- Prepared by
- Mugalula Nickson and Tendo Michael
- Organisation
- Loft Technologies, June 2026
- Series
- White paper · Stadium access intelligence
Executive note
This paper is about learning from real stadium access pressure
This white paper is written from the practical side of stadium access control. It is not presented as a laboratory paper on simulation. It is a lessons paper: what we have learned from planning and operating digital ticketing, scanning, turnstiles, handheld validation and live access control for high-pressure stadium events, and what other venues can take from that experience.
Our strongest lesson is that access control is not only a question of whether a ticket is valid. It is a question of whether a valid ticket can be turned into a physical admission quickly enough, at the right gate, under real arrival pressure, with enough stewarding and hardware capacity to keep queues moving.
Boids-based crowd modelling is useful because it translates individual movement into system behaviour. A fan chooses a route, slows down near a queue, follows a group, reacts to signage, avoids dense areas when possible, and is influenced by nearby people. These simple behaviours can produce pressure pockets, blocked approaches and cross-flows that are not obvious from ticket sales data alone.
- 60%arrival shock in final 30 minutes
- scan-to-admitthe real access metric
- boids + logsmovement model grounded in operations data
Reader value
What the reader can pick from this paper
The reader does not need to adopt every concept in this paper. A stadium operator, federation, ticketing provider, security commander or venue consultant can pick one practical improvement and apply it immediately.
For stadium operators
Stop planning with total capacity alone. Break every event into gate-by-gate demand, gate-by-gate throughput and gate-by-gate risk. A stadium with enough total lanes can still fail at one overloaded gate.
For federations and event owners
Ticket categories, complimentary allocations, accreditation flows and sales closure rules are crowd-control decisions. They must be approved early enough to influence gate design and communication.
For security and stewarding teams
Failed admissions are not only ticketing problems. Every failed scan consumes lane time, creates argument time and pushes pressure backwards into the queue.
For technology teams
The access-control dashboard should show admission pressure, not only scan totals. The useful question is what is forming now and what will happen in the next 10 minutes.
The best stadium access-control system is not the one that only counts entries. It is the one that helps the operations team understand pressure early enough to act.
Access pressure
The central lesson: pressure forms when demand outruns admission capacity
Crowd risk near gates is rarely caused by attendance alone. It is caused when too many people arrive at the same place faster than the venue can validate, admit and disperse them. The failure point is the gap between arrival demand and effective admission capacity.
Late-arrival stress assumption: 60% in final 30 minutes
For a 30,000-attendee match, this assumption places 18,000 people into the last 30 minutes. That is 600 arrivals per minute across the whole stadium footprint. If those arrivals are unevenly distributed, one gate can become dangerous even if total stadium capacity looks adequate.
| Planning view | What it misses | Better question |
|---|---|---|
| Total tickets sold | When people arrive and which gate they use | How many people arrive per gate per minute? |
| Total turnstiles installed | Open lanes, device uptime and human passage time | How many actual admissions can each gate process now? |
| Scan count | Failed attempts, wrong-gate pressure and physical passage delays | How many scans become admitted people? |
Formula framework
The practical maths of stadium ingress
The formulas below are not meant to make crowd control look complicated. They are meant to force the right operational questions. Every variable should be measurable from ticketing logs, device logs, steward reports or field observation.
Arrival rate at gate g in time window t. This converts attendance expectation into operational pressure.
Total admission cycle time. This is the full time from ticket presentation to the lane being ready for the next person.
Effective gate throughput. It combines active lanes, cycle time, success rate, uptime and crowd-readiness conditions.
Queue growth. If arrivals exceed throughput, the queue does not merely remain. It compounds over time.
Worked example
A 30,000-person match under late-arrival pressure
Assume a high-demand football match with 30,000 expected attendees and a late-arrival pattern where 60 percent arrive in the final 30 minutes. The stadium therefore needs to absorb 18,000 people in 30 minutes.
| Input | Assumption | Operational meaning |
|---|---|---|
| Expected attendance | 30,000 | Total people to admit |
| Late-arrival share | 60% | 18,000 people in final 30 minutes |
| Peak arrival rate | 600/min | Demand across the footprint |
| Active lanes | 63 | Turnstiles and equivalent admission lanes |
| Cycle time | 6.4 seconds | Presentation, validation, unlock, pass and reset |
| Success rate | 92% | Share of attempts that become clean admissions |
| Uptime | 95% | Devices and lanes available |
| Crowd-readiness factor | 92% | Tickets ready, phone brightness, clear movement, low confusion |
Throughput sensitivity
Small quality losses become large capacity losses
The admission system is sensitive to scan-to-admit success rate. When a venue is under late-arrival pressure, a few percentage points of failure can be the difference between flow and backlog.
Success-rate sensitivity: small drops become large capacity losses
Success-rate loss is not only caused by fraudulent or invalid tickets. In field operations it can come from wrong-gate presentation, fans searching for tickets at the scanner, unreadable screens, duplicate screenshots, category mismatch, slow network response, steward uncertainty, or physical hesitation at the turnstile.
Operational leaf
Report success rate by gate and by 10-minute interval. A stadium-wide average hides the gate where pressure is forming.
Technology leaf
Treat every failed scan as a queue event. Failed attempts should feed the live pressure model, not only the audit report.
Queue compounding
Delayed admission turns small deficits into visible crowd pressure
A gate can appear stable for a few minutes and then become difficult quickly. The reason is compounding. When capacity is below demand, the unprocessed people do not disappear. They remain in the approach, making every next minute harder to manage.
Queue growth under a 600 people/minute peak arrival shock
Admission pressure ratio. Values above 1.0 indicate that the gate is receiving more pressure than it can process.
Lane planning
The number of open lanes must follow the peak, not the average
The lane requirement formula helps convert peak demand into an operational staffing and equipment plan. It also shows why slow physical admission is expensive. Each extra second in cycle time increases the number of lanes required.
Minimum active lanes required for a target peak arrival rate. This should be calculated per gate, not only for the whole stadium.
Lane requirement rises sharply as the admission cycle slows
This chart gives a simple lesson: improving cycle time can be as valuable as adding hardware. Faster ticket retrieval, better feedback screens, upstream filtering and clearer gate assignment can reduce the number of lanes required to handle the same crowd.
Time to clear
Backlog must be translated into minutes, not only people
Queue length matters, but decision makers often respond better to time-to-clear. A queue of 2,000 people means different things depending on whether the gate has a net clearing capacity of 50 people per minute or 200 people per minute.
Time-to-clear estimate. This applies only when throughput is greater than arrival rate. If it is not, the queue is still growing.
Time-to-clear makes backlog visible to decision makers
Time-to-clear should be shown on the live dashboard because it converts technical conditions into operational urgency. If the gate will clear after kickoff, the venue must decide whether to reroute, open more lanes, delay entertainment sequencing, strengthen outer filtering or communicate clearly with the crowd.
Boids translation
How boids becomes useful for stadium access
Boids modelling starts with simple local rules: separation, alignment and cohesion. For stadium operations, these behaviours must be translated into practical crowd movement logic.
Boids layer: individual rules reveal system-level flow
Each agent has a ticket, route, walking speed, group behaviour and response to congestion.
| Boids concept | Stadium translation | Operational use |
|---|---|---|
| Separation | People avoid direct collision and uncomfortable density | Estimate where queues widen, bend or push into roads |
| Alignment | People follow the visible flow of nearby fans | Predict herd movement toward wrong gates or short-looking queues |
| Cohesion | Groups try to stay together | Model family, fan club and group arrival behaviour |
| Destination seeking | People move toward assigned gate or visible entrance | Test whether signage and ticket instructions are clear enough |
| Rerouting | People change path when congestion becomes obvious | Prevent uncontrolled movement from one overloaded gate to another |
Boids formula
The movement layer should be explainable
A useful simulation must be explainable to operations teams. The velocity formula below is a simplified way of saying that each fan’s next movement is influenced by nearby people, group behaviour, gate destination and rerouting signals.
Simplified agent movement update. Separation, alignment, cohesion, destination-seeking and rerouting forces combine into the next movement direction.
The weights should not be treated as universal constants. They should be calibrated by venue type and event context. A derby match, family event, concert, international fixture and VIP-heavy event will all produce different movement behaviour.
Before the event
Use the model to test gate assignment, approach routes, barricade placement, signage points and steward positions.
During the event
Use actual scan rates, queue estimates and density observations to update risk assumptions and recommend interventions.
Spatial risk
Risk is spatial, not only numerical
A gate with 2,000 people waiting may be manageable in a wide, well-barricaded forecourt. The same number can be unsafe if it is compressed into a narrow approach road, stair base, wall edge or merge point. That is why the model must show spatial pressure.
Density in a zone. The same number of people can be safe or unsafe depending on the area and movement conditions.
Spatial pressure map: the danger is where density forms
Boids-based modelling adds value here because it shows how people redistribute when one route becomes dense. If the simulation predicts that an overloaded gate spills into a road or blocks an adjacent gate, the operations team can change the geometry before event day.
Delayed admissions
What actually slows the lane
Delayed admission is a full-chain problem. It can start before the fan reaches the scanner and continue after the ticket has been accepted. The access-control team should therefore measure causes of delay in the whole gate environment, not only at the validation API.
Where delayed admissions can come from
| Delay source | What it looks like | Countermeasure |
|---|---|---|
| Ticket not ready | Fan reaches scanner then searches WhatsApp, email or gallery | Pre-gate reminders, holding screens, ticket-help desks upstream |
| Wrong gate | Valid ticket presented at wrong category or zone | Gate printed on ticket, WhatsApp reminder, outer steward filtering |
| Validation delay | Scanner waits for response or retries | Edge validation, offline cache, redundant networks |
| Turnstile hesitation | Accepted ticket but user delays passage | Feedback screens, better lane design, steward cueing |
| Exception handling | Dispute, override, accreditation issue or duplicate ticket | Remove exceptions from main lane immediately |
Turnstile behaviour
The turnstile is both a validator and a traffic signal
The moment after a valid ticket is presented is critical. The system must unlock fast enough, stay open long enough for safe passage, reset quickly, and give clear feedback if the ticket is not accepted. The turnstile should not be treated as a passive barrier. It is a traffic signal for the fan and the steward.
Open time
Too short creates repeated attempts, confusion and assisted passage. Too long slows the lane and increases tailgating risk.
Feedback screen
Clear messages such as accepted, wrong gate, already used, invalid event or seek steward reduce argument time and keep the lane moving.
Alarm logic
Alarms should distinguish forced passage, tailgating, wrong direction and repeated failed attempts. Not every issue needs the same steward response.
Accessibility
PWD, VIP, family and assisted-entry routes need different cycle assumptions and should not be judged only against general admission speed.
Cycle time must include the physical action of the turnstile and the human passage, not only server validation time.
Live control model
What the command centre should see
A good access dashboard should answer four questions: where is pressure forming, why is it forming, what happens if nothing changes, and what intervention is most likely to help?
Composite risk score. The weights should be calibrated by venue experience, but the structure forces the dashboard to combine pressure, success rate, wrong-gate behaviour, density and exception load.
| Dashboard signal | Meaning | Decision it supports |
|---|---|---|
| Pressure ratio | Demand versus effective capacity | Open lanes, reroute, add handhelds |
| Time-to-clear | How long the backlog will last | Escalate if clearance goes beyond kickoff |
| Wrong-gate rate | Gate assignment or communication problem | Push messages, move stewards upstream |
| Manual override rate | Rules, devices or ticket types causing exceptions | Create exception desk, adjust validation rules |
| Density zone | Spatial crowd pressure | Change barricades, open spillover areas |
The dashboard should not only say what happened. It should indicate what will happen next if operations do nothing.
Intervention playbook
How to act before pressure becomes unsafe
Interventions should be pre-agreed before the event. During pressure, teams do not have time to debate authority, permissions or messaging. The model should recommend actions from a prepared playbook.
| Signal | Likely cause | Recommended action | Owner |
|---|---|---|---|
| High arrivals, high success | Pure volume surge | Open reserve lanes, deploy handheld pre-validation | Access control lead |
| High wrong-gate rate | Communication or signage failure | Push gate reminders, move stewards to outer ring | Ticketing + stewarding |
| Low readiness factor | Fans reaching scanners without tickets open | Create ticket preparation area and broadcast instructions | Fan support team |
| High exception rate | Duplicate, comp, accreditation or rule issue | Move exceptions to separate desk away from queue | Ticketing supervisor |
| Density pocket | Bad geometry or blocked route | Adjust barriers, widen queue, remove crossing flow | Security commander |
Before kickoff
Interventions should be judged by whether they reduce pressure within the next 5 to 10 minutes.
After kickoff
The priority shifts from speed alone to safe clearance, accurate reporting and controlled late entry.
African stadium realities
The model must respect local event behaviour
Many imported stadium models assume highly predictable transport patterns, early arrival, strong compliance with signage and uniform digital readiness. African stadium operations need a model that reflects how fans actually arrive and behave.
Late clustered arrival
Traffic, workday schedules, public transport, pre-match social activity and confidence that gates will remain open can concentrate arrivals very close to kickoff.
Mixed ticket formats
Fans may present WhatsApp tickets, PDFs, screenshots, printed tickets, accreditation badges, wallet passes or box-office receipts.
Protocol pressure
VIP, VVIP, sponsors, media, teams, officials and complimentary-ticket holders can create competing flows if zoning is weak.
Mobile money effects
Payment confirmation, ticket retrieval, phone battery and customer-support issues can create queues outside the official access-control line.
The lesson is not that local conditions make modelling impossible. The lesson is that the model must be calibrated locally. Every event should update the assumptions for arrival curve, wrong-gate probability, readiness factor, exception rate, lane cycle time and steward response time.
Implementation roadmap
How a stadium can adopt this practically
A stadium does not need a full simulation platform on day one. It can build toward predictive access control in stages.
| Stage | Build | Output | Minimum data needed |
|---|---|---|---|
| 1. Measure | Gate-by-gate scan logs, rejection reasons, device uptime, active lane count | Truthful post-event report | Access logs + device logs |
| 2. Forecast | Arrival curve, gate allocation, peak demand estimate | Pre-event access plan | Ticket sales + historic arrivals |
| 3. Stress test | Late-arrival scenario, queue growth, lane requirement | Capacity risk assessment | Throughput assumptions |
| 4. Simulate | Boids agents, route geometry, barricades, rerouting | Spatial pressure map | Venue map + behavioural assumptions |
| 5. Control | Live pressure dashboard and intervention playbook | Real-time operational decisions | Live scans + steward observations |
| 6. Learn | Compare forecast against actuals | Improved next-event model | Post-event review |
Data requirements
What should be captured at every event
The quality of the model depends on the quality of the event data. The goal is not to collect data for reporting decoration. The goal is to make the next event safer and easier to manage.
Ticketing data
Sales by gate, category, channel, issue time, ticket type, complimentary allocation, accreditation group and ticket delivery method.
Access-control data
Scan time, gate, lane, device, outcome, rejection reason, retry count, admission confirmation, override status and offline validation status.
Hardware data
Turnstile open time, reset time, forced passage alarms, tailgating alarms, device uptime, battery, network latency and scanner health.
Operations data
Active lanes, steward count, barricade layout, weather, kickoff time, road closures, announcements and intervention timestamps.
KPIs
The access-control metrics that matter most
The best post-event report should help the venue understand whether the access plan worked, where pressure formed, why it formed and what should change next time.
| KPI | Definition | Why it matters |
|---|---|---|
| Peak arrival rate | Highest arrivals per minute by gate | Shows the true stress point |
| Effective throughput | Clean admissions per minute adjusted for success and uptime | Shows real capacity |
| Scan-to-admit success | Accepted admissions divided by total attempts | Shows quality of flow |
| Wrong-gate rate | Wrong-gate attempts divided by total attempts | Shows communication and zoning quality |
| Average cycle time | Presentation to lane ready for next person | Shows speed of the access lane |
| Time-to-clear | Estimated clearance time for existing queue | Supports live decisions |
| Intervention response time | Time from signal to action | Shows command-centre effectiveness |
Limitations
What the model should not pretend to know
A credible model must state its limits. Boids and formulas can help the team reason better, but they do not replace trained stewards, safety officers, police, emergency planning, venue design or professional judgement.
Behaviour is probabilistic
People do not move like machines. Crowd mood, weather, transport disruption, team rivalry, misinformation and panic can change behaviour quickly.
Density estimates need validation
Camera-based or manual density estimates should be checked against field reports, not treated as perfect truth.
Ethics and privacy matter
Any camera, AI or analytics layer should be designed around safety, minimisation, access control and clear governance.
Simulation is only useful if acted upon
A model that predicts pressure but has no intervention authority is only a report. The command structure must be ready to act.
The point of the framework is not to automate crowd safety. It is to give trained operators better early warning, better explanations and better options.
Closing position
From ticket validation to access intelligence
The future of stadium access control is not simply faster scanning. It is an integrated access-intelligence layer that combines ticketing, turnstiles, handheld scanners, gate operations, crowd modelling and post-event learning.
Boids gives stadium teams a way to think about how individuals become a crowd. Throughput formulas give teams a way to understand whether gates can cope. Live access data shows what is happening now. The learning loop turns every event into better planning for the next one.
- Predictarrival pressure
- Measurescan-to-admit reality
- Learnfrom every event
References
This paper combines Loft Technologies operational learning with established literature and guidance on boids, sports ground safety, stadium security and pedestrian movement.
Sources used to ground the framework
- 1Reynolds, C. W. (1987). Flocks, Herds, and Schools: A Distributed Behavioral Model. ACM SIGGRAPH Computer Graphics, 21(4), 25-34. DOI: 10.1145/37402.37406. dl.acm.org ↗
- 2Sports Grounds Safety Authority. Guide to Safety at Sports Grounds, Sixth Edition, also known as the Green Guide. sgsa.org.uk ↗
- 3FIFA. Stadium Safety and Security Regulations. digitalhub.fifa.com ↗
- 4Confederation of African Football. CAF Safety and Security Regulations. cafonline.com ↗
- 5Confederation of African Football. CAF Safety and Security Officers Handbook. cafonline.com ↗
- 6Fruin, J. J. (1971). Pedestrian Planning and Design. Metropolitan Association of Urban Designers and Environmental Planners.
- 7Still, G. K. (2014). Introduction to Crowd Science. CRC Press.
- 8UEFA. UEFA Safety and Security Regulations. documents.uefa.com ↗