints. Every simulation is calibrated for the precise decision on your mind.
The dynamic agents scope with you the parameters of a simulation adapted to your situation, ahead of the decision. From initial brief to first deliverable, allow 20 to 30 minutes, depending on the case's complexity and the breadth of the populations to model.