The research objectives require several types of companies:

l venture capital-backed start-ups

-successful and unsuccessful (in order to determine whether a grouping based on outcomes has better explanatory power than a grouping based on initial conditions).

-technology-based and non-technological (in order to determine whether apparent differences between VC-backed start-ups and others stem from the effects of technology or from the effects of size.

l small start-ups

-a reasonably random sample of such start-ups encompassing the usual range of outcomes.

-a set of highly successful small start-ups (in order to determine whether such companies match large start-ups better than the general population of small start-ups from which they came.

In all, the sample includes 52 companies, as described in Table 1. The technology-based companies are a random sample, stratified on performance from the portfolios of four venture capitalists. They were collected by Keeley & Roure (1993). The non-technological, VC-backed companies are also from four venture capitalists. The majority of VC-backed companies are technology-based, so the non-technological sample proved difficult to obtain. We used the only non-technological investments of three funds, and a block of five investments from the fourth fund. Data on the non-technological companies were gathered by the authors for this paper.

The small companies are from a mailing list of several hundred such companies compiled by a local entrepreneurship center. They were randomly selected from the list. They were collected by Keeley & Knapp (1993). The high performers among the small start-ups represent rare events, and as such are harder to sample randomly. The authors used the INC 500, and the local business press as sources of companies. Data for nine of the companies were collected by Keeley & Knapp (1994), with the remaining five collected immediately prior to this study. The large start-ups all received at least several million dollars of private funding, and had the avowed intent of seizing a major share of a large market. The small start-ups received between 0 and $1 million in total equity funding--with only one receiving more than $250,000.

This study uses Eisenhardt's (1989) inductive research method. As she notes the method is particularly suited for developing theory, including testing on an exploratory basis, from case studies. Semi-structured interviews serve as the primary method for gathering data this study. During the interviews the founder (or in about 20 percent of the cases a director or investor) assumes the role of company historian. Interviews averaged about 1.5 hours. Most were tape recorded--in some cases equipment failure or background noise prevented capturing a full transcript--and in all cases the interviewers kept detailed notes. In all four interviewers assisted with the data collection (Keeley, Knapp, Rothe and Roure). All interviewers participated in at least two interviews jointly in order to assure reasonably consistent coverage. The interviews avoid asking interpretive questions in which the "historian" would need to use personal judgment. Instead they focused on events and on statements of the principals--e.g., "When did you start the company?" "Who got the idea for the product? How?" "At the time you started the company, what reason did you give others for starting it?"


Description of the Sample

  Small Start-ups VC-backed
  Small High Growth Non-tech High-tech
Number 15 14 8 15
Average age 5.3 11.5 8.2 7.2
Average number of founders1.36   1.42 1.62 6.1
Percent successful 80 100 50 57
Average sales  
First year $67K $467K $1250K 0
Fifth year $235K $11.6M $36M $27M
Average amt. equity financing $29K $166K $11.6M $15M
Industry membership (% in each)  
Construction/development 7 22    
Mfg: non-tech. 22 28    
Tech (include. software)       100
Retail 14 7 75  
Consumer services 7 22 25  
Business services 50 22    


The analysis of the data proceeds in two stages: identifying features of possible interest, and then assessing their importance and their genesis for all of the companies in the sample. Identification of features begins with findings from the literature on entrepreneurship and organizational research. Other features may emerge from the data in the study. Thus the process is circular: a review of the data suggests a possible feature or the research literature suggests looking for a specific feature; then the sample is checked as to the prevalence of the feature and in the process other characteristics may emerge for addition to the analysis.

The dataset is longitudinal for each company, covering a multi-year period from inception through initial success or failure. The analysis of the reasons for success or failure searches for events that in hindsight seemed to heavily influence the company's fortunes. Then a preceding event was sought that influenced later history. Finally, elements of the initial conditions and processes are examined for possible causal roles in the set of events. This "chain of events" is identified by the authors, not by the "historian." To have asked the "historian" to interpret cause-and-effect could have biased the responses in the authors' views.


Top of Page
Previous Page | Main Menu |Next Page
Return to 1996 Topical Index

1997 Babson College All Rights Reserved
Last Updated 5/1/97 by YuBei Teng.

To sign-up for the Center for Entrepreneurial Studies' publication lists,
please register with the
Entrepreneurship WebTeam.