Customer Not Available — A Business Continuity Plan for Last-Mile, Collections and Healthcare Operations in India 2026
A control tower operator at a 3PL in Gurgaon pulled up her morning exceptions dashboard on a Wednesday. 6,200 shipments flagged. Roughly 4,100 of them sat under one disposition: "customer not available." A separate dashboard in the collections division at a Mumbai NBFC showed 11,800 EMI reminders dialled the night before with a 41% "customer not available" rate. A third dashboard, at a Bengaluru diagnostic lab, showed 1,400 home blood-collection appointments where the customer hadn't answered the phlebotomist's call before arrival. Same disposition, three different operations, one underlying problem: the customer wasn't on the other end when the system tried to reach them, and nobody had a real plan beyond "try again later."
This is the question buyers Google when they type "customer not available business continuity plan." It is not a feature ask. It is an operational problem — recurring, expensive, and unsolved across categories — where voice AI offers a different shape of answer than what most playbooks deploy.
This post is the operator's view of a customer-not-available business continuity plan that holds up across Indian last-mile, collections, healthcare and field-service operations. The retry policy that actually works. The alternate-contact escalation tree. The hold-at-hub and reschedule loops. The voice AI orchestration that converts "not available" from a terminal disposition into a routable signal.
Why "customer not available" is the most expensive disposition you have
In Indian operations, the cost of a "customer not available" event compounds in ways most ops leads don't model.
Logistics. A failed delivery (NDR — non-delivery report) where the customer wasn't reachable triggers a re-attempt the next day at ~₹40–80 per re-attempt depending on metro vs tier-3. After two re-attempts, the package goes back to the hub. RTO (return to origin) costs the seller ₹120–280 per shipment plus the lost revenue. A 4,100-NDR day at a mid-size 3PL is ₹3.5–9 lakh in re-attempt cost compounding into ₹15–22 lakh in RTO if unresolved.
Collections. A "not available" disposition pushes the borrower into the next attempt cycle, increasing time-in-bucket and degrading cure-rate by 6–11 percentage points per slipped attempt. On a ₹1,500-EMI book at 47,000 monthly cases, this translates to ₹4.8–8.6 crore of impaired collection per quarter.
Healthcare. A diagnostic home-collection where the phlebotomist couldn't reach the customer is a wasted trip — ~₹140–220 per visit including travel — plus a slot the lab can't reuse and a customer who often abandons re-booking.
Field service. A technician dispatched for an AMC service call where the customer wasn't reachable is a half-day write-off for the technician's productivity.
The pattern: "customer not available" isn't one event; it's the first event in a cost compounding sequence. The business continuity plan exists to short-circuit that sequence before the second cost hits.
Why the standard playbook fails
Most operations teams treat "customer not available" with a fixed retry policy: try again in 2 hours, then 4, then 24. This worked when:
- Calls were made by human agents who exercised judgement.
- The retry queue was small enough to manage manually.
- The alternate-contact list was reliable.
None of these hold in 2026.
Human agents are not making 30,000 outbound calls a day on a single book — voice AI is, and a generic retry policy applied at that volume produces dialler storms, spam-flag risk on the outbound number and customer-fatigue events. The retry queue at modern volumes is too large for a human dispatcher to make routing decisions. And the alternate-contact list captured at order placement is stale within 60 days for ~22% of records.
A business continuity plan for "customer not available" in 2026 has to do three things the standard playbook doesn't: classify the not-available reason, escalate to alternate contacts intelligently, and route to non-call resolution paths when call retries won't work.
The classification layer — what "not available" actually means
A useful business continuity plan starts by recognising that "not available" collapses at least seven distinct underlying states. The voice AI agent (or the ops dispatcher) should be classifying which one is happening.
| Underlying state | Signal | What this needs |
|---|---|---|
| Phone genuinely not reachable | Network error, no ring | Retry on alternate number, then SMS+WhatsApp |
| Phone rings out unanswered | Full ring duration, no pickup | Time-of-day retry, then alternate contact |
| Voicemail picked up | IVR / answering machine detected | No-op leave brand callback; SMS handoff |
| Truecaller-blocked / spam-flagged | Auto-reject pattern | Rotate outbound CLI, retry with verified caller-ID |
| Customer rejected the call | One ring, immediate drop | SMS+WhatsApp handoff; no immediate retry |
| Wrong number | Verified by callback | Mark in CRM, escalate to operations |
| Number changed (off-network) | Repeated reachability fail | Source new number via alternate contact or KYC update |
Each of these has a different right next step. Treating them all as "try again in 2 hours" wastes spend and risks brand-damage from over-dialing.
The voice AI platform should be writing the underlying state to the disposition, not just "not available." This is the single most important data architecture decision in this category and the one most vendors get wrong by default.
The orchestration model
[Voice AI attempt 1] │ ├─ Connected → run script └─ Not available → classify reason │ ├─ Network error → retry alt number in 30 min ├─ Rings out → retry in 4 hr different time-window ├─ Voicemail → leave brand callback; SMS+WhatsApp ├─ Spam-flagged → rotate CLI; retry within 24 hr ├─ Rejected → SMS+WhatsApp only; no more dial today ├─ Wrong number → ops queue for verification └─ Off-network → alternate contact escalation │ ┌─────────────────────┘ ▼ [Voice AI attempt 2] │ └─ Not available → escalate to alternate contact OR route to non-call resolution path │ ▼ [Alternate contact path] - Secondary number on file - Family member (collections) - Buyer's office (logistics) - Marketplace-recorded number (D2C) │ ▼ [Non-call resolution paths] - Hold at hub for self-pickup (logistics) - WhatsApp self-serve reschedule - SMS deferred-payment link (collections) - Locker / branch drop (healthcare)
The two parts most ops teams underbuild are alternate-contact escalation and non-call resolution paths. Without the former, the bot just dials the same dead number five times. Without the latter, every unresolved exception becomes a human dispatch problem.
The alternate-contact escalation tree
Every Indian operations vertical has its own alternate-contact structure. The business continuity plan should hardcode the right tree per vertical.
Logistics. Marketplaces typically pass two numbers on a shipment: the buyer's primary and a backup (often family or office). Some carry a secondary "delivery instruction" contact. After two failed primary attempts, dial the backup; after two failed backup attempts, escalate to hold-at-hub.
Collections. Indian retail lenders typically capture borrower phone + co-applicant phone + reference contact (often a family member). After two failed primary attempts, the bot should never silently dial the reference — DPDP and RBI Fair Practices Code restrict third-party contact for collections. The escalation goes to a human collections agent who decides whether the reference contact is warranted.
Healthcare. Diagnostic and pharmacy orders often have a delivery contact different from the patient — the patient's son or daughter-in-law is sometimes the operational contact. Family-phone structure is the norm. Identity verification on the first 6 seconds determines who's actually being talked to.
Field service. AMC and warranty contracts typically have the registered owner plus the residence contact plus sometimes the building security. Escalation to the residence/building contact is permitted under most service agreements.
The rule: alternate-contact escalation is verticalised. A single retry policy across operations doesn't survive contact with the consent overlay each vertical carries.
Non-call resolution paths — the underbuilt layer
The single largest leverage in the business continuity plan is to route customer-not-available cases to non-call paths that don't depend on reaching the customer at all.
Logistics: hold-at-hub for self-pickup. If a customer is genuinely unreachable on the day, hold the package at the nearest partner pickup point (Amazon Hub, Blue Dart Express centre, India Post outlet, kirana partner). Notify via WhatsApp with a one-tap location pin and pickup window. Conversion to successful delivery: 38–54% on holdable shipments. Saves ₹120–280 per RTO avoided.
Collections: deferred-payment link with hardship form. Borrower not reachable for 3 attempts; route to a WhatsApp/SMS deferred-payment link with a short "tell us what's blocking you" form that captures hardship signal. Inbound replies route to a human agent. Converts 14–22% of unreachable cases into a structured next step instead of a dead exception.
Healthcare diagnostics: locker or branch drop. Home sample collection failed because phlebotomist couldn't reach customer; offer drop-off at the nearest lab branch via WhatsApp with a one-tap location. Conversion: 22–34% on labs with branch density.
Field service: self-serve reschedule. Service appointment missed; offer WhatsApp-based self-serve reschedule with technician slots. Conversion: 31–48%, much higher than waiting for the customer to call back.
These paths convert unreachable cases into resolved cases without burning more outbound dial spend. Most ops teams haven't wired any of them.
Indian-specific realities
Time-of-day cadence by vertical. Logistics customers reachable 11am–1pm and 5pm–8pm. Hindi-belt collections borrowers don't pick up before 10:30am or after 9pm. Healthcare customers reachable mid-day weekdays, weekends concentrate 10am–noon. Generic 9am–5pm retry windows underperform by 18–32%.
Tier-2/3 number churn. ~22% of customer phone numbers in tier-2/3 cities change within 6 months due to porting, lost SIM, family-member-handover. The business continuity plan needs an alternate-contact-refresh trigger on first detection of off-network.
WhatsApp share is high. WhatsApp penetration in Indian metros is 70%+ on Android handsets. A customer-not-reachable on voice is often reachable on WhatsApp. Non-call resolution paths should default WhatsApp-first.
SMS deliverability variance. SMS deliverability in India varies 76–94% depending on operator, time and DLT template quality. SMS-only fallback as the last-resort path is not reliable; pair with WhatsApp.
Truecaller dominance. Truecaller has ~80% Android-handset penetration in India. An outbound number flagged as spam (which happens within 8–14 days of running 80,000+ weekly dials from one CLI) tanks reachability across the entire book, not just on flagged numbers. CLI rotation and Verified Business Caller registration are baseline.
What goes wrong
Generic retry policy. A single retry cadence applied across verticals burns spend on collections (where third-attempt fatigue lowers PTP capture) and underspends on logistics (where 3 quick same-day attempts work). Tune by vertical.
Silent alternate-contact dialling in collections. Some ops teams configure auto-escalation to reference contact after primary failures. This violates RBI Fair Practices Code on third-party contact and triggers regulator complaints. The reference contact path must go through a human decision, not bot automation.
Voicemail brand leak. Voicemails referencing the borrower's loan, the patient's order or the consignee's shipment leak private information into a shared family inbox. Voicemail messages should be brand-only callback with no specifics.
Spam-flag cascade. A single hot CLI flagged by Truecaller drags down reachability across the book for 7–14 days. Without CLI rotation policy, a busy 3rd-of-month dial cycle can flag the entire book.
Disposition coarse-grain. Operations team reports a flat "customer not available" rate without underlying-state breakdown. Continuous improvement is impossible. Mandate seven-state classification in disposition logs from day one.
Non-call path config drift. Hold-at-hub partner pincode coverage changes; deferred-payment link expires; locker drop-off branches close. Wire these reference data sets to the live ops system, not static config.
The numbers that matter
Realistic ranges from production deployments across Indian 3PLs, NBFCs and healthcare platforms running an integrated customer-not-available continuity plan for 90+ days.
| Metric | Acceptable | Good | Best-in-class |
|---|---|---|---|
| "Not available" classification depth | 4 states | 6 states | 7+ states |
| Disposition write latency | < 30s | < 10s | < 4s |
| Alternate-contact resolution rate | 14% | 24% | 36% |
| Hold-at-hub / locker conversion (logistics/healthcare) | 22% | 38% | 54% |
| WhatsApp self-serve reschedule conversion | 18% | 31% | 48% |
| RTO reduction (logistics) | -14% | -28% | -41% |
| Time-in-bucket lift (collections) | +6 pts | +11 pts | +17 pts |
| Privacy incident rate (voicemail leak etc) | < 0.5% | < 0.1% | 0% |
The RTO reduction in logistics and the time-in-bucket lift in collections are what get the CFO's attention. The privacy incident rate at the bottom is the hard constraint — any rate above 0% is one screenshot away from a brand event.
For broader product context, see the logistics control tower playbook, the voice AI + WhatsApp collections orchestration playbook, and the healthcare cart recovery playbook.
The 45-day implementation playbook
Days 1–7. Pull 90-day disposition data. Calculate current "customer not available" rate per vertical and cost-per-incident. Map alternate-contact data sources in the operations system.
Days 8–14. Define the seven-state classification taxonomy. Wire the disposition write-back to the operations system with the new states. Build the CLI rotation policy and register Verified Business Caller status.
Days 15–25. Build the verticalised escalation tree (logistics: backup → hold-at-hub; collections: human-decision gate; healthcare: family contact verified → locker drop). Wire non-call resolution paths (hold-at-hub partner API, deferred-payment link generator, WhatsApp self-serve reschedule template).
Days 26–35. Pilot at 10% of "customer not available" traffic on one vertical. Daily review of classification accuracy, escalation outcome and non-call path conversion. Tune the time-of-day retry windows by region.
Days 36–45. Roll to 100% on the chosen vertical. Hand over to the operations team with a daily dashboard. Plan rollout to the second vertical for the next quarter.
By day 45 the Wednesday morning exceptions dashboard at the 3PL still shows 6,200 flagged shipments, but only 1,800 sit as terminal "customer not available" exceptions — the other 2,300 have been routed to alternate contacts, hold-at-hub partners or WhatsApp self-serve reschedules. RTO drops by 28%. The control tower team handles judgement calls instead of dispatch rote.
Compliance and audit trail
RBI Fair Practices Code (collections). Third-party contact for borrowers is restricted. Bot-driven escalation to reference contacts without human-decision gate is non-compliant. Audit logs must show human approval before any non-borrower dial.
DPDP Act 2023. Each dial must operate under purpose-bound consent. Cross-vertical escalation (e.g., a customer's number used for both collections and cross-sell) requires separate consent. Voicemail messages with order/loan details may constitute disclosure to third parties — keep them brand-only.
TRAI DLT. Transactional templates for retry communication must be DLT-registered. WhatsApp self-serve reschedule templates follow Meta's category policy, separate from DLT.
RBI on outbound numbers. Some banks are formalising Verified Business Caller frameworks. Compliance with these reduces spam-flag risk and is becoming table stakes for NBFC outbound dialling.
Build vs buy
A 4-engineer team can build the seven-state classification and a verticalised retry policy in one quarter. Adding the alternate-contact escalation tree, the non-call resolution paths (hold-at-hub partner integrations, WhatsApp self-serve flows, locker drop APIs) and the CLI rotation policy is one more quarter. For multi-vertical operations (a logistics company that also runs collections), buy a platform with verticalised templates rather than building seven escalation trees.
What changes in the next 12 months
RCS adoption. Rich Communication Services adoption is climbing in India through 2026. Non-call resolution paths will increasingly default to RCS over SMS for richer interaction. Voice AI platforms that integrate RCS templates inside the customer-not-available flow will edge competitors.
Verified Business Caller across telcos. Jio, Airtel and Vodafone-Idea are rolling out sender verification frameworks similar to Truecaller's. Spam-flag risk reduces; reachability of unbranded calls reduces further. Sender verification becomes mandatory.
ABDM HealthID for identity. Healthcare verticals will move identity verification from phone-based to HealthID-based, simplifying the family-phone identity ambiguity.
Account Aggregator for collections context. AA-shared cash-flow data will let collections bots understand when "not available" likely reflects hardship (income gap, recent layoff signal) versus inconvenience. Routing to a hardship-aware path becomes more accurate.
Bottom line
"Customer not available" is the most expensive disposition in Indian operations because it compounds — failed delivery becomes RTO, missed collection becomes bucket slip, unreachable patient becomes wasted trip. The standard "try again in 2 hours" playbook doesn't survive 2026 volumes. A useful business continuity plan classifies the underlying state, escalates to alternate contacts within consent rules, and routes unreachable cases to non-call resolution paths — hold-at-hub, WhatsApp self-serve, deferred-payment link, locker drop — that resolve the case without burning more dial spend. Get this right and RTO falls 28%, collections time-in-bucket lifts 11 points, and the ops team stops drowning in exceptions.
If you run last-mile, collections, healthcare or field service in India and your "customer not available" rate is the single largest disposition in your reports, talk to us — we'll show you a seven-state classification dashboard from a live deployment, not a slide.
Frequently Asked Questions
Tags :
