Insights

Telecom writing by the people who carry the traffic

Every article here is written by a Telecorenetwork voice engineer, carrier relations manager or compliance lead — people who debug one-way audio at 03:00 and negotiate the interconnects behind 190+ countries. No vendor-neutral filler, no reheated press releases, and numbers you can check against your own call records.

Featured analysis

Why your answer rates fell in 2026 — and the four fixes that actually move them

Answer rates on outbound enterprise traffic dropped again this year, and most of the blame lands on analytics engines that never see a full attestation on your calling numbers. We pulled anonymised data from 2.4 billion minutes of Telecorenetwork-carried traffic to separate the causes that matter from the ones consultants like to sell.

Read the full analysis
Library

Browse by subject

Twelve working guides on the problems our customers actually raise — trunk sizing, attestation, porting, recording retention, queue maths and media streaming.

Sizing SIP channels from CDR data instead of guesswork

The one-channel-per-four-users rule was written for offices that no longer exist. Here is how to extract true peak concurrency from ninety days of call detail records and size a trunk against measured demand.

Read article

10DLC rejection codes: what each one actually means

Campaign registries return terse codes that rarely explain the real problem. We decode the rejections we see most often and show exactly which field in your brand or campaign submission to change.

Read article

MOS is a lagging indicator — measure these four things instead

By the time a mean opinion score drops, the call has ended and the customer has already formed an opinion. Jitter buffer depth, burst loss distribution, post-dial delay and RTCP reporting gaps all warn you earlier.

Read article

STIR/SHAKEN attestation levels explained without the acronym soup

A, B and C attestation are not quality grades — they describe what your carrier can prove about the calling number it is signing. Here is what each level does to answer rates and how to earn full attestation.

Read article

Planning a number port with zero downtime

Ports fail on paperwork far more often than on technology. A verified customer service record, batched cutover windows and a genuine parallel run will move four thousand numbers without dropping a single call.

Read article

The PRI decommissioning checklist nobody hands you

Cancelling the circuit is the last step, not the first. Fax machines, lift phones, alarm panels, franking meters and door entry systems all live on those channels, and every one needs somewhere else to go.

Read article

Per-channel or per-minute: choosing the billing model that fits your traffic

Predictable office telephony almost always wins on per-channel; spiky international and outbound campaign traffic almost always wins on per-minute. The break-even point is far easier to calculate than vendors imply.

Read article

Designing an IVR that customers do not hate

Menu depth, not menu length, is what drives opt-out rates. Cap the tree at two levels, name the destination before the digit, and always send an unrecognised input to a human rather than back to the top.

Read article

One-way audio: exactly what to log before you call support

Nine times in ten the cause is NAT, an SDP mismatch or a missing firewall pinhole. Capture the full SIP dialog, both SDP bodies, the RTP five-tuple and sixty seconds of pcap from each side and the fix takes minutes.

Read article

GDPR and call recording: how long can you actually keep it?

There is no single retention figure in the regulation, which is precisely the difficulty. Retention has to be derived from your lawful basis, and a recording kept “just in case” has no lawful basis at all.

Read article

Building a voice AI agent on a live media stream

Piping audio to a model over WebSocket is the easy part. Barge-in handling, endpoint detection and a hard budget on round-trip latency are what separate a convincing demo from something you can put in front of customers.

Read article

Callback maths: when queueing a call back beats making people wait

Virtual queueing only pays off when your promised callback window sits below the point where your abandonment curve turns sharply upward. Here is the arithmetic, using Erlang C and your own abandon data.

Read article

Sizing SIP channels from CDR data instead of guesswork

The one-channel-per-four-users rule was written for offices that no longer exist. Here is how to extract true peak concurrency from ninety days of call detail records and size a trunk against measured demand.

Read article

MOS is a lagging indicator — measure these four things instead

By the time a mean opinion score drops, the call has ended and the customer has already formed an opinion. Jitter buffer depth, burst loss distribution, post-dial delay and RTCP reporting gaps all warn you earlier.

Read article

Planning a number port with zero downtime

Ports fail on paperwork far more often than on technology. A verified customer service record, batched cutover windows and a genuine parallel run will move four thousand numbers without dropping a single call.

Read article

The PRI decommissioning checklist nobody hands you

Cancelling the circuit is the last step, not the first. Fax machines, lift phones, alarm panels, franking meters and door entry systems all live on those channels, and every one needs somewhere else to go.

Read article

Per-channel or per-minute: choosing the billing model that fits your traffic

Predictable office telephony almost always wins on per-channel; spiky international and outbound campaign traffic almost always wins on per-minute. The break-even point is far easier to calculate than vendors imply.

Read article

10DLC rejection codes: what each one actually means

Campaign registries return terse codes that rarely explain the real problem. We decode the rejections we see most often and show exactly which field in your brand or campaign submission to change.

Read article

STIR/SHAKEN attestation levels explained without the acronym soup

A, B and C attestation are not quality grades — they describe what your carrier can prove about the calling number it is signing. Here is what each level does to answer rates and how to earn full attestation.

Read article

GDPR and call recording: how long can you actually keep it?

There is no single retention figure in the regulation, which is precisely the difficulty. Retention has to be derived from your lawful basis, and a recording kept “just in case” has no lawful basis at all.

Read article

Designing an IVR that customers do not hate

Menu depth, not menu length, is what drives opt-out rates. Cap the tree at two levels, name the destination before the digit, and always send an unrecognised input to a human rather than back to the top.

Read article

Callback maths: when queueing a call back beats making people wait

Virtual queueing only pays off when your promised callback window sits below the point where your abandonment curve turns sharply upward. Here is the arithmetic, using Erlang C and your own abandon data.

Read article

One-way audio: exactly what to log before you call support

Nine times in ten the cause is NAT, an SDP mismatch or a missing firewall pinhole. Capture the full SIP dialog, both SDP bodies, the RTP five-tuple and sixty seconds of pcap from each side and the fix takes minutes.

Read article

Building a voice AI agent on a live media stream

Piping audio to a model over WebSocket is the easy part. Barge-in handling, endpoint detection and a hard budget on round-trip latency are what separate a convincing demo from something you can put in front of customers.

Read article

10DLC rejection codes: what each one actually means

Campaign registries return terse codes that rarely explain the real problem. We decode the rejections we see most often and show exactly which field in your brand or campaign submission to change.

Read article
Newsletter

One useful telecom email a month

New guides, rate-deck changes worth knowing about and regulatory deadlines before they bite. Written by engineers, sent once a month, unsubscribed in one click.

You are subscribed. The next Telecorenetwork Insights email goes out on the first Tuesday of next month — check your inbox for a confirmation from insights@telecorenetwork.com.
Please enter a valid work email address.

No sales sequences · No list sharing · See our Privacy Policy