Abstract
SwiftRide, a fast-growing mobility platform in Lagos, has introduced an algorithmic scheduling system to optimise rider utilisation, reduce idle time, and improve profitability. From a data perspective, the rollout is a success—dashboards reflect strong performance, and executives hail the system as a leap forward in operational discipline. But beneath the numbers, riders describe a different reality: erratic shift patterns, unsafe driving conditions, physical exhaustion, and the constant fear of deactivation. As reports of accidents and absenteeism quietly rise, Aisha Bello, the company’s Lagos Head of Operations, finds herself caught between data that signals success and workers whose wellbeing is visibly deteriorating. With pressure mounting to scale the system nationwide, Aisha must decide whether to adjust the algorithm, enforce stricter compliance, or reintroduce human oversight—knowing that each option carries risks to margins, morale, or managerial credibility. The case examines the tensions between algorithmic control and human judgement, and the ethical, strategic, and organisational dilemmas that surface when digital optimisation collides with lived experience.
Keywords
Learning objectives
By the end of this case, students will be able to: 1. Analyse the implications of algorithmic management for worker autonomy, managerial control, and operational efficiency in dynamic, high-variability environments. 2. Evaluate the limitations of data-driven performance metrics in capturing human costs, contextual complexity, and unintended outcomes in platform-based service delivery. 3. Critically assess ethical responsibility in digitally mediated work systems, particularly where accountability is diffused across algorithms, managers, and frontline workers. 4. Apply socio-technical systems thinking to explore how algorithmic tools interact with organisational culture, informal practices, and lived experience. 5. Formulate strategic responses to digital transformation dilemmas that balance scalability, performance, trust, and worker wellbeing.
Setting the scene: Scaling control in motion
Lagos wasn’t just busy—it was unruly. The city’s infamous traffic gridlock had humbled even the most aggressive tech interventions, and for ride-hailing platforms like SwiftRide, the streets were both opportunity and chaos. What looked like predictable logistics on a dashboard often dissolved into sweaty hours trapped on Third Mainland Bridge or navigating flooding in Apapa. Yet, in this volatility, SwiftRide had grown fast—fuelled by investor appetite, a rising urban middle class, and the promise of digital solutions that could tame informal systems.
From her office in Ikeja, Aisha Bello, the Head of Operations for SwiftRide Lagos, was one of the key architects of that growth. She had joined the company during its early expansion phase, bringing with her a background in operations management and a reputation for making complex systems legible. For the first 18 months, SwiftRide had relied on semi-manual scheduling and performance tracking—dispatchers with deep local knowledge, WhatsApp coordination, and late-night spreadsheet reviews. It worked—sort of—but it couldn’t scale. The company needed a way to “professionalise” its rider network without constantly expanding the operations team.
Enter the algorithm.
Headquarters in Nairobi had commissioned an AI-powered scheduling system—designed to allocate jobs, calculate rider scores, and maximise throughput. The system, they promised, would bring order, fairness, and data-backed accountability. It was fast, objective, and scalable. And crucially, it removed the bottleneck of human judgement.
For Lagos, the timing made sense. The city’s expansion zones—like Lekki, Festac, and parts of Ajah—were exploding with demand. Manual scheduling was already faltering under the weight of growth. Aisha had pushed for Lagos to be the first to adopt the new system, confident that it would bring relief to both managers and riders.
Lagos operating context snapshot.
Operations performance dashboard (month 3—Lagos region).
Rider voice extracts (anonymised WhatsApp logs).
Internal log—incidents, absences, and exits (post-rollout trends).
Pilot results comparison—Surulere (human override) versus Lekki (algorithm-only).
Back in Ikeja, Aisha found herself increasingly in two worlds: a metrics-driven command centre that promised efficiency—and a volatile, human street system that stubbornly refused to be automated. The question was no longer whether the system could scale, but at what cost.
The algorithm arrives
The rollout was slick—on paper. SwiftRide’s new algorithmic management system, internally branded as OptiRoute, promised dynamic shift scheduling, real-time job allocations, and performance-based incentives calibrated to rider behaviour. The logic was simple: the more efficiently you moved, the more you earned. Acceptance rates, response times, average speed, idle time—all fed into a rider score that shaped the next day’s assignments. The system was designed to learn and adapt, constantly optimising for throughput and profit.
For Aisha Bello, the transition felt like a professional win. She had advocated for Lagos to be the pilot city, pitching the algorithm as a way to relieve dispatchers, reduce operational drag, and introduce “fairness by formula.” No more complaints about favouritism or late assignments. No more panicked calls during shift changes. Everything—from route allocation to bonus eligibility—would now flow through clean, auditable code.
Initially, there was some resistance from the field. Riders were asked to download a new version of the app, which disabled manual shift preferences and introduced mandatory check-ins at the start and end of each shift. But the onboarding came with incentives: early adopters got guaranteed bonuses, fuel vouchers, and premium job allocations. Within weeks, the old dispatch channels went silent.
At headquarters, executives celebrated. Rider coverage improved by 23% in high-demand zones. Idle time dropped. Cancellations reduced. And for the first time in SwiftRide’s history, operations in Lagos were hitting all their KPI targets ahead of schedule.
But the celebrations masked a deeper shift. Supervisors no longer decided who got what shift. Field managers had less discretion. Complaints about traffic conditions were harder to escalate—after all, OptiRoute had already “optimised” for that. Human feedback had become a postscript, not a variable (Figures 1 and 2). SwiftRide scheduling logic (simplified flow diagram). Draft policy memo—rider performance & compliance framework.
For Aisha, the challenge wasn’t that the algorithm wasn’t working—it was. It just wasn’t seeing. By Month 2, riders were taking more trips but logging fewer breaks. Average delivery speeds had increased. But then came a string of warning signs: a near-miss accident in Ikoyi, a heat exhaustion case in Surulere, and multiple riders privately messaging their former supervisors with one-liners like, “I’m not a robot.”
The system was delivering efficiency. But to whom?
What the dashboard says
By the third month of rollout, the data was singing. Aisha Bello’s inbox was full of congratulatory emails from regional leadership. The Lagos office had not only recovered from last quarter’s margin dip, it was now outperforming Accra and Nairobi on key efficiency indicators. The new algorithm appeared to be doing exactly what it was designed to do: eliminate slack, reward fast movers, and squeeze more productivity out of each rider shift.
At the monthly performance review, the Lagos executive dashboard was projected onto the wall. Aisha watched as charts glowed green across every zone—Apapa, Ikeja, Lekki, even notoriously difficult Festac. She could feel the room’s tone shifting. Where Lagos had once been a problem market—too unpredictable, too “Nigerian”—it was now the model for disciplined, data-led delivery.
Her own line manager beamed. “This is the first time Lagos has posted three consecutive weeks above 90% utilisation,” he said. “Whatever pushback you’re hearing from the field, ignore it. This is real progress.”
And that was the heart of it: progress, measured in clean, compelling numbers. The dashboard gave no indication of near-misses, fatigue, or silent resentment. It saw utilisation, order fulfilment, revenue per rider. It didn’t see back pain, skipped meals, or emotional burnout.
Still, in a metrics-driven culture, the numbers ruled. They were portable, impressive, and—most dangerously—believable.
Psurpose of exhibit
Showcases the seduction of surface-level performance data while quietly planting the contradiction—the one red flag (“Rider Absences”) that hints at hidden costs. This contrast sets up the next section, where rider voices push back against the clean metrics.
As Aisha glanced again at the absence data, she made a quick note in her planner. Flag attrition trend. Follow up on causes. She knew what it looked like to everyone else: a blip in an otherwise excellent quarter. But she had a growing sense it was something more. Something the algorithm wasn’t tracking.
What the riders say
By Week 10, the field was restless. Riders who had once joked with dispatchers or lingered at base for tea between jobs were now ghosting the office entirely. The WhatsApp groups Aisha used to monitor chatter had gone quiet, replaced by side conversations she couldn’t trace. When she finally invited a few long-term riders in for a casual lunch meeting, the floodgates opened.
“I don’t rest anymore,” said Dimeji, who had been with SwiftRide since the early days. “Before, I’d take a 15-minute break after a run to Lekki. Now, if I stop, the system drops my score.”
Amaka, one of the few female riders on the team, nodded in agreement. “The shift schedule doesn’t consider weather, flooding, or roadblocks. If your app says ‘start,’ you better start—even if Obalende is underwater.”
Another rider, Bashir, leaned in: “There was an accident two weeks ago. Jide fell trying to beat the time window. Nobody reported it because he didn’t want to lose points.”
Aisha sat quietly, phone in hand, scrolling through internal metrics that showed no such incident. Jide’s performance was listed as “highly efficient.”
The room fell into an uneasy silence.
When Aisha reported the concerns at the next ops meeting, the response was muted. “If they’re not coping, we can onboard new riders,” said one exec. “Don’t let anecdotal stuff derail optimisation.”
But it wasn’t anecdotal to Aisha. It was accumulating. And it wasn’t just about physical strain—it was about voice, trust, and being seen. The algorithm didn’t know what it didn’t know. But she did.
Accidents, absences, and the invisible costs
The data didn’t lie—but it also didn’t speak.
Aisha began keeping her own quiet log. It started after a Saturday morning email from HR noting an “unusual cluster” of sick leave in the Mushin and Yaba zones. Then came a call from a field supervisor about a minor crash near Alausa—“no injury, but the rider was exhausted.” The system hadn’t flagged it. There were no exceptions logged, no delays recorded. On paper, everything looked fine.
She dug deeper.
Since the algorithm’s rollout, rider absences had jumped nearly 60%. Attrition was creeping upward too—often with no formal resignation, just dropped shifts and unreachable phones. Field managers were demoralised. They had gone from being respected problem-solvers to passive observers. One even described the system as “management by blindness.”
When Aisha asked the tech team for explanations, she got defensive jargon: “We’ve optimised for delivery density and cost per kilometre, not health outcomes.” When she suggested adding a variable for fatigue or traffic-related stress, she was told there wasn’t enough structured data to support it.
What she saw were bodies burning out under the weight of a system that saw only numbers. The real costs—physical, emotional, relational—were nowhere on the balance sheet.
Reveals the creeping operational and human costs that the main dashboard obscures. Highlights pattern recognition that only emerges when someone is looking outside the system. Opens the door to debate on what counts as a “performance indicator.”
Aisha stared at her spreadsheet, then at her wall-mounted performance dashboard. The green bars still shone proudly. But her gut was uneasy.
How much longer could this go unseen before something snapped?
Management interpretation: Data or discipline?
When Aisha raised the rising sick days and attrition numbers at the next management call, the room responded with polite indifference. One director suggested riders were exaggerating to avoid tougher shift schedules. Another implied that “soft managers” were enabling complaints instead of enforcing standards.
“Look, Aisha,” her regional lead said flatly, “We’re not running a co-op. We’re scaling a logistics company. If the metrics are green, the system is working. Maybe we need stronger compliance, not softer processes.”
This was the tone that had quietly taken root: data over doubt. Any deviation from the algorithm’s logic was framed as inefficiency—or worse, resistance. Suggestions for rider feedback loops or schedule adjustments were met with concern about “undermining objectivity.” Aisha began to notice how quickly conversations turned defensive the moment the algorithm was questioned.
She reviewed her notes. Across five internal meetings, the term “data-driven” appeared 17 times. “Wellbeing”? Just once.
The dominant framing was clear: • The algorithm = neutral • The riders = unreliable • Management’s job = enforcement
Any other interpretation was seen as sentimental or unscalable.
Yet Aisha couldn’t shake the dissonance. She knew these riders. She’d sat with them during onboarding, visited their homes, listened to them explain how they calculated fuel costs versus earnings. The algorithm hadn’t met their children or seen the state of their tyres. It didn’t know about flooding, local protests, or the unspoken pressure to smile through a 14-h shift.
But as far as senior management was concerned, these things didn’t belong in a performance review. They were distractions from the data.
The temptation of compliance
The pressure to conform wasn’t subtle.
By Month 4, headquarters was asking Aisha to produce a new “discipline framework” for rider management. The Nairobi team had rolled out a stricter set of rules in Kampala—with reported success—and Lagos was expected to follow suit. This time, the directive came with teeth: escalate penalties for missed shifts, reduce tolerance for app non-responsiveness, and reassign low-score riders to “cool-off” zones with fewer orders.
Internally, it was framed as “tightening the system.” But to Aisha, it felt like punishing people for trying to survive.
Still, the logic was hard to argue with. More compliance meant more predictability. More predictability meant smoother scaling. And for a platform chasing regional dominance, scalability was sacred.
She drafted the memo. It outlined proposed thresholds and penalties, all justified with clean KPIs. It looked good. Rational. Manageable.
But as she read it back to herself, something turned in her stomach. There was no mention of fatigue, Lagos traffic, or the riders who quietly logged 12-h shifts to keep scores high enough to feed their families. The document spoke only of “targets,” “behavioural trends,” and “system compliance.”
Aisha stared at the screen. She hadn’t sent the memo yet. It was ready—but she wasn’t sure she was. If she clicked send, the riders wouldn’t even blame her. They’d just adjust. Or leave.
The risk of human overrides
It started with one zone—Surulere.
Aisha instructed a trusted field manager to trial a minor change: manual shift overrides for riders flagged as high-risk—those with recent absences, repeated lateness, or informal complaints about exhaustion. For 1 week, the algorithm’s recommendations were reviewed by a human before final dispatch. Routes with known flooding, bottlenecks, or security risks were reassigned. Breaks were built in manually. It was small. Quiet. Invisible to headquarters.
The results weren’t revolutionary—but they were revealing.
Utilisation dipped slightly. Fulfilment rates dropped by 5%. But the riders who’d been adjusted for? No accidents. Fewer sick calls. A 100% attendance rate that week. And—perhaps most telling—two riders who had ghosted the previous month returned to active duty.
Field managers reported improved morale. “We felt like we were managing again,” one said. “Not just watching the system punish people.”
But the shift exposed the real dilemma: discretion slowed things down. Overrides broke the promise of algorithmic purity. Human decision-making wasn’t scalable, and SwiftRide’s regional expansion plan depended on replicability, not nuance.
And yet… the override worked. It made things better. Just not faster.
Aisha now had proof: injecting human judgement reduced harm. But it also reduced efficiency. The algorithm punished variance. But maybe Lagos needed variance—flex, judgement, care.
If she scaled this override model, she’d have to justify it to HQ. If she didn’t, she’d be complicit in ignoring what the data refused to see.
Her next move wouldn’t just affect metrics. It would shape how the company understood work itself.
The ethical fault line
Aisha couldn’t sleep.
Each version of the dashboard replayed in her mind: the green bars, the rising margins, the smiling performance curves. She thought about Dimeji’s cracked helmet, Amaka’s swollen ankles, Bashir’s silence after three missed shifts. None of it showed up on OptiRoute. The system had no field for exhaustion. No data type for fear. No dashboard tile for dignity.
And yet, the system was working. At least, that’s what she was expected to believe—and defend.
The real weight of leadership was becoming clear to her now. It wasn’t in the targets or the budget lines. It was in the moments where silence felt easier than speaking. Where the data offered plausible deniability. Where she could look away and no one would notice—because the algorithm had already decided what counted.
But she noticed.
If a rider got injured while trying to meet a performance threshold the system had defined, who was responsible? The rider who accepted the shift? The manager who enforced the policy? The engineer who built the logic? Or the executive who ignored every red flag that didn’t appear in a quarterly report?
SwiftRide prided itself on being ethical. It had a code of conduct, a values statement, even a CSR programme supporting clean water access in underserved communities. But none of that answered the question now haunting Aisha:
What does it mean to lead ethically in a system designed to optimise everything except care?
There was no memo for this. No override.
Just a growing awareness that silence was a kind of complicity—and efficiency, when pursued blindly, could become violence in disguise.
Decision point: What should Aisha do?
It was Monday morning, and Aisha had 48 hours before the regional leadership review. She was expected to present Lagos’s Q4 performance—and make recommendations for rollout across additional Nigerian cities. Everything pointed to a green light. Margins were up, KPIs were on track, and headquarters wanted a replicable success story. Lagos had become that story.
But Aisha knew the foundation was cracking.
Riders were burning out. Incidents were underreported. Attrition was climbing. And her quiet experiment in Surulere had revealed what the dashboards refused to acknowledge: when you let humans override the system, safety and dignity improved—even if profitability didn’t.
Now, she had to decide what kind of leader she wanted to be.
Option 1: Enforce compliance
She could finalise the draft policy memo and scale strict enforcement. The data would hold, investor confidence would rise, and Lagos would remain the poster child of platform discipline. Riders would adapt—or leave. Either way, the system would stay “clean.”
Option 2: Adjust the algorithm
She could push for a recalibration of the algorithm itself—downgrading speed, building in rest buffers, and including fatigue proxies. But this meant lower margins, delays in expansion, and intense resistance from the Nairobi tech team.
Option 3: Institutionalise human overrides
She could scale her Surulere experiment, embedding local discretion into daily operations. This would require rethinking what scalability meant—and accepting that some parts of the system couldn’t be automated without human cost.
Each option came with trade-offs: political, financial, ethical. None were risk-free. And Aisha wouldn’t get to revisit this quietly in a backroom strategy session. She had to speak, with clarity and conviction, in front of people who didn’t see what she saw.
She stared at her laptop. Three slides. One recommendation. No way back.
Discussion questions
1. How should managers navigate situations where algorithmic data signals success, but lived experience from frontline workers points to harm, and what organisational, ethical, and strategic risks arise when one form of knowledge is consistently privileged over the other? 2. In platform-based systems like SwiftRide’s, where responsibility is diffused across code, leadership, and frontline actors, how can ethical accountability be meaningfully assigned—and what are the dangers of relying on “neutral” algorithms to make performance decisions? 3. Given the trade-offs between efficiency, scalability, and human discretion revealed in the Surulere pilot, how can organisations design algorithmic management systems that incorporate contextual judgement without sacrificing operational control or strategic clarity? 4. What would it take for a performance dashboard to measure sustainability and care—not just utilisation and output—and how might such a redefinition of value challenge existing power structures and decision-making norms in digital firms? 5. If Aisha chooses to institutionalise human overrides or recalibrate the algorithm, how might that reshape SwiftRide’s operational model, investor narrative, and internal culture—and what does this reveal about the real costs of optimising at scale in volatile environments like Lagos?
Footnotes
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.