|Title: ||Systematic Lead User Identification: Industrial Case Study Framework|
|Authors: ||Pajo, Sanjin|
De Lameillieure, Maka
|Issue Date: ||Jul-2014 |
|Conference: ||12th Annual Open and User Innovation Conference location:Boston, USA date:28-30 July 2014|
|Abstract: ||In a rapidly changing marketplace companies are constantly looking to meet customer needs and stay competitive by bringing new solutions and products to the market. A small subgroup of customers, called lead users, experience needs before the rest of the marketplace and stand to benefit greatly by finding solutions to those needs. They actively engage in innovation and are a great source of new and commercially successful ideas for companies. Additionally, lead users can help diffuse new products into the marketplace due to their influential position within the customer network. To identify such a valuable human resource, companies have utilized surveying approaches like broadcasting, screening, pyramiding and crowdsourcing. Although effective, these methods are often time and resource consuming: the identification process can last up to 6 months and experts are required to analyze vast amounts of user information collected in interviews or through questionnaires.
To systematically and quickly identify lead users, an approach utilizing data mining and machine learning methods to find lead users online, called Fast Lead User IDentification (FLUID), is proposed. Social networking sites like Twitter provide vast amounts of rich data through a structured interface and are suitable target platforms. In order to train the algorithm, a set of Twitter lead users and non-lead users was collected by utilizing a surveying approach. This set acts as a stepping stone that allows us to build an early classification model. The FLUID tool will be tested on a number of industrial cases executed in close cooperation with industrial partners. For each case, a set of keywords is requested from the company to be used as query terms for a lead user search on Twitter. Each term is given a weight value that is indicative of its significance to the background or interest of the targeted lead users by the company design team. In the next step, data collection, user metadata and tweets are retrieved through the structured interface. The amount of data collected depends on the company’s specific needs. Additional filters may be applied, for example language filters, limiting the user group to a particular targeted subgroup. Thereafter, instantaneous classification is performed using the previously generated classifier model, which separates Twitter users into lead and non-lead users, and provides a confidence level score for each user's classification. This allows for the making of a ranked list of top lead users. To further improve the classification process and to tailor it to the needs of the industrial partners, the company representatives are asked to evaluate the metadata and tweets of lead and non-lead users uniformly picked from the ranked user list. For each of these users, the company indicates a novelty degree of ideas from lead and non-lead users and their interest in contacting the user for further collaboration in their product development process. The evaluated set of users is added to the training set to further improve the classification model. It is envisaged, that a number of iterations will be necessary in order to optimally rank the users according to the company’s interest. Finally, the obtained set of lead users is validated through an idea evaluation process performed together with the company design team. The approach contributes to the lead user methodology by offering a systematic and fast approach to identifying lead users through social networking sites. Additionally, authors look to further validate the methodology by identifying lead users online for a number of target fields and products.
|Publication status: ||accepted|
|KU Leuven publication type: ||ER|
|Appears in Collections:||Centre for Industrial Management / Traffic & Infrastructure|