Needs should be quantified in a way that can help to determine the dimension of the challenge, so that resources will be allocated properly.

An example is the random sample survey of bound periodicals to determine patterns of use, which in turn can provide information needed to design appropriate bindings or to suggest that certain titles not be bound at all. However, the way the data is gathered must ensure that the sample is randomly selected and that the number of items examined reflects the group as a whole. Moreover, the questions answered must be the right ones. In a bound periodical survey of this kind, the data sought can be quite simple: number of times circulated, evidence of use, and evidence of binding wear or stress. Early analysis of commercial library binding costs and procedures leads inevitably to the establishment of basic preservation operations that not only generate significant cost savings to further advance the program, but also, almost incidentally, help to build the essential infrastructure on which the program can develop.

The size of the sample to be surveyed has been a matter of some concern to librarians, especially as statisticians tell us that the relationship of the sample size to the size of the population (collection) is irrelevant. Carl Drott, the writer most widely cited by librarians, sets forth the principles on which sample size is based by describing a system involving "confidence level" and "tolerance." Confidence level is the measure of how certain the surveyor is that the result falls within the limits of tolerance, and tolerance is a measure and expression of the range of accuracy of the result. Drott recommends the base number of 400 items for most collection surveys.

The method of sample selection is of vital importance to the design and validity of sample surveys. For random sampling, each item in the collection must have an equal chance of being included in the survey. This is to ensure that the random sample is representative of the entire population. There are a number of ways to ensure randomness, the most objective being through tables of random numbers. Although many library preservation surveys have utilized tables of random numbers, the samples have usually been conditioned by "stratification," that is, a proportionate number of items from each stratum (subject area, collection, branch library, etc.) is surveyed to produce more coherent and particular results. Given the complexities of research library collections, the variety of formats and multiplicity of use patterns, careful stratification seems essential if the results are to be useful.

Other common methods of sample selection are systematic and fractional sampling. Typically, systematic sampling involves drawing the requisite number of items at a fixed interval from a list. For example, in a sample of 400 to be drawn from a shelf list card file of 100,000, every 250th card would be pulled, beginning at a random point in the file. Another, less tedious approach would be to take the linear measure of the cards in the shelf list and divide this by the sample number. Thus a 30-meter shelf list would yield a sample card every 7.5 centimeters. Fractional sampling uses the same basic procedure, except that the sample card is selected at random from within the interval, that is, from within each 7.5-centimeter group. The validity of systematic and fractional sampling depends on the unbiased ordering of the shelf list. It is generally assumed that conventional library shelf lists are ordered in an unbiased fashion.

The types of questions to be answered by a sample survey, and how many questions, obviously depend on the range of data gathered and the survey's purpose. If the survey incorporates a decision-making or prioritizing model, an attempt should be made to combine risk factors that place the surveyed item at jeopardy (such as lack of fire protection, inadequate environmental controls, and absence of a disaster plan) with vulnerability (such as high frequency of use, poor security, and high value) and condition.

Condition surveys are designed to identify the physical condition of the collection in a much more specific manner, often combining observations on damage and deterioration with incidence and type of use. The greater the number of condition categories, the longer the survey will take, and a balance has to be struck between the value of the data and the cost of the survey. During planning, the results of the survey can help to determine priorities, and can also serve as a rhetorical device to rally support for the program.