I’m doing an “analytics” project in which I have to collect data from small businesses (SMEs) in South America and try to come up with some insights about what are their needs in general. The project’s coach basically wants to collect data from social media and business association websites and do some text analytic on it to see if we can gain some insights.

Ideally, the idea of the project is that given a country and a sector, find what the companies in those sectors are up to and see what are their problems.

This is what I've thought:

  1. Find keywords in the twitter API and facebook API.

  2. Find News about SMEs in a given country and crawl the content.

  3. Find associations web sites. I dis this but the content here is useless most of the time.

The goal of this project is to gain some insights and see if some new products can be customized to these companies.

Thus, my question is: how would you solve a problem like this if you do not have any data? I was also thinking about getting some contact information from a sample of this companies and send them all a survey or something.

  • $\begingroup$ Can you simulate data if you don't have any data? $\endgroup$
    – Spacedman
    Commented Jun 22, 2016 at 22:28
  • $\begingroup$ I don't have any data on the problem. I was given the problem statement but it's quite open. How would you simulate it? $\endgroup$
    – nhern121
    Commented Jun 22, 2016 at 22:33

3 Answers 3


The World Bank runs Enterprise Surveys that 'provide the world's most comprehensive company-level data in emerging markets and developing economies. Business data are available on 130,000 firms in 135 countries'. You can get access by requesting on the site subject to some confidentiality terms.

The data are fairly comprehensive and would give you a good starting point to form hypotheses as which areas (access to credit, regulation, corruption, etc) matter most that you could then try to test by mining news or social media.

The World Bank also collects data on Doing Business that, while not specific to small companies, highlight country-specific challenges.

Plenty of research has been published around the two programs above.


Your best bet is to conduct focus groups or conduct market research, and that will probably be very expensive. You are not likely to find anything useful on Twitter, Facebook, or the news. No amount of text mining will solve your problem. This is not the answer that you want to hear, but your project is not set up for success, and you should look for/propose a new project rather than potentially waste more of your valuable time on this project.


I would also look at the academic articles, if they have done such analysis before or they have related data. If the topic may not be related but you can find some related data and get in touch with authors.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.