It appears that one is required to estimate a proportion (of companies that have a given property) by sampling. The method is to draw a number of samples, and then find out how many companies in the sample have the property, and find the estimated proportion. Given the number of samples and the estimated proportion, we can draw a confidence interval for the actual proportion. You can lookup this article for information on computing the confidence interval. The higher the confidence value required, the wider is the interval. The more the number of samples, the smaller is the interval.
Here, we are given that based on the earlier study the proportion may be around 0.70. This is one of the factors in computing the confidence interval. Using this, the confidence interval width and the confidence value, we can compute the number of samples required.