These are the top languages

This experiment was based on a random sample of 22.632.977 Tweets collected using Twitter Streaming API, 23.4.-7.5.2015.

We used the language attribute (“lang”) which comes as a data field of a Tweet to determine the language.

This experiment is still ongoing, the official results will be published in a research paper.

Advisory boards help startups to take right strategic decisions

In advisory board typically consists of a team of 3-5 people who use their individual expertise to help a companies’ management team to take the right strategic decisions. By asking the right questions they challenge the management’s assumptions and provide advice ideally leading to improved success. When a startup placed the first product on the market, got some first traction and has the first investor on board, it is the right time to set up a team of advisors. It is of great importance that advisors provide different types of expertise in disciplines which complement the fields of the management and are critical for business development. There are different remuneration schemes, one popular option is to pay advisors with 0.5-3% of company shares, In return advisors would meet with the management team on a regular basis every 1-2 month to discuss business development and assist with critical questions on demand. The relation between the company and the advisors is coined by respect and mutual trust, which makes Non-Disclosure Agreements superfluous as trust and discretion are basic preconditions. An advisor is different from a consultant, in a way that there is a more independent relationship, an advisor does not work for you, but is a long-term companion to the company who provides critical advice as needed.

The importance of advisors and finding the right ones.

As a potential future startup founder I learned from this talk that it is important to keep in mind already in the beginning that the startup you are founding is not going to stay a small, flexible team of friends for a long time, but in the best case is growing fast. Which creates the need for organization on the one hand and the responsibility to take the right decisions is increasing. As a manager you are expected to stick to your decisions and to not change decisions arbitrarily, at the same time you have to deal with a lot of uncertainty. Every professional advice you can get that helps to take the right, or at least prevent you from taking bad decisions is of critical importance. This assumes however that you picked the right people to get advice from, which might not be so easy. Advisors should not only fit in terms of expertise and knowledge to the company but also with their type of personality and communication. My conclusion from this is that you should see any business conference, talk or meeting also as an opportunity to spot potential advisors, maintain a list of professional, potential advisors already early one (maybe even before you found a company), which might help you to find the right advisors, once you need them.

Who dares wins.

Thomas Peruzzi left his secure job in a large IT company to found his own business in a time when his family was expecting the first child and was building a house. While many people would assume this time to be an inappropriate moment to start a new business, Mr. Peruzzi saw in this moment an opportunity to change his path entirely before settling down. While doubling his income in each of the following 8 years he never regretted his decision. For me this was a very impressive moment of this talk and encourages me to question our assumptions on what we consider appropriate.

Talk of Thomas Peruzzi, on 11 March 2015, Public Lecture Series on Innovation by i2C Innovation Center @ Vienna University of Technology

In the end we got the opportunity to pitch our idea in 3 minutes to a jury of investors and experts and with my project StrikeSensor I won the High Potential R&D idea Award! Yeha!

Source: I2C Facebook Page

My notes and lessons learned from the Vienna University of Technology “i2c StartAcademy”

General

• Keep it simple. If you cannot explain it in a simple way you haven’t thought about it long enough. Sit down and try to get it clear in your mind.

Start-up – Business Plan Development

• “The lean start up” philosophy is the most promising approach up to date to start a business. It means you should check your hypotheses about customers needs as early as possible, find a lead customer, evaluate the scope of the problem in reality and test and co-develop your solution iteratively.
•  Startups are not smaller versions of large companies, one of their main target is to search effectively for problems, solutions, people and customers.
• Most startups fail because their first contact with the customer is too late
• Startups don’t actually need a CFO, CIO, CTO …. but they do neet a customer development team
• Try to get a customer sign a constract before the product is available
• Find out why your customers would love to do business with you, thats your value proposition.
• Find out how you can generate excitement, thats the value architecture.
• Find  out how you can earn money, thats your revenue model.
• Do not confuse product features with customer value. Find out what you get done for your customer. Find out the pains and the gains of your customer.
• The Business Model Canvas from Osterwalder is a good tool to structure your business model

Business Model Canvas for my project StrikeSensor

Pitching

• The purpose of a pitch is not only to sell a solution – but you sell your product, your technology, your passion and your team.
• A good structure for your pitch is (1) One-Liner e.g. “StrikeSensor detects labor strikes around the world in real-time” (2) Problem (3) Solution (4) Future Steps (5) Request
• Then follows the Question & Answer session, anticipate the questions prepare back-up slides!
• Important things to mention in your pitch: Whats the target customer and the size of the market? Whats special about the team, competences? What makes you different from other solutions? How can you access the target market, what is your network?
• If you don’t know the answer to a question refer to someone else in your team who knows it.
• Before the pitch: (1) Check the microphone (2) Check the clicker (3) Make eye-contact with the person moderating to get a sign to start
• Make the target clear to yourself, visualize the target in your mind, what do you want to achieve? you need to be 100% clear about that yourself.
• Make sure every part of your pitch, every slides is directed towards that goal.
• Show facts and metrics! State the business case, how much value/money can be gained/saved by your solution?
• After each pitch, reflect: what worked what didn’t work! Make it better next time!
• On the last slide: provide your logo, your company name, your name, your contacts!

Pricing & finances

• Main costs: R&D, Labor, Marketing, Production, Sales + Personal Expenses of founders (don’t forget that you need to live from it as well!), customer acquisition, Travelling
• Income: Price * customer
• Remember that there might be a difference in time, when costs occur and when you get the revenues
• Investors can raise the equity (in return for shares) or liabilities, low equity but high reliabilities can make it harder to get money from the bank in the future
• Good tools for Financial Planning: Plan4You provided by WKO, or BACA Business planer
• Don’t state numbers too detailed that seems unrealistic and unprofessional
• the goal of the business plan is to provide a rough picture where the journey is going, and how much money you need to request from investors.
• Sources of funding in Austria
  • AWS: PreSeed: has to adress Venture Capitalists as investors, you should have an exit strategy
  • FFG: funds 70% of costs as liability, 30% is covered by yourself, focus on applied science
  • Others: Business Angels, Family Offices, Venture Capitalists, Banks, Crowd, Founders, Family

Marketing

• Start with active marketing & sales activities early!
• Useful Tools:
  • hootsuite for cross-platform posting, helps you to manage social media audiences
  • TweetDeck Monitor multiple keywords or timelines on Twitter
  • Sproutsocial Social Media Management
  • If this then that – IFTTT lets you create recipes (stored procedures) that automatically perform tasks for you, e.g. save your email attachments into dropbox, get an email when a profile pic changes..
  • getsatisfaction.com – Social Media marketing software
  • pr.co – A tool to format Press Releases
• Provide a press-kit on your website: logo, one-liner, short description of company (3-5 sentences)
• Read about Growth Hacking (marketing strategy for start-up, grow as fast as possible with low resources)

These were the most important lessons I took from last week, and at this point I want to thank all the mentors and i2c Innovation Center for providing us with that knowledge and the unique opportunity to take part in great programs like this one.

From March on I will attend a 3-semester course on Innovation, I am looking forward!

Many international brands are supplied by factories in Indonesia, consumers often do not know much about the conditions under which products are produced. In a recent study on factories in Indonesia we found that many supplier factories in Indonesia are represented on Foursquare. This indicates that the manufacturing industry is already reflected – to some extent – on social media.

In a next step, I want to analyze, whether problems with working conditions are apparent in online data. However, first of all, it is hard to define what a “problem” actually is. I decided to particularly look at labor strike events, because these are events were workers themselves stand-up in order to raise public awareness for circumstances they find problematic in form of a protest.

This diagram shows the increase of Tweets during a labour strike in an Adidas Factory on the island Batam mentioning the brand or the island.

I repeatedly observed peaks in the amount of Tweets mentioning a factory name or city at the time of the event of a labour-strike.

Some of the questions I want to address next are:

• Who are the people tweeting about strikes? (media, workers, workers unions?)
• What are the topics discussed before | during | after a strike?
• Are there certain phases that can be observed across several strike events?
• Which language(s) are used? (dialects, formal, informal)

In the middle of November I get the chance to travel to Indonesia to the International Conference on Data and Software Engineering 2014 in Bandung and I will stay about four weeks in Indonesia. I am looking forward to listen to the opinions of Indonesian researchers.

It seems there are several challenges which make it hard to verify whether and how well proposed methods actually work:

• It is expensive to obtain historic Twitter data therefore experiments can not be repeated under same conditions
• A multitude of decisions have to be taken during data collection (Which API is used?, Which keywords or filtering criteria are used? Which time period is captured?) often these decisions are not sufficiently documented which make it hard to repeat experiments and to apply the method in different settings
• Many of proposed methods require a predefined list of keywords to filter tweets (e.g. “flu”, “cough”, “H1N1″ … if you want to track a disease) however it’s not quite clear how to compile these lists, so methods rely on the ability of the researcher to define such lists and it is difficult to apply methods in a different context, e.g. countries with a different language.

Given this multitude of decisions and predefined knowledge that is required to conduct the experiments combined with the difficulty to repeat experiments for other researchers, it seems in Twitter prediction research could be at risk to be influenced by the observer-expectancy effect, which means that the researcher subconciously effects the research result.

Or as David Hand wrote, in other words:

“It is quite possible that the most interesting patterns we discover during a data mining exercise will have resulted from measurement inaccuracies, distorted samples or some other unsuspected difference between the reality of the data and our perception of it.” [8]

My colleague Amal Almansour from Kings College in London and I, we were particularly interested into the decisions made during Twitter Prediction research, and we just finished a literature survey and cricially analyzed 24 existing Twitter Prediction studies. In this study, we identified the different actors involved in the typical Twitter research process and their potential impact on the prediction method and respectively the prediction result.

This study is currently in the peer-review process, results will be stated here soon.

[1] T. Sakaki, M. Okazaki, and Y. Matsuo, “Earthquake shakes Twitter users: real-time event detection by social sensors,” Proceedings of the 19th international conference on world wide web, pp. 851-860, 2010.
[Bibtex]

@article{Sakaki2010,
abstract = {Twitter, a popular microblogging service, has received much attention recently. An important characteristic of Twitter is its real-time nature. For example, when an earthquake occurs, people make many Twitter posts (tweets) related to the earthquake, which enables detection of earthquake occurrence promptly, simply by observing the tweets. As described in this paper, we investigate the real-time interaction of events such as earthquakes in Twitter and propose an algorithm to monitor tweets and to detect a target event. To detect a target event, we devise a classifier of tweets based on features such as the keywords in a tweet, the number of words, and their context. Subsequently, we produce a probabilistic spatiotemporal model for the target event that can find the center and the trajectory of the event location. We consider each Twitter user as a sensor and apply Kalman filtering and particle filtering, which are widely used for location estimation in ubiquitous/pervasive computing. The particle filter works better than other comparable methods for estimating the centers of earthquakes and the trajectories of typhoons. As an application, we construct an earthquake reporting system in Japan. Because of the numerous earthquakes and the large number of Twitter users throughout the country, we can detect an earthquake with high probability (96\% of earthquakes of Japan Meteorological Agency (JMA) seismic intensity scale 3 or more are detected) merely by monitoring tweets. Our system detects earthquakes promptly and sends e-mails to registered users. Notification is delivered much faster than the announcements that are broadcast by the JMA.},
archivePrefix = {arXiv},
arxivId = {0808.0743v3},
author = {Sakaki, Takeshi and Okazaki, Makoto and Matsuo, Yutaka},
doi = {10.1145/1772690.1772777},
eprint = {0808.0743v3},
isbn = {9781605587998},
issn = {1605587990},
journal = {Proceedings of the 19th international conference on World wide web},
keywords = {event detection,location estimation,social sensor,twitter},
pages = {851--860},
pmid = {14716836},
title = {{Earthquake shakes Twitter users: real-time event detection by social sensors}},
url = {http://portal.acm.org/citation.cfm?doid=1772690.1772777},
year = {2010}
}

[2] X. Zhang, H. Fuehres, and P. A. Gloor, Predicting Stock Market Indicators Through Twitter “I hope it is not as bad as I fear”, 2011.
[Bibtex]

@misc{Zhang2011,
abstract = {Procedia - Social and Behavioral Sciences, 26 (2011) 55-62. doi:10.1016/j.sbspro.2011.10.562},
author = {Zhang, Xue and Fuehres, Hauke and Gloor, Peter A.},
booktitle = {Procedia - Social and Behavioral Sciences},
doi = {10.1016/j.sbspro.2011.10.562},
isbn = {18770428},
issn = {18770428},
pages = {55--62},
title = {{Predicting Stock Market Indicators Through Twitter “I hope it is not as bad as I fear”}},
volume = {26},
year = {2011}
}

[3] A. Tumasjan, T. Sprenger, P. Sandner, and I. Welpe, “Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment.,” Icwsm, pp. 178-185, 2010.
[Bibtex]

@article{Tumasjan2010,
abstract = {Twitter is a microblogging website where users read and write millions of short messages on a variety of topics every day. This study uses the context of the German federal election to investigate whether Twitter is used as a forum for political deliberation and whether online messages on Twitter validly mirror offline political sentiment. Using LIWC text analysis software, we conducted a content analysis of over 100,000 messages containing a reference to either a political party or a politician. Our results show that Twitter is indeed used extensively for political deliberation. We find that the mere number of messages mentioning a party reflects the election result. Moreover, joint mentions of two parties are in line with real world political ties and coalitions. An analysis of the tweets’ political sentiment demonstrates close correspondence to the parties' and politicians’ political positions indicating that the content of Twitter messages plausibly reflects the offline political landscape. We discuss the use of microblogging message content as a valid indicator of political sentiment and derive suggestions for further research.},
author = {Tumasjan, Andranik and Sprenger, To and Sandner, Pg and Welpe, Im},
doi = {10.1074/jbc.M501708200},
isbn = {0894439310386},
issn = {00219258},
journal = {ICWSM},
keywords = {Twitter,data mining,elections,microblogging,politics,sentiment analysis},
pages = {178--185},
pmid = {16046402},
title = {{Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment.}},
url = {http://www.aaai.org/ocs/index.php/ICWSM/ICWSM10/paper/viewFile/1441/1852},
year = {2010}
}

[4] A. Signorini and A. M. Segreldots, “Using Twitter to Estimate H1N1 Influenza Activity,” 9th annual conference of the \ldots, 2010.
[Bibtex]

@article{Signorini2010,
abstract = {Objective This paper describes a system that uses Twitter to estimate influenza-like illness levels by geographic region. Background Twitter is a free social networking and micro- blogging service that enables its millions of users to send and read each other's “tweets, ” or short ...},
author = {Signorini, A and Segre\ldots, A M},
journal = {9th Annual Conference of the \ldots},
title = {{Using Twitter to Estimate H1N1 Influenza Activity}},
url = {http://www.cs.uiowa.edu/~asignori/papers/using-twitter-to-estimate-H1N1-activity.pdf$\backslash$npapers://d0a46af5-98a1-4365-adb7-c8ea03c45bf3/Paper/p10221},
year = {2010}
}

[5] S. Asur and B. Huberman, “Predicting the future with social media,” in \ldots agent technology (wi-iat), 2010 ieee \ldots, 2010, pp. 492-499.
[Bibtex]

@inproceedings{Asur2010,
abstract = {In recent years, social media has become ubiquitous and important for social networking and content sharing. And yet, the content that is generated from these websites remains largely untapped. In this paper, we demonstrate how social media content can be used to predict real-world outcomes. In particular, we use the chatter from Twitter.com to forecast box-office revenues for movies. We show that a simple model built from the rate at which tweets are created about particular topics can outperform market-based predictors. We further demonstrate how sentiments extracted from Twitter can be further utilized to improve the forecasting power of social media.},
archivePrefix = {arXiv},
arxivId = {arXiv:1003.5699v1},
author = {Asur, Sitaram and Huberman, BA},
booktitle = {\ldots Agent Technology (WI-IAT), 2010 IEEE \ldots},
doi = {10.1109/WI-IAT.2010.63},
eprint = {arXiv:1003.5699v1},
isbn = {978-1-4244-8482-9},
issn = {03062619},
pages = {492--499},
title = {{Predicting the future with social media}},
url = {http://ieeexplore.ieee.org/xpls/abs\_all.jsp?arnumber=5616710},
year = {2010}
}

[6] A. Jungherr, P. Jürgens, and H. Schoen, “Why the pirate party won the german election of 2009 or the trouble with predictions: a response to tumasjan, a., sprenger, to, sander, pg, & welpe, im ?predicting elections with twitter: what 140 characters reveal about political sentiment?,” Social science computer review, vol. 30, iss. 2, pp. 229-234, 2012.
[Bibtex]

[7] P. T. Metaxas, E. Mustafaraj, and D. Gayo-Avello, “How (Not) to predict elections,” in Proceedings – 2011 ieee international conference on privacy, security, risk and trust and ieee international conference on social computing, passat/socialcom 2011, 2011, pp. 165-171.
[Bibtex]

@inproceedings{Metaxas2011,
abstract = {Using social media for political discourse is increasingly becoming common practice, especially around election time. Arguably, one of the most interesting aspects of this trend is the possibility of ''pulsing'' the public's opinion in near real-time and, thus, it has attracted the interest of many researchers as well as news organizations. Recently, it has been reported that predicting electoral outcomes from social media data is feasible, in fact it is quite simple to compute. Positive results have been reported in a few occasions, but without an analysis on what principle enables them. This, however, should be surprising given the significant differences in the demographics between likely voters and users of online social networks. This work aims to test the predictive power of social media metrics against several Senate races of the two recent US Congressional elections. We review the findings of other researchers and we try to duplicate their findings both in terms of data volume and sentiment analysis. Our research aim is to shed light on why predictions of electoral (or other social events) using social media might or might not be feasible. In this paper, we offer two conclusions and a proposal: First, we find that electoral predictions using the published research methods on Twitter data are not better than chance. Second, we reveal some major challenges that limit the predictability of election results through data from social media. We propose a set of standards that any theory aiming to predict elections (or other social events) using social media should follow.},
author = {Metaxas, Panagiotis T. and Mustafaraj, Eni and Gayo-Avello, Daniel},
booktitle = {Proceedings - 2011 IEEE International Conference on Privacy, Security, Risk and Trust and IEEE International Conference on Social Computing, PASSAT/SocialCom 2011},
doi = {10.1109/PASSAT/SocialCom.2011.98},
isbn = {9780769545783},
issn = {1457719312},
pages = {165--171},
title = {{How (Not) to predict elections}},
year = {2011}
}

[8] D. J. Hand, “Principles of data mining,” in Drug safety, 2007, pp. 621-622.
[Bibtex]

@inproceedings{Hand2007,
abstract = {Data mining is the discovery of interesting, unexpected or valuable structures in large datasets. As such, it has two rather different aspects. One of these concerns large-scale, 'global' structures, and the aim is to model the shapes, or features of the shapes, of distributions. The other concerns small-scale, 'local' structures, and the aim is to detect these anomalies and decide if they are real or chance occurrences. In the context of signal detection in the pharmaceutical sector, most interest lies in the second of the above two aspects; however, signal detection occurs relative to an assumed background model, therefore, some discussion of the first aspect is also necessary. This paper gives a lightning overview of data mining and its relation to statistics, with particular emphasis on tools for the detection of adverse drug reactions.},
author = {Hand, David J.},
booktitle = {Drug Safety},
doi = {10.2165/00002018-200730070-00010},
isbn = {026208290X},
issn = {01145916},
pages = {621--622},
pmid = {17604416},
title = {{Principles of data mining}},
volume = {30},
year = {2007}
}

In this study, we explored how social and semantic data can be used to monitor risks around supplier factories. We focused our study on Indonesia, as it exhibits both an important position as an outsourcing country for several major brands as well as a high social media usage.

Data sample

We compiled a sample of 139 factories in Indonesia supplying 4 popular companies in the textile, sports and electronics industry. Each factory is described by its name and its address. All data was retrieved from the respective company website.

Main research question

1. Can user-generated data help to determine the physical location (GPS-coordinates) of supplier factories?
2. How could we link semantic data to attain risk information about supplier factories?

The most interesting facts and results

1. Mapping Services could map only few factory addresses

Using Google Maps, Nokia Here Maps, Bing Maps, Open Street Maps (Nominatim) to transform the address information into GPS-coordinates we could only retrieve accurate GPS-coordinates for few (20/139) factories. There were considerable differences in the number of addresses which could be transformed to GPS coordinates, and precision levels.

2.Most of the factories in our sample have a Foursquare profile

For most of the factories (122/139) we could find a profile on the geo-social network “Foursquare”. Foursquare profiles are created by users, those might in this case be workers or people living around the production site.
Typically users register a location with its name and purpose using mobile devices. Thereby maps are created collectively.

3.Most of the factories were tagged on Wikimapia

Most of the factories (94/139) were tagged by users on the crowdsourced map “Wikimapia”. On Wikimapia users can tag buildings with their names or purpose on satellite pictures, thereby they create maps.

4. Factory locations in Indonesia can be best determined using User-Generated Information

By comparing the returned GPS-coordinates with a manually compiled groundtruth dataset, we found that User-generated Information (Foursquare, Wikimapia) can provide more accurate location results than geo-coding services (Google, Nokia). The first diagram below shows for how many factories we retrieved a result, wheras the second one shows how many of these were correct. Correct means within 1km distance of our ground-truth data set.

5. User-Generated information is not 100% reliable

The biggest advantage of user-generated information is at the same time also its biggest disadvantage – everyone can contribute information. Allthough collective data collection might result in huge resources, it is hardly validated resulting in lower data quality. During our reserach we often found multiple profiles for the same factory, or multiple venues at one location. Therefore Information Credibility is certainly an issue here.

Further Information

If you want to read more about it please take a look at the whole paper, or send me an email.

The publication has been developed jointly with my colleagues at the Information & Software Engineering Group @ Vienna University of Technology:

Madlberger, L., Hobel H., Thöni, A., Tjoa, A.M. Analysing Supplier Locations Using Social and Semantic Data: A Case Study Based on Indonesian Factories 12th International Conference on Knowledge Management and Knowledge Technologies, 2014 Download (pdf)