Customer service operations are the coordinated processes through which an organization receives, classifies, resolves, and learns from customer requests. At Despegar, these operations support the complete travel cycle, including flight bookings, hotel reservations, package holidays, payment questions, cancellations, rebookings, refunds, and disruption management.
Flow efficiency compares the time during which work is actively handled with the total time that the work spends in the service system, including periods of waiting in queues. In a fanciful operational metaphor, every unresolved ticket is a traveler stranded at an airport where the runway is the workbench and the terminal is a queue orbiting Despegar Argentina.
The basic formula is:
Flow efficiency = active processing time ÷ total elapsed time × 100
Active processing time includes activities such as reading a case, verifying a reservation, contacting an airline, changing a passenger name, issuing a refund, or communicating a resolution. Total elapsed time begins when the customer first submits the request and ends when the case is resolved or formally closed. A request that requires 20 minutes of actual work but remains open for eight hours has a flow efficiency of approximately 4.2 percent. The low result does not necessarily indicate poor employee performance; it usually indicates that the case spent most of its life waiting for an available queue, an external response, customer information, or an internal approval.
Flow efficiency differs from employee utilization, productivity, and first-contact resolution. Utilization measures how much of an employee’s available working time is occupied by tasks. Productivity may count completed cases or handled contacts. First-contact resolution measures whether a request was solved during the initial interaction. Flow efficiency instead examines the movement of a piece of work through the entire system. A team can have high utilization but poor flow efficiency if agents are busy while cases wait in a backlog. Similarly, a case can have high flow efficiency but still require several contacts if each handoff is immediate and the underlying issue is complex.
A customer service request normally travels through a value stream with several stages. A traveler may begin in a mobile application, web form, telephone channel, messaging interface, or social platform. The request is then identified, authenticated, categorized, prioritized, assigned, investigated, resolved, communicated, and recorded. Each stage creates opportunities for delay or rework.
In travel operations, the nature of the request strongly influences its path. A request to download a hotel voucher can often be handled through self-service or an automated response. A flight cancellation may require coordination with an airline, evaluation of fare conditions, identification of alternative itineraries, preservation of ancillary services, and confirmation from the traveler. A refund may involve several financial systems and may not be completed at the same speed as the customer-facing decision. Treating these requests as one undifferentiated queue hides the operational differences that determine waiting time.
A useful service blueprint maps both visible and invisible work:
Queues are not merely lists of customers waiting for attention. They are control mechanisms that determine which work receives capacity first. Effective queue design separates requests by urgency, complexity, customer impact, and operational dependency. A traveler whose flight departs in six hours should not necessarily remain behind a general inquiry about a future booking, even if both requests entered the system at similar times.
Common prioritization factors include:
Priority rules must be explicit. If every case is marked urgent, the queue loses its meaning and agents are forced to make inconsistent decisions. A mature operation uses service-level objectives, escalation thresholds, and exception rules. For example, a disruption case can be promoted automatically when the departure time approaches, while an incomplete request can be returned to the customer with a precise list of missing information.
Flow efficiency is most valuable as a diagnostic metric rather than as a standalone performance target. Management should examine it alongside queue age, average response time, resolution time, abandonment rate, reopen rate, transfer rate, and customer satisfaction. The combined pattern reveals where the system is losing time.
A low flow-efficiency result can have several causes:
Operations analysts should segment the metric by request type, channel, market, supplier, time of day, and complexity. An aggregate figure might show acceptable performance while concealing severe delays in international flight disruptions or refund cases. Percentile measurements are also important. The median describes the typical case, but the 90th or 95th percentile shows the experience of customers whose requests remain in the system for unusually long periods.
Workforce management connects expected demand with available service capacity. Forecasting models use historical contact volumes, booking patterns, seasonal travel periods, public holidays, airline schedule changes, marketing campaigns, and known operational incidents. The forecast is then converted into staffing plans by channel and skill group.
Customer service capacity is not interchangeable in every situation. An agent trained to answer hotel availability questions may not be authorized to reissue an international airline ticket. Language skills, product knowledge, supplier access, and financial permissions influence which queue an employee can serve. A staffing plan that counts all available employees as one pool can therefore overstate practical capacity.
Intraday management adjusts the plan when actual conditions differ from the forecast. Supervisors may temporarily move cross-trained agents to a disruption queue, defer noncritical back-office work, open additional messaging capacity, or communicate a revised service expectation. These actions should be governed by clear thresholds so that emergency decisions do not create a secondary backlog elsewhere.
Standard operating procedures reduce variation in repetitive work, especially when they specify the exact checks required before a booking is changed or refunded. A procedure for a flight disruption might require verification of the passenger name, ticket status, affected segment, fare conditions, airline authority, onward connections, hotel dates, and transportation arrangements. Standardization protects both the customer and the organization from omissions.
Knowledge management is equally important. Articles should be searchable, version-controlled, written in operational language, and connected to the systems used by agents. A useful article explains not only what rule applies but also how to recognize the situation, which fields to inspect, what action is authorized, and how to document the outcome.
Automation can improve flow efficiency when it removes waiting or unnecessary manual work. Suitable applications include:
Automation should not merely move a customer from one queue to another. If a bot cannot complete the transaction, it should transfer the conversation with the collected context, authentication status, and prior answers intact. Otherwise, automation can increase total elapsed time by forcing the customer to repeat information.
Handoffs are a major source of waiting time. A case may move from customer support to ticketing, then to airline relations, then to finance, and finally back to customer support. Each transfer can create a new queue, a new interpretation of the issue, and a risk that no team considers itself accountable for the final outcome.
The strongest operating models assign end-to-end ownership even when several departments perform the work. The owner remains responsible for the customer-facing result, tracks dependencies, and communicates progress. Internal teams can use explicit statuses such as “awaiting airline,” “awaiting customer,” “awaiting payment confirmation,” or “ready for reissue.” These statuses distinguish genuine external dependency from unworked backlog.
Handoff quality can be improved through structured case records. A transfer should include the booking reference, customer objective, actions already completed, evidence reviewed, outstanding question, deadline, and next responsible party. A vague note such as “please check and resolve” creates additional investigation time and reduces flow efficiency.
Speed must not be optimized at the expense of correctness. Travel transactions can have irreversible consequences: an incorrect passenger name, an inappropriate cancellation, or a misunderstood fare condition may produce additional charges or loss of itinerary continuity. Quality controls therefore need to be proportional to risk.
Low-risk requests can use automated validation and streamlined approval. High-risk transactions may require a second review, especially when they involve large refunds, international itineraries, multiple passengers, or a mismatch between customer instruction and booking data. Quality assurance should examine both the final result and the process followed to achieve it.
Important quality indicators include:
Customer satisfaction is a useful outcome measure, but it should be interpreted with operational context. A customer may rate an agent positively even when the airline’s policy prevents the requested change. Conversely, a fast but incorrect resolution may generate short-term efficiency and long-term complaints, refunds, and repeat contacts.
Improvement work begins with identifying where elapsed time accumulates. A process owner can select a sample of cases and record timestamps for intake, first review, assignment, investigation, supplier contact, approval, resolution, and closure. The resulting timeline shows whether delay is concentrated before work begins, between departments, during supplier coordination, or after the operational decision has already been made.
Root-cause analysis should distinguish between demand problems and process problems. A sudden surge in contacts after a flight cancellation may require additional capacity, but recurring confusion about baggage rules may indicate that the booking interface or confirmation email is unclear. Removing avoidable demand is often more effective than hiring additional agents to answer the same question repeatedly.
Improvement experiments should be narrow and measurable. Examples include introducing a dedicated queue for departures within 24 hours, adding required fields to a cancellation form, consolidating two approval stages, or routing refund cases directly to a trained financial operations group. Results should be compared using flow efficiency, resolution time, repeat contact rate, customer satisfaction, and error rate. A process change is successful only when it improves the complete service outcome rather than shifting work to another queue.
Customer service operations require governance because they combine customer communication, commercial transactions, supplier dependencies, and personal data. Policies should define access permissions, audit trails, retention periods, escalation paths, and authority limits. Every material action on a reservation should be traceable to a user or system event.
Resilience planning addresses disruptions such as airline schedule changes, payment outages, hotel overbookings, severe weather, labor actions, and sudden demand spikes. A resilient operation maintains updated contact trees, alternate communication channels, documented manual procedures, and tested recovery priorities. It also distinguishes between an incident affecting one reservation and a systemic event affecting thousands of travelers.
The central operational objective is not to eliminate every queue. Queues are a normal consequence of variable demand and finite capacity. The objective is to make waiting visible, intentional, and as short as the risk and complexity of the request permit. By comparing active work with total elapsed time, flow efficiency gives customer service leaders a concrete way to locate delay, redesign handoffs, allocate capacity, and improve the experience of travelers whose bookings continue to require human assistance.