I discovered this by accident despite my having written this website and being technically knowledgeable. I noticed that I have not received one telemarketing call in the last two months. (This was originally written 3-99.) I'm wasn't sure how that happened, but I have a few clues:
skip to what I did.
skip to similar things that may work.

1. I don't think it's this website. I believe I have changed some telemarketing practices (most notably muted tape warning beeps and called attention to continued refusal to identify their telemarketing business), not to mention the very real effect of the "Do Not Call  List. It would make some business sense for them to maintain a blacklist of troublemakers. But my new phone number isn't listed under my name, and therefore would not have made it to a blacklist.
2. It may be my answering machine. Since I no longer use my phone for business (unless you count a small aircraft parts business), I have a non-standard voice pattern. See below.
3. I never gave a valid home phone number to un-trusted commercial entities. - not it - My phone number belonged to someone else until recently. Maybe he didn't give his phone number out either, but it was listed.
4. Space aliens. could be
5. (FWIW, I do not request telemarketers to be placed on the "Do Not Call" list for the telemarketing company itself. That's an effective tool, but it would prevent me from hearing their latest tricks, such as lying about their identity)

I believe it's the answering machine, because I believe it's a basic issue with "call progress detection" software.

This sounds esoteric, mainly because it is esoteric and arcane. So bear with me or skip to the next section.

I believe my answering machine causes a fault in the "answering machine detection" portion of the scum's "call progress detection" software.

Since I no longer use my phone for business, I have my dog "answering" on the machine.

Well, my dog has a one-word vocabulary. Actually, I can count at least five vocalizations, which he uses to form at least three times that many expressions, but you get the idea -- one word. I think it starts with an 'F'. There are some people who have a similar one word vocabulary, also with a few additional vocalizations, such as "you." But back to the answering machine...

"Dude" (1986-2006)

It seems that when dogs record on an answering machine, they do so the same way they answer live (bark!, pause bark!). So the vocal patterns match that of a live human (or dog) answering the phone.

The result is a short "hello", followed by a 0.7 sec. pause, followed by another short "hello". The total time period is rather short, despite a pause between the two vocalizations. The call then goes to the answering machine tone, but for some reason the software can't handle this.

The total time period of the vocalizations is less than that of a long residential "greeting".

The result is that the scum's "call progress detection" software fails to detect a standard answering machine message. When the software fails to detect the answering machine, it passes the call to a live agent. After a few dead connections, the number is flagged as out-of-order.

• The timing of the vocalizations results in human voice detection. This includes:

  a short initial pause

  brief vocalizations

  a brief total time for both vocalizations

  a short time interval between vocalizations

  high sound level peaks after an initial pause (low modulation levels followed by high modulation levels)

• The sharpness of the barking mimics an "unknown" SIT tone from the phone company
• The combination of the sharp barking and the answering machine "beep" causes a data analysis fault
• The "Call Progress Detection" software detects a live human voice pattern in combination with the answering machine "beep", which causes a data analysis fault.
• the dog's voice mimics an "Intercept SIT" tone.

The "beep" from the answering machine should indicate an answering machine, but that occurs after a live human (or canine) response is detected.

This isn't easy to correct because many answering machines have a lag time between the message and the tone. The tone is perhaps identifiable but this additional step consumes time. Worse yet, if the call is handed off to the "agent" (the flunky), this consumes additional time.

The telemarketers program their equipment to hang up on the answering machine. When "Dude" answers, their equipment assumes it's a live person to disturb and hands off the now-dead call to a telemarketer. The telemarketer then catalogues the number as "dead" because the dog never picks up.

Despite their protestations to the contrary, telemarketers consider their time to be too precious to consume on what they call non-productive time. They simply waste yours.

What I suspect happens is that "call progress detection" conflicts get resolved as a fault. Typically this would be reported as "Invalid for Calling". After all, why should the software tell the telemarketing scum that it couldn't accurately perform the answering machine detection?

So apparently an answering machine which mimics a live person (human or canine) will force your number off their master lists.

I don't know if this works in combination with a more conventional answering machine message, but I suspect it does. I don't think the telemarketers are smart enough to compare such a bad connection with a previous valid connection. I mean any good Unix programmer would figure that one out, but we're talking about lowlife scum here.

1. The predictive dialer calls the number.
2. Answering machine picks up and transmits a greeting (the dog)
3. Call progress detection (part of the predictive dialer) determines dog is live telemarketing victim.
4. Predictive dialer hands off call to flunky.
5. The flunky tries to talk with the dog but the dog is ignoring him/her/it because dogs generally don't like telescum either.
6. The flunky records the call as incomplete.
7. The predictive dialer logs the call. This is recorded in a failed attempt memory store.

   Alternatively the predictive dialer responds to the beep, but since this was supposedly a live victim, the predictive dialer records an error.

8. The predictive dialer compares the failed attempt memory store with a predetermined maximum. If the number of failed attempts exceets the maximum, the number is removed from the telescum's master list. (This is the same process which limits the number of repeat calls when the flunky flags your number to "call back".)

For some reason I've started to get some "Hello...Hello??...Haveaniceday" calls on my machine.

Any marginally intelligent human being would figure out they're talking to a dog and that the dog is a recording on an answering machine. Okay, some people are a bit miffed by the dog, but they figure it out. But telescum try to talk to the dog. Not the dog; the machine recording of the dog. If there are any telemarketers reading this, pay attention here...

That dog will not hold a conversation with you. It's a machine. Try to grasp the concept.

The real dog is probably lying down somewhere thinking how stupid you are. He understands the concept and doesn't alert to the machine when he hears himself.

Everyone else -- don't say "hello" more than once. (If it turns out to be a friend on a bad cellphone connection, you'll probably recognize the voice.)

As I mentioned the scum program their equipment to hang up on answering machines. Normally "answering machine detection" is performed by:

A length of time of the victim's "hello" greeting.

This can sometimes be used to detect a business greeting, although some businesses have greetings as short as residential greetings. A typical residential "hello" is less than 1.3 seconds, but the "answering machine detection" will typically allow a little over 2.5 seconds. This is ample time for a person to say, "Hello", "Are you there?", "Linda Tripp Residence",   or "Hello. Monica Speaking.". An answering machine is expected to exceed 2.5 sec.

A pause after the victim's first "hello".

Usually a live person will wait for a response.

Pause between pickup and "hello"

A live person will normally provide a "hello" in less than 0.5 sec. of

A pause between vocalizations.

Silence is considered to exceed 0.7 sec. Inter-word gaps are typically less than 0.3 sec.

An initial hiss pattern

This is now pretty much obsolete. "Answering Machine Detection" devices used to listen for a hiss pattern caused by the answering machine tape winding up to speed. Many newer answering machines use solid state memory, which has no detectable hiss. Similar hisses are caused by cordless phones. People will often answer their calls with a speakerphone (even if they then pick up the receiver). So hiss patterns aren't particularly reliable.

Tones

These include SIT tones such as "busy", "reorder busy", "intercept" and a beep used by some telcos to signal use of their voice messaging system. In other words, SIT tones are a sort of prenatal testing of phone calls.

So??

If the telemarketing scum detect an answering machine, their predictive dialers will keep trying several times until they get through. If their equipment fails to detect the answering machine, then this causes a fault.

The fault is probably reported as an invalid number, to give the telemarketing company the impression that their equipment is running smoothly. If the call does get handed off to a live agent, then the agent would have to manually reset the software to re-dial the number, which wastes too much of their valuable time. (Remember their time is valuable; not yours.)

The people who program predictive call detection equipment are somewhat limited in their ability to get their "call progress detection" software to work correctly. Coding predictive dialer software, if not the lowest programming job, is pretty close to whatever is lowest.

But that's not the problem. Let's presume the programmer is competent but just has low self-esteem. The problem is how to detect a real person has answered the phone, based on speech and noise patterns. It isn't easy.

So if most calls are initially mis-detected as to call outcome type, then the scum's equipment will toss your number off of their master list!

Can't Predictive Dialers Distinguish Recordings?

Yes.

The problem is that in order to distinguish recorded people (or dogs) they'd unduely complicate their equipment. This worked better when tape artifacts or digital compression artifacts could be easily detected. To some extent this is still true when the victim is connected through a typical SS7-type switch, but there are too many variables. (Variables include cordless phones, poor connections, the limited audio response of a typical POTS phone line, and "tricks" used by long distance carriers to establish connections such as grey market VoIP.) If a predictive dialer is programmed to distinguish recorded dogs from live dogs answering the phone, it is liable to lose live victims.

I've had success with the dog barking

See above for the pattern.

A short courteous answering machine message.

Your callers know what an answering machine is and long distance callers don't want to hear you read off the owner's manual. Try "Wait for beep." or just "Please wait."

Be creative!

If you have a quiet dog, use a cow (fold hands prayer style, cupped over your mouth, and say "r-r-r-r-r").

Inter-word pauses in excess of 0.7 sec.

Insert pauses between sentences and keep it short.

Use an Intercept SIT tone.

I have seen references to equipment which plays an intercept SIT tone (Zapper, $30 at Radio Shack) and (www.dialerbuddy.com - two models).

Record "disconnect" SIT tones at the beginning of your answering machine message.

sample (15k), from www.twpyhr.com If it doesn't play clearly directly from the net, try downloading it.. Here's a digitally synthesized (45k) sample (courtesy of Dale).

Anecdotal reports are that the first beep controls predictive dialer response. This contradicts Dialogic's literature, and so I recommend this only if you don't want the three tones.

If you are able to assign telemarketers to a particular answering machine message (see mailboxing answering machines), then you can assign these calls to a Predictive Dialer mailbox.

The Predictive Dialer mailbox includes:

a brief "hello"
followed by
a long pause ( >3 seconds)
followed by
jibberish or several requests to talk louder

Assign known telemarketers and calls with a combination of [CID number and "unavailable" name] to that mailbox. Keep the message clean just in case a legitimate call comes in that way.

The idea is to prevent the scum's equipment from detecting your answering machine. That substantially increases the difficulty of their targeting you.

Normally the "predictive dialing" software will determine if a call is answered by an answering machine. But before this happens, the software must determine if the initial call progress. Items such as "no answer" (more than 5 rings), modems, busy (engaged) signals and operator intercept tones are easy, but what happens if the call outcome category is undefined?

It turns out that if the software detects a particular type of "call outcome" such as a live victim, but the result is a dead connection, then the software will recategorise the number as "invalid for calling" or "SIT, vacant". (The predictive dialing software normally doesn't advise the telemarketing scum that their equipment failed to detect the call outcome.)

So if most calls are initially mis-detected as to call outcome type, then the scum's equipment will toss your number off of their master list!

(This page will by Y2K compliant by 1st quarter, 1901.)

Future developments are discussed at the tech data page

References are on the Dry Technical Data page.

back to Counter-Telemarketing Tactics - (this site)
back to Dry Technical Data

Courtesy of Stan Protigal

site first posted November 3, 1996; this page 1999
rev December 12, 2011 This page copyright 1999, Stan Protigal
This website can be found by searching for "The Telemarketing Scum Page" on a search engine.

Seattle Community Network
SCN Member Pages