Editor's note: Dale Kulp is president of, and Amy Starer is director of sales marketing for, Marketing Systems Group, developers of the GENESYS. sampling system.

Over the past few months we here at Marketing Systems Group have been amused at the estimates of "dialing costs" quoted in ads and published in trade magazines. We felt this subject was long overdue for an injection of reality and common sense.

It could be, of course, that the person who developed those estimates had little or no data collection background, had an axe to grind, or possibly, their calculator just malfunctioned. Nonetheless if it really costs these phantom research firms $.69, let alone $.90, to dial a non-working number, we suggest they have some serious internal problems.

What makes us think we have any better handle on these costs? Well, our staff has over sixty years of experience in market and survey research, including data collection at some large interviewing facilities. We have also consulted with a number of WATS managers to arrive at a reasonable consensus on the cost of dialing non-working and other non-productive numbers.

Those unfamiliar with data collection might be unaware of or confused about some very basic facts:

1) There are productive and non-productive portions of an interviewing hour. Averaging the "total cost" over just the unproductive part of the hour grossly inflates the "cost per dialing," producing unrealistic and unobtainable "savings."

2) There is a big difference between the selling price of a WATS hour, and the direct internal cost. Averaging the "margin" into costs will also overstate the cost per dialing and potential savings.

3) "Non-productive dialings" are not all the same. Some involve only interviewer labor, others incur both labor and phone charges.

The point is that one can easily inflate the costs associated with unproductive dialings by 300 to 400%, if that is the objective.

Why our concern with dialing costs? We have invested significant effort in evaluating the costs associated with dialing non-productive sample. Just accepting grossly inflated estimates might make our job easier, but it wouldn't do much for our firm's reputation.

The following paragraphs will focus on two methods to estimate the savings one can reasonably expect from eliminating non-working numbers (i.e., the number of interviewer hours actually saved by not having to dial non-working numbers). Since there are various methods used in estimating the number of data collection hours required, our approaches may not exactly match yours. However, our common sense approaches are fairly generic and should be applicable to any operation.

The first method, used quite frequently to estimate the required number of interviewing hours, employs a standard known as an "effective interviewing hour." Basically this is an estimate of the average number of minutes in each hour that interviewers actually spend administering questionnaires.

A consensus of research firms indicates that a 25 minute interviewing hour is generally applicable for RDD household samples, while a 28 minute hour is a reasonable benchmark for a good list-based household sample. The basic difference between these samples of course is the relative absence of non-working numbers in the latter.

A strong case can then be made that the difference in the "effective interviewing hours" is due to this absence of non-workings in listed samples. One can then reasonably conclude that approximately three minutes of each interviewing hour is spent dialing these non-working numbers. In other words, if it were possible to identify and remove all non-working numbers, one could expect a maximum savings of about 12% in the interviewer hours required on any RDD sample project.

A hypothetical 100 interviewing hour project would then require just 88 hours. But, what does this 12 hour savings really represent? The hours "saved" by not dialing non-working numbers have not decreased any line charges-phone costs do not accrue for dialing non-working or disconnected numbers. Assuming your interviewers are paid $7.50 per hour including benefits, your actual, direct "savings" could total $90.00.

A second method of assessing "savings" would be to develop a dialing rate per hour just for non-working numbers. This is very difficult to do in practice since interviewers are not normally dialing just non-working numbers for any extended period of time. However, there is again, a common sense approach to the dialing rate problem:

How many dialings are made per hour? This will vary by study and project, but a conservative estimate is about 30. If we again assume that 25 minutes out of each hour is actually spent interviewing, we are left with another 25 minutes of time spent on the phone. In other words, interviewers spend about 25 minutes interviewing live households, and another 25 minutes finding those respondents. What if interviewers spent no time interviewing and spent the entire 50 minutes just "dialing?" Would the dialing rate double to 60 per hour? Probably not, but a reasonable assumption might be 50 dialings per hour, or about one per minute.

If we use the 50 per hour dialing rate and $7.50 per hour for interviewer wages, each non-working number eliminated or, not dialed, would "save" about $.15 in direct interviewer cost.

Going back to the first method, we estimated a potential cost savings of 12 hours and about $90.00 for that 100 hour project. If we again assume that the average dialing rate would have been 30 per hour, the project would have required about 3,000 total dialings (100 hours X 30 dialings per hour), of which roughly 600 of those dialings resulting in non-working numbers, again the result is a direct cost of $0.15 for each dialing.

Eliminating non-working numbers does not decrease line costs, but one might argue that we should include supervisor, hiring, training, administration and selling costs in estimating the total savings. Let's assume the real cost per WATS or interviewer hour is $16.00-the actual cost, not selling price. If we deduct $3.00 per hour for phone charges, we are left with $13.00, or just $0.26 per dialing.

Of course, there are also indirect benefits resulting from increased productivity: it may mean less overflow work, more projects completed on time, and, assuming a constant workload, 10% fewer interviewers, or 10% greater capacity.

It should be quite clear that there are significant benefits and savings possible by eliminating non-working numbers from RDD samples. However, one must be wary of non-researchers' naive assessments of "costs and savings." Using a marked-up selling price and dividing by an average dialing rate will of course provide a cost per dialing-an inflated and meaningless cost, but a cost nonetheless.

So the next time someone tries to tell you that a "reasonable" cost per dialing is $0.90 or even $0.69, you might want to question their motivations. Using a low estimate of 30 dialings per hour, the "cost" of dialing non-productive numbers is either $27.00 or $21.00 per hour. Every company in the industry must be losing money; by their logic the "costs" of just the unproductive dialings is equal to or greater than the going WATS hour rate.