How to Design, Prioritize, and Measure a Telesales Call Queue for Thai Online Shops

A Concrete Starting Point: Malee Home Décor's Monday Morning Problem

Imagine Malee Home Décor, a fictional Thai online shop with three representatives: Fah, Som and Nop. Their Monday list mixes agreed callbacks, an unresolved delivery complaint, buyers with a longer-than-usual purchase gap, a person who declined further sales contact and records already handled by another colleague. Without a shared record, the team can miss an appointment or repeat an unwanted call even when every representative is working hard. The example is invented; no customer records are used.

This guide turns that mixed list into a practical queue with handling rules, owners, due times, a capacity calculation and honest measurement. CasperBrain supports analysis and planning; CasperCall is a separate execution product. The workflow below is an operating example, not a promise that either product automatically enforces every rule or field.

RFM Segmentation Foundations and Queue Eligibility Rules

CasperBrain uses four quartiles for recency, frequency and monetary value: 1 is the strongest purchasing rank and 4 the weakest. Read a code such as 114 as three labels. A 111 customer belongs to the strongest quartile on all three dimensions, not necessarily the single best customer. A 444 record is in the weakest quartile on each dimension relative to the scored population and window. Monetary value means total spending within the scope, not profit or average basket value. Vendor scales may reverse the direction or use five levels, so check the convention before copying a segment label.

Eligibility comes before ranking. Exclude further sales outreach when the recorded preference refuses it, and hold ambiguous identities for review. Route an unresolved complaint to the authorized service owner before any sales decision. Then check appointments and recent contact. A fictional 48-hour cooldown below applies to unsolicited sales contacts by any representative; it must not erase a specifically requested callback or necessary service follow-up. Document the purpose and permitted route instead of treating all contact as interchangeable.

Queue Priority Logic: Eight Fictional Record Decision Table

The following eight records are fictional teaching examples. Codes are supplied example quartile labels, not scores calculated from this eight-record queue. Assume the underlying scoring population puts the recent first-time buyer 144 in low frequency and monetary quartiles. The Decha 211 example combines a less-recent rank with strong frequency and total-spending ranks; the separate 45-day elapsed interval provides the reason to investigate timing.

The first column is a handling lane, not the chronological order of calls. Due times govern appointments, while service and identity work may be handled in parallel by the appropriate owner. Held or excluded records are not numbered sales priorities. Before acting, apply the eligibility gates to the whole record: a customer with both a 10 a.m. appointment and an open complaint needs a service conversation, not an automatic upsell.

Capacity Planning: A Worked Arithmetic Example

Before filling the queue, the team needs to know how many calls are actually achievable. The following calculation uses Malee Home Décor's hypothetical planning inputs. It is a teaching example, not a benchmark or industry standard. Teams should derive their own inputs from their own call logs.

Fictional inputs: 3 representatives × 4 usable hours × 60 minutes ÷ 8 minutes per slot=90 slots. The 8 minutes must cover preparation, conversation or attempt, and recording the result; it is an assumed planning input, not a measured benchmark. Reserve 18 slots for due or overdue callbacks, leaving 72. Assign 40 additional records and retain 32 unassigned slots for longer conversations, coordination and urgent work. Schedule appointments at their promised times rather than filling every available minute with an interchangeable call.

The buffer is 72 − 40=32 slots, or 32 ÷ 72 ≈ 44.4% of the remaining capacity. Long calls, wrong numbers and service handoffs make the estimate uncertain. Assigning all 72 raises the risk that later records remain unworked; it does not prove that this will happen. Compare planned and actual time, then adjust the next day’s estimate. If service cases take much longer than sales check-ins, give them a separate time allowance.

Queue Decision Diagram: Eligibility Flowchart

Apply these gates to each record before a sales call. The sequence separates eligibility, service work, commitments and discretionary outreach. It is the fictional team’s process. Microsoft’s cited queue and assignment documentation illustrates prioritisation and capacity concepts in its own systems; it does not prescribe this outbound hierarchy or establish CasperBrain behavior.

First check the recorded permission and contact preference for the proposed purpose; a sales refusal blocks sales outreach. Hold uncertain identities. Next route any unresolved complaint to the authorized service owner, including records with an existing sales appointment. For otherwise eligible records, honor a specifically agreed callback and repair missed commitments using the agreed retry policy. Apply the 48-hour recent-contact hold only to unsolicited sales outreach, checking all representatives’ logs. Use RFM and purchase context to rank only the remaining discretionary work.

Sources: Dynamics 365: Assignment Methods in Unified Routing · Dynamics 365: Create and Manage Queues for Unified Routing

Pre-Call Preparation and Customer Discovery

Read the activity record before dialing: purchase context, previous conversation, preferences, agreed next action and open service issues. The HubSpot discovery-call source supports preparing for relevant questions and listening to the customer’s situation. It does not establish a universal script or guaranteed sales improvement. A shared spreadsheet can support the discipline when ownership and updates are clear; a tool alone cannot make missing notes reliable.

Decode the example before choosing the question. For 114, strong recency and frequency with lower total spending does not by itself tell you basket size, profit or appetite for a volume offer. Inspect actual order details. For the fictional 144 new buyer, first ask about the purchase experience; do not assume replenishment is due after seven days. For 211 Decha, the elapsed interval invites a question about remaining stock, not a claim that the customer has disengaged.

Pre-call checklist: read the latest notes; verify the identity and permitted purpose; check the due time and owner; inspect any service issue; decide the one useful question. Keep the planning allowance in mind without rushing a customer or promising an unauthorized remedy. Record the outcome before moving to the next record so colleagues can see what was agreed.

Sources: HubSpot: Discovery Call Questions That Reveal Real Pain Points

Call Handling: Scripts, Outcomes, and Owner Assignment

Scripts in telesales are best treated as conversation guides rather than rigid word-for-word recitations. The goal is consistency of intent—what the rep is trying to learn or offer—not uniformity of phrasing. The following script outlines cover the four most common call types in Malee Home Décor's queue.

Agreed Callback (Priority 1 — example: Khun Araya): 'สวัสดีค่ะ / ครับ คุณอารยา ดิฉัน / ผม [ชื่อ] จาก Malee Home Décor โทรมาตามที่นัดไว้นะคะ / ครับ วันนี้สะดวกคุยไหมคะ / ครับ?' (Hello, Khun Araya, this is [name] from Malee Home Décor calling as arranged. Is now still a good time?) Open by honouring the appointment. Do not pivot to selling before confirming the customer's objective for the call. If the customer has changed their mind about the call, thank them and log the updated preference—do not push.

Service Issue (service lane — example: Khun Chanida): 'สวัสดีค่ะ / ครับ คุณชนิดา ดิฉัน / ผม [ชื่อ] จาก Malee Home Décor ทราบว่ามีปัญหาเรื่องสินค้าที่ได้รับ ขอโทษมากนะคะ / ครับ รบกวนเล่าให้ฟังได้เลยค่ะ / ครับ' (Hello Khun Chanida, I understand there was a problem with your order. I'm sorry to hear that. Please tell me what happened.) Listen fully, then explain the handoff or remedy the service owner has actually authorized. Do not attempt any upsell on a service call. If the issue requires a specialist (e.g., logistics or accounts), hand off immediately per the service protocol in the next section.

Timing follow-up—Khun Decha: ‘สวัสดีค่ะ / ครับ โทรมาสอบถามว่าสินค้าที่ซื้อไปยังมีใช้อยู่ไหม และสะดวกให้ติดต่ออีกครั้งเมื่อไรคะ / ครับ’ (Hello, may I ask whether you still have the product and when another conversation would be useful?) Ask whether now is convenient. Record the answer and a requested date; do not turn the 45-day gap into a pressure tactic or an unsupported claim that the customer needs more stock.

New-buyer follow-up—Khun Emonrat: if the purpose and contact route are appropriate, ask one open question about the order experience. Let the customer describe any problem. Record what they actually say and any next step, without converting the representative’s impression into a customer-satisfaction score or a promise of future demand.

Outcome logging is as important as the call itself. Every call should end with the rep recording one of five outcomes in the activity record: Reached and completed (note key points), Reached and rescheduled (log new date/time), Voicemail left, No answer (note attempt count), or Customer declined / requested no further contact (immediately escalate to Do Not Call process). Logging every attempt—including no-answers—ensures the next rep has accurate context and prevents duplicate or conflicting outreach.

Sources: HubSpot: Discovery Call Questions That Reveal Real Pain Points

Service Handoff Protocol

A telesales representative is not a service specialist. When a call reveals an unresolved complaint, a logistics failure, a billing dispute, or a product defect, the rep's role is to acknowledge, apologise, and hand off—not to improvise a solution. A clear handoff protocol prevents customers from being bounced between teams without resolution.

A practical handoff has four parts: acknowledge the issue; connect to the authorized service owner or create a case; tell the customer the next confirmed step; record the owner and due time. Promise a resolution deadline only after the service owner confirms it. Otherwise agree on a specific status-update time. Keep the case visible and require the team to check it before further sales outreach. A flag helps coordination; it does not guarantee that another system or representative will enforce the hold.

The record should contain a customer alias, relevant order ID, a concise factual description, destination owner and the next confirmed update time. Check whether the handoff was accepted. If the owner is absent, a supervisor assigns cover and passes the same context rather than asking the customer to explain everything again.

Metrics, Measurement, and Causal Caveats

Use a separate fictional 40-customer assigned group, fixed at Monday 09:00. Observe all 40 through Wednesday 09:00—the same 48-hour window, whether reached or not. During the window 30 distinct assigned customers receive an attempt, 18 have a live conversation, and six distinct customers place at least one valid order. Assume all six purchasers are among the 18 reached and none among the other 22. These are supplied teaching counts, not causal results. Multiple orders from one person count as one purchaser.

Coverage: 30 distinct customers attempted ÷ 40 assigned=75%. It measures how much of the assigned group received an attempt. Repeated calls to one customer do not increase the numerator. Investigate time estimates, availability and unexpected work before judging the team from one rate.

Contact rate: 18 distinct customers reached ÷ 30 attempted=60%. Twelve attempted customers were not reached. Track wrong numbers, no-answer and other outcomes separately if the distinction helps the next decision; the aggregate rate alone does not explain why contact failed.

Observed purchaser share of assigned: 6 ÷ 40=15%. This uses every originally assigned customer as its denominator and the fixed 48-hour window. Record valid-order rules before calculating it. A cancelled transaction or multiple line items must not silently create another purchaser.

Observed purchaser share of reached: 6 ÷ 18 ≈ 33.33%, using the stated assumption that all six purchasers are in the reached group. This describes a selected subgroup. If purchasers also appeared among unreached or unattempted customers, count each subgroup separately rather than reusing the same numerator.

Neither rate establishes incremental sales caused by calls. Some buyers could have ordered anyway. One stronger design is to randomly assign otherwise eligible customers prospectively to the contact program or a holdout, use the same observation window and compare by original assignment. Do not withhold promised callbacks or necessary service to create a test. Account for uncertainty and possible contact through other channels. Merely selecting similar-looking customers does not remove selection bias.

Additionally: percentages and percentage points are different. If the purchase rate among reached customers rises from 33% to 38% between two weeks, that is a 5 percentage-point increase, or approximately a 15% relative increase. Mixing these framings inflates or deflates apparent results.

Metrics Reference Table

The table below consolidates the measurement framework so teams can reproduce it with their own numbers each week. Fill in your actual figures in the 'Your Week' column.

Weekly Iteration and Team Review Process

A call queue improves only when the team reflects on what happened and adjusts. Malee Home Décor's fictional team holds a 30-minute review every Friday. The agenda has five fixed items, each mapped to a specific metric or qualitative observation.

Review five things: coverage and contact with their denominators; accepted service handoffs and promised update times; any contact that should have been excluded; missed commitments; and purchase-context changes after new orders are imported and rescored. Any 75%, 60% or 40% threshold is an illustrative choice for this fictional team, not an industry target. A score can change because of elapsed time, new history or a changed scoring population; the new code alone does not establish urgency.

The output of the Friday review is a short written summary: coverage rate, contact rate, observed purchase rate (with the causal caveat noted), any compliance failures, and the adjusted capacity assumption for the following week. This document is the team's institutional memory and the basis for any reporting to management.

If a segment is difficult to reach, inspect the time, purpose, route and record quality. Test a small change only where communication preferences permit it. Trying a LINE message before a call is an optional hypothesis, not a way to bypass a refusal. Write the expected observation and comparison window before looking at the result.

Limitations, Ethical Boundaries, and Troubleshooting

No queue design eliminates uncertainty, and several limitations apply to everything described in this guide. First, RFM segmentation is a descriptive tool. It summarises past behaviour; it does not predict individual future behaviour with certainty. A customer’s refusal of further contact takes precedence over any RFM score, including '111'. Second, the capacity arithmetic presented here depends on a single estimated average call duration. If some call types (e.g., service escalations) routinely take 20 minutes while new-buyer follow-ups take 4, a single average will either underestimate or overestimate usable capacity. Teams should track actual call durations by type and use type-specific estimates.

Missing notes can create repeat contact, so a supervisor can inspect a small sample of records for completeness. Measurement also depends on the window: 48 hours counts purchases observed during that period; it does not attribute them causally to a call. Extending the same cohort to seven days can include additional purchases, or the count may stay unchanged. Declare the horizon and use equivalent windows for comparisons.

Troubleshooting: if a colleague already contacted the person, check the existing commitment and apologize for an accidental duplicate; changing the cooldown alone cannot repair a missing log. If a complaint appears during a sales call, stop selling and hand off to service. If an agreed callback is unanswered, log the outcome, follow the agreed retry policy and release unused time. If two profiles share a phone, hold uncertain contact and investigate rather than assuming they are the same person.

CasperBrain provides analysis and planning; CasperCall is a separate execution product. Verify the tools and operating procedures your team actually uses. A priority label, service hold or preferred-contact field described here should not be assumed to trigger an automatic action in either product.

Sources and editorial method

Examples are fictional and contain no customer records. Prepared by the CasperBrain team with AI drafting assistance; facts and language are checked before publication.

• Dynamics 365: Assignment Methods in Unified Routing
• Dynamics 365: Create and Manage Queues for Unified Routing
• HubSpot: Discovery Call Questions That Reveal Real Pain Points