Explanation on using special interviewer types

The first and main thing to understand about special interviewer type numbers is that the phone system was NOT originally designed to handle lots of these numbers. If your sample is over a certain percentage (10%?) of special numbers than you may experience some unexpected results in your dialing patterns. The true issue here really does not depend upon the total percentage of special numbers in that file, rather it is how that number compares with the percentage of interviewing time allocated to interviewers with that special time. The higher the special interviewer number percentage is compared to the available time for that interviewer type, the more issues you will have to deal with. Any time you start seeing lots of numbers in the special interviewer stacks you are probably not dialing the sample as effectively as you should, unless you are using that stack as a “holding” stack.

The next biggest point to understand is there is often confusion between the special interviewer flag (22 of the phone record) and the special interviewer stacks (#318-326). The flag marks that record as only being deliverable to a special interviewer. The stack contains the special interviewer numbers that are available right now.

When the system goes through the algorithm to choose a number to dial and picks a special number, it puts that number in the special stack associated with that number instead of passing it on the interviewer. It then continues on looking at the next number until it finds a non-special number. If the next 100 numbers are all special numbers, it will move all 100 of them (silently) into the special stack. If the interviewer who is requesting a number is of a special type, it looks first in the special stack before going to the algorithm.

Example:

Suppose you have 20 phone records all in the Eastern time zone and there are no markets. The first 3 numbers are all non-special, records 4-7 are special type 1, records 8-10 are non-special, 11 and 12 are special type 1, 13-16 are non special, 17 is special type 1, and 18-20 are non special. When the numbers are added into the file all 20 numbers will go into the new bucket for the Eastern time zone.

Phone screen ‘A’ – Call scheduling status – (SPECINTV) (13 AUG 2003 14:27)
next: ** none set ** later: ** none set ** last put: ** none set **
Fresh: -1 Freshfirst: 0 Replicates: -1 In the Air: 0
Timed today: 0 Timed Later: 0 Resolved: 0
Onthefloor: 0 Hidden: 0 Free: 0
Owned recs: 0 Errors: 0 Duplicates: 0

Special 1: 0 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Zone: -5 -6 -7 -8 TOTAL Description
20 0 0 0 20 0) Fresh numbers

A normal interviewer logs in and gets the first 3 numbers and resolves them. Note for purposes of this example all numbers are being statuses as resolved to make the screen easier to understand, but it would make no difference if the numbers were given active statuses instead. When she asks for the 4th number, records 4-7 are transferred to the special interviewer stack, and the 8th number is delivered to the interviewer. The SPI A screen would look like this.

Special 1: 4 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Special 1: 6 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Special 1: 6 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Special 1: 5 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Special 1: 0 2: 0 3: 0 4: 0 5: 0 6: 0 7: 0 8: 0 9: 0

Notice that the count for the special stack is now zero. If this special interviewer resolves that number and asks for the next number she will be immediately be given record number 17 even though it is a special type number. There is no need to put the number in the special stack because this interviewer has the proper ability. After resolving that number the interviewer will just then get the last 3 numbers in the fresh bucket (record #18-20) before running completely out of sample.

This all works well as long as you do NOT have a lot of numbers building up in the special stack. Once a number is in that stack it really should have already been dialed and the longer it sits there the further away from that original time it will actually be dealt with. This is especially a problem for timed calls. We deal with this by putting the timed calls at the top of stack when they are moved to the special stack, but this does not help if you have lots of timed calls or if no special interviewer of that type is logged on.

For example, suppose your shop is in the Central time zone and you are working a consumer job that is called in the evening between 5 and 9 (local time) in all 4 US time zones. At 7:00pm CST a interviewer asks for a number, the system decides on a special interviewer number in the Eastern Time Zone, so it puts it into the appropriate special stack. No special interviewer is able to get to that number until 8:30pm CST (9:30 EST) which is now too late to call that number, so the system will set up a soft-call back for 4:00pm CST (5 CST) the following day. This is an example of just one phone number, but it could just as easily have happened to hundreds of numbers if your fone file is loaded with lots of special interviewer type numbers.

The problem can be further exacerbated if your day part times are not set carefully on the job. In above example, suppose daypartime1 was set to 9:00am instead of 5:00pm, because occasionally interviewers log on during the day to make a call or two. In that case, the number(s) would have been rescheduled for 9:00am respondent time (8:00am local time) the following morning. If no interviewer logged on till 5:00, all those numbers would probably immediately be rescheduled for 9:00 on the following morning because they are now too old to be called based on maximum timed callback age. To avoid this problem, you need to do one of the following:

1) Make sure your maximum timed callback age is set very high.
2) Make sure your day part times (especially time1) is set correctly.
3) Log 1 non-special interview on at 9:00am and all those numbers will migrate back to the special interviewer stack.

To minimize the consequences of the above design, if you have a file that has a high percentage of special interviewer type numbers you should do either one of the following to get the most efficient use of your sample.

1) Use markets to put each special interview type into its own markets. You can then use the Phone,L question to have each interviewer only retrieve a number of his market/special type. This will substantially reduce the number of numbers that will be put in the special buckets as only a timed call-back for a interviewer type different than the current interviewer will ever go in there and if there are any interviewers of that type logged on, they will get that number when they ask for it.

2) Make sure that your sample is randomly sorted so that the special interviewer numbers are NOT in blocks in the file. Then schedule your interviewing staff (as best you can) such that your percentage of special interviews matches the percentage of special interviewer numbers in the file. For instance, suppose 50% (English) (type 0), 30% is Spanish (type 1), and 20% is French (type 2). If you wanted a shift with 20 interviewers on this job, you would be best to schedule 10 English speakers, 6 Spanish, and 4 French.

In either case you should monitor the number of numbers in the special stacks and if it ever starts getting too large, you are likely to see at least some of the above issues.

Setting the special interviewer flag

There are 4 ways to set/change the special interviewer flag on a record.

1) You can set it by putting the code 0-9 in column 22 of the raw or ASCII phone record and then add the record in using Fonebuld. These numbers will be loaded into the fresh bucket with all the other sample. If you want to directly load these numbers into the special stack you will need to put SPECIAL=# in columns 31-46 of the raw record as well.

2) You can use a phone,O,<code> statement in the questionnaire. Records marked in this way will be put down in one of the timed or system call back stacks like any other number. You could use this for example to mark all suspends as special type 1 so that only experienced interviewers can ever resume suspended interviews.

3) You can also use a phone,S with a status code of 191-199 or 201-209 to set it in the questionnaire. Records marked this way will be put down directly into the special stack. You could use this for example to mark a Spanish speaking household that a non-Spanish speaking interviewer just called. You may want to call them right back since you know they are home.

4) You can use the # or f option in Foneutil to find a record and then navigate to the special interviewer flag field on the record and change it.

Assigning special types to an interviewer

There are 3 ways to assign a special type to an interviewer. Note that an interviewer can be multiple special types, but a phone record can only be a single one. For instance, an interviewer may be trilingual and speak, English, French, and Spanish, but a give phone number can only be assigned to a single language.

1) You can use columns 32-40 of the employee file to permanently set this. The employee file is called employee.xxx for DOS/UNIX and is typically found in the Ipcfiles sub-directory of wherever the CfMC Software is installed. For MPEXL the file is just called employee and can be found in the Control group of wherever the software is installed.

2) The supervisor can type S=<values> at the end of a Start/STS/ATM command to assign all of the started interviewers those special statues. To start up stations 11-20 on the bank job and assign those interviewers special statuses 3 and 5, the command would look like

Sts 11-20 bank s=35

3) The interviewer can type S=<values> after the name at the enter interviewer ID prompt. For instance, if your ID was 0001 and you wanted to be assigned special types 2 and 4, you would enter

Type interviewer ID [,special type] or “quit” –> 0001,S=24

If more than one of the above have special settings, you get the last one set. In other words, the settings at the interviewer ID prompt override any previous settings, and the settings from the supervisor override any settings in the employee file. Note, you can say S=0 to run as a normal interviewer even if values are set in the employee file.

Other features that interact with special interviewer types