Software marketing on the Internet: the use of samples and repositories

Abstract

This paper examines one of the most important marketing strategies by software producers on the Internet. That is whether to offer free samples and, if so, whether to list the samples on shareware repositories. I show that firms with higher value products have a greater incentive to offer free samples but are more reluctant to do so if they are well known, and even when they do are less likely to be listed on shareware repositories. I then proceed to use four types of Probit-based models to corroborate the findings from the theoretical model.

Keywords:

• shareware
• software
• Internet
• brand
• search
• sample

JEL Classification :

• D42
• D43
• D82
• D83
• L13
• L15
• L81
• L86

Acknowledgements

Thanks go to Bruno Jullien for advising me on a previous version of this paper, to Tina Chang and Peter Moffatt for their advice on the empirical section, and to Bruce Lyons and anonymous referees for suggestions. Support from the ESRC Centre for Competition Policy is gratefully acknowledged.

Notes

See Bhargava and Feng (2006) and Bhargava and Feng (2007) for an analysis of possible fee-induced bias in search results.

See Chevalier and Mayzlin (2006) for an application to the online context.

Consumers tend to choose relative popularity as their main way to arrange software listing when searching a repository (Duan, Gu, and Whinston 2006) and will tend to ignore any software that does not appear on the first page of results. One could also have selected software by the total number of downloads on the repository. However, the length of time the product was available differs greatly firm by firm, so a product that appears to have been downloaded often may have been so because it has been listed for a long time. As an alternative, taking the daily average of downloads does not take into account the difference between older products that were present when download.com was starting, and newer products that benefit from the present popularity of this site. Moreover, recently listed products often display high daily download numbers because of novelty, and this number then tapers off in a way that is difficult to predict. Using the ‘number of downloads last week’ suffers from the same novelty bias.

Other data on CNET download.com were collected as well, including the size of the downloads, the number of downloads made since listing, ‘last week's downloads,’ and the number of ratings made.

Consider, for example, sites 1 and 2, where n ₁ sites link to site 1 and n ₂ sites link to site 2. Suppose that there are N=n ₁+n ₂ sites and all sites link either to site 1 or 2. In a random search, the probability one finds a site that links to site 1 is The logarithm of n ₁ approximates the shape of this function.

There are some examples of the use of price as a measure of value in the literature on cross industry analysis of intra-industry trade (Greenaway, Hine, and Milner 1995).

Ratings based on less than five opinions were ignored unless, in the case of amazon.com, a sufficient number of consumers signalled that the reviews were useful (indeed, a review can be so ‘definitive’ that no reviewers would submit additional reviews).

User ratings do not necessarily express the absolute value level of the software, or even its value in relative terms vis-à-vis the competition, but its value vs. what the consumer expected from the product. Products that are good value for money or exceed expectations will therefore tend to be highly rated, rather than products that are of higher value.

One could also assume that firms first choose whether to list on the repository. If they do so, they have to offer a sample. If they do not do so, they can still choose to offer a sample on their website. However, one cannot observe variations in the choice whether to offer samples or not among those selected firms that choose to list on the repository. This is why one has to first evaluate the decision whether to offer a sample or not, and then the decision whether to list on the repository.

Twelve from Microsoft, six from Avanquest, and five from Adobe and from Symantec.

In the case of download.com, publishers are able to add downloads of older versions of their products to the downloads of their newer versions, thus keeping the ranking of the older version.

Another common strategy is to start with a low-range and cheap product and then gradually learn through experience what one's needs are so as to later on buy a more sophisticated and more adequate product.