Cutting new-customer activation from 3 days to 20 minutes
A growing GPON operator replaced manual OLT provisioning with NetXol NMM auto-provisioning and TR-069 ACS — collapsing activation time and field errors and freeing the NOC to focus on growth.
−98%
Activation time
−92%
Provisioning errors
+11%
ARPU uplift
A privately-held FTTH operator serving a mixed urban-suburban region in South Asia, with 80,000 active subscribers across four cities, had grown faster than its operations could keep up with. New-subscriber activation — the single most repeated workflow in the business — was queue-bound and taking up to three days. Every additional sale was adding to the queue. The operations leadership knew that scaling staff was not the answer.
Background
The operator runs a mixed-vendor PON estate: roughly 60% Huawei OLTs, 30% CDATA, and the remaining 10% V-SOL in newer territories. The ACS was a 2018-era commercial product, used essentially for firmware pushes and WiFi credential synchronisation. New activations were a manual ritual: an order comes in via the CRM, a NOC engineer SSHes into the right OLT, creates the ONU, attaches a service profile, configures the CPE, validates by ping.
Every step was a place to make an error. By the operator's own admission, about 7% of activations had at least one defect in the first week — wrong package speed, wrong VLAN, sometimes the wrong subscriber on the wrong port. Those defects converted directly into customer-care calls, second activations, and a small but persistent stream of avoidable truck rolls.
The challenge in numbers
| Metric | Before | After | Δ |
|---|---|---|---|
| Engineer-minutes per activation | 22 min | 0 (only exceptions) | −100% nominal |
| Wall-clock activation time | ~3 days (queue) | ~20 min | −99% |
| First-week defect rate | 7% | 0.6% | −92% |
| Activations per NOC engineer / day | 12 | 95+ | +690% |
What we did
Implementation phases
- Connected NetXol NMM to all four OLT vendor estates with native adapters (Huawei, CDATA, V-SOL).
- Imported the existing service profiles into NetXol's vendor-agnostic profile model.
- Validated discovered topology against the operator's ground-truth maps; reconciled 142 mismatches.
- Provisioned NetXol ACS in parallel with the incumbent for 7 days under split traffic.
- Migrated CPE TR-069 management URL to NetXol via firmware-side rotation; rollback path tested at 10% fleet.
- Validated zero-touch first-time configuration for 5 representative CPE models across vendors.
- Wired the CRM order intake to NetXol's order API. Every paid order now triggers a single workflow.
- Defined exception escalation — any failure surfaces with diagnostic evidence to a single queue.
- Ran 4,800 production activations under the new flow with engineers on standby; 23 exceptions escalated.
- Engineers stood down from the provisioning queue. The queue is now < 1 active item at any time.
- AI OLT Engineer enabled for natural-language bandwidth changes ("upgrade 485754 from 20 to 50 Mbps").
- Activation NPS surveys auto-triggered 24h post-activation; first-month NPS rose 14 points.
Anatomy of an activation, after
- 1CRM order created with subscriber, plan, address and assigned CPE serial number.
- 2NetXol receives the order, reserves an ONU index on the right OLT, writes the whitelist entry.
- 3CPE arrives at the address, connects to the OLT, informs the ACS within 60 seconds.
- 4ACS pushes the full configuration in one transaction: VLAN, QoS, WiFi SSID + PSK, firmware if needed.
- 5Synthetic test measures speed against the contracted plan and writes the result back into the CRM.
- 6Order auto-closes; subscriber receives "you are online" notification by SMS.
Where it almost did not work
Two weeks into Phase 2, a batch of 1,400 CPEs running an older firmware build went into an inform-loop after the ACS migration. The root cause was a vendor quirk in handling the connect-request port renegotiation. NetXol's ACS supports automatic detection of inform-loop patterns and quarantines affected CPEs, which limited the customer impact to under 0.3% of the fleet for under 90 minutes. The vendor firmware was patched the following week; the affected CPEs were auto-upgraded out of the quarantine.
The lesson the operator carried forward
Always run the ACS swap behind an inform-loop detector and a quarantine policy. The 1,400-CPE incident would have been an outage on the previous platform; on NetXol it was a non-event for 99.7% of the fleet.
Outcomes
−98%
Activation wall-clock
3 days → 20 minutes
−92%
First-week defect rate
7% → 0.6%
+11%
ARPU uplift
Through accurate package enforcement + upsell flow
+14
NPS lift in first 30 days
From 28 to 42
Tech stack used
- NetXol NMM — multi-vendor OLT/ONU management (Huawei, CDATA, V-SOL).
- NetXol ACS — TR-069 with TR-369 ready, with idempotent profile pushes and quarantine policy.
- NetXol AI OLT Engineer — natural-language ONU operations.
- CRM integration via TM Forum-compatible order API.
- Synthetic test layer using HTTP/UDP probes from the CPE.
What we would do differently next time
- Start the firmware staging campaign one week earlier — outdated firmware was the single largest source of exceptions.
- Move the topology reconciliation into Phase 1 day 1, not day 12. The 142 mismatches included some that affected provisioning paths.
- Onboard finance/billing earlier — invoice reconciliation needs the activation IDs from the new flow.
Put your ISP on autopilot
See NetXol on your own network in a live demo — or send us your RFP and let our team scope the whole project for you.