sent. Thus a considerable fund of
information is obtained for future use. This information, however, has
to be very carefully handled because it may easily become
misinformation, for we cannot forget the appeal of the product itself.
No one as yet has ever been able to gauge in advance the appeal of a
product.

Some apparently very bad letters have sold very good products. Some
apparently very good letters have quite failed to sell what turned out
to be bad products. Therefore, the information that is obtained in the
circularizing and sale of one product has to be taken warily when
applied to another product. It should be taken only for what it is
worth, and that is as a general guide.

[Illustration: Specimens of business letterheads]

Several concerns with a mind for statistical information have in
the past so carefully compiled the effectiveness of their letters,
but without regard to the product, that they have discovered an
inordinately large number of things that cannot be done and extremely
few things that can be done. This is the danger of placing too much
faith in previous experience. One of these companies entirely discarded
its records of what could not be done and started afresh. They found
that several of the methods which they had previously used and
discarded happened to do well under changed conditions and with
different products.

If any large expenditure be contemplated then many tests should be made.
The kind of envelope, the manner of addressing, the one cent as opposed
to the two-cent stamp, the kind of letterhead, the comparative merits
of printing, multigraphing, or electric typewriting, the length and
composition of the letter, the effect of the return card, the effect of
enclosing a stamped return card or a stamped return envelope, the method
of signing, and so on, through each detail, must be tried out. No test
is ever conclusive, but very little information of value is to be
obtained by circularizing less than five hundred names. These names may
be taken sectionally or at random. The sectional method is somewhat
better, for then comparison of results in several sections may be made,
and it may turn out that it would be well to phrase differently letters
for different sections.

The returns on the letters are not of themselves conclusive. If one
section responds and another does not, it is well to look into business
conditions in the sections. It may be that in one section the people