False Answer Supervision (FAS): How to Detect and Stop It

What false answer supervision is

In SIP, a call becomes billable when the far end sends 200 OK to the INVITE. Answer supervision is the signal that the called party has picked up. False answer supervision happens when a carrier in the path sends 200 OK itself, without the called party answering. It then plays something to keep you on the line: recorded ringback, silence, a fake busy tone or a "number not available" message.

There are three common variants:

• Early answer: the call is answered during ringing, and the real answer (if any) comes later. Every answered call is billed for extra seconds.
• Answer on no answer: calls that would have rung out are answered with a recorded ringback or message, so unanswered calls become billable.
• Answer on failure: calls to invalid or unreachable numbers are answered and played a tone instead of failing.

Why it matters more for dialers

A conversational campaign might lose a few seconds per call to early answer. A predictive dialer can lose far more: every unanswered attempt that is falsely answered becomes a billable call, and agents are connected to ringing tones instead of people. ASR rises, which looks like a better route, while productivity and cost per contact get worse. If you only watch ASR, FAS looks like success.

Signals in your CDRs

SELECT LEFT(dst, 5) AS prefix,
       COUNT(*) AS answered,
       ROUND(100.0 * SUM(billsec <= 6) / COUNT(*), 1) AS pct_under_6s,
       ROUND(AVG(billsec), 1) AS avg_billsec
FROM cdr
WHERE disposition = 'ANSWERED'
  AND calldate >= NOW() - INTERVAL 1 DAY
GROUP BY prefix
HAVING answered >= 100
ORDER BY pct_under_6s DESC
LIMIT 20;

Compare the result with the same destination on a different route or the same route a week earlier. A destination where the share of calls under six seconds jumps from 8% to 30% overnight deserves a test.

Proving FAS with test calls

1. 1

   Use handsets you control

   Get SIMs or fixed lines on the destination network, or use a test-call service with real numbers there.

2. 2

   Let calls ring out

   Place calls and deliberately do not answer. On a clean route these calls end with 480, 487 or similar. With FAS they show as answered.

3. 3

   Answer late

   Answer after a measured delay, for example 15 seconds. Compare the handset's answer time with the 200 OK timestamp in your capture.

4. 4

   Call an invalid number

   Call an unallocated number in a valid range. It should fail with 404 or similar, not answer.

5. 5

   Capture everything

   Record a pcap with sngrep or tcpdump for each call so you have SIP timestamps and audio as evidence.

# signalling and media for the carrier host
sudo tcpdump -i any -s 0 -w fas-test.pcap \
  host sip.example-carrier.net or portrange 10000-20000

# or watch the SIP ladder live
sudo sngrep host sip.example-carrier.net

What to do when you find it

1. Stop or reduce traffic to the affected destination on that route.
2. Open a ticket with the test results, pcaps and a list of affected Call-IDs.
3. Request a credit for the affected calls. Reputable carriers credit proven FAS.
4. Ask what changed. FAS appears when traffic is moved to a different partner, so the carrier should be able to identify and remove it.
5. Add the destination to weekly sampling so you catch any recurrence.

Estimating what FAS costs you

Work out the cost in minutes, not currency, so you can compare across destinations. Take a dialer campaign that makes 50,000 attempts a day to one destination with a true ASR of 25%. If a carrier in the path falsely answers half of the 37,500 unanswered calls and holds each for an average of 12 seconds before the dialer or agent hangs up, that is 18,750 extra billable calls and about 3,750 extra minutes a day, all of them worthless. On 60/60 billing, each of those calls is charged as a full minute, which turns 3,750 minutes into 18,750.

The indirect cost is larger. Agents connected to ringback or silence are not talking to customers, so cost per contact rises sharply even though ASR looks better than ever.

Detecting FAS from the audio

The most reliable automated test listens to the first seconds after 200 OK. A real answer produces irregular speech energy, or a voicemail greeting. A false answer produces a periodic ringback cadence (for example 400 + 450 Hz, 0.4 s on, 0.2 s off, 0.4 s on, 2 s off in the UK; 440 + 480 Hz, 2 s on, 4 s off in North America), a continuous tone, or digital silence. Simple tone detection on a sample of answered calls flags most cases.

Silence alone is not proof: NAT and firewall problems also produce one-way audio. Confirm that RTP packets are arriving before drawing conclusions. If packets arrive but carry silence, suspect FAS; if no packets arrive, check your network first.

Contract terms worth having

• An explicit prohibition on false answer supervision anywhere in the provider's path.
• A credit commitment for calls proven to have been falsely answered.
• A dispute window long enough for you to analyse CDRs, typically 30 days or more.
• Access to CDRs with SIP response codes and timestamps so you can reconcile.

A weekly FAS routine

1. Run the short-answered-calls query for every destination with meaningful volume and compare with the previous four weeks.
2. Pick the three destinations with the largest increase and place five ring-out test calls to each.
3. Review agent dispositions for "ringing", "dead air" and "answering machine" on those destinations.
4. Sample ten answered calls per destination for tone analysis or manual listening.
5. Log the results in the same sheet you use for acceptance tests, so trends are visible over months.

This takes an engineer about an hour a week and catches most problems within days of them appearing, long before they would show up in a monthly invoice review. On the CLIRoutes portal, short-call share and ACD per destination are already calculated, which removes the first step.

How CLIRoutes prevents FAS

We run automated test calls on every active destination, including ring-out and invalid-number tests, and compare SIP answer time with media analysis that detects ringback tones after a 200 OK. Partners that show FAS are removed from the route, and affected customers are credited. You can see short-call ratios and ACD per destination in the portal, so you can spot anomalies yourself rather than wait for a monthly report.

FAS checks are part of our 7-day route acceptance test plan. For the metrics behind them, read ASR, ACD, PDD and NER.