Since beginning of 2010 I have been researching on how companies can incorporate sustainability in their business operations, within the OEPI project and several SAP-internal projects with LoB sustainability. In order to improve their operational efficiency and overall sustainability, companies rely on exchanging Environmental Performance Indicators (EPIs), since internal measures are incomplete and often only display a fraction of a product’s overall emissions . In our projects we have been investigating how to design inter-organizational Information Systems to exchange EPIs, Sustainable Business Networks (SBNs). SBNs, like their big counterparts in the consumer world (Facebook, Twitter & Co.) can only be successful if they attract a high volume of contributions from their participants. Unlike in social networks, the focus of these contributions is not on entertainment but on quality. But how can this quality be monitored, disseminated, and finally improved? Could the Wisdom of the Crowd, enabled by a crowdsourcing process be a solution to this problem?
The Expert Crowd
Crowdsourcing, as it is applied in Web 2.0 applications, is defined as “the act of taking a job traditionally performed by a designated agent (usually an employee) and outsourcing it to an undefined, generally large group of people in the form of an open call. ” This definition already reveals the difference between classical crowdsourcing and expert crowdsourcing in Business Networks:
- Expert crowdsourcing in SBNs often enables new capabilities that a company did not have before and therefore only in few cases replaces jobs traditionally performed by employees.
- Expert crowdsourcing in SBNs does not address an undefined, generally large group of people but a specified group of experts. This limits the potential amount of task performers and to some extent also their diversity, but on the other hand increases their average competence and accuracy.
Some of the well-known examples of crowdsourcing in fact illustrate the potential of limiting the participation to experts: When the USS Scorpion, a US Navy nuclear submarine, went missing in 1968, the search continued without success for about 5 months. Since underwater listening systems had recorded several explosions, a radius of 20 miles for the location of the wreck could be determined. Dr. John Craven, who led the search for the U.S. Navy, finally selected a team of diverse marine experts, and made all information available to them. He then allowed the experts to bet on the parameters speed, direction, and angle of the submarine after the first explosion. The ship could finally be located only 220 yards from the position calculated by aggregation of these parameters.
In fact, in these cases, instead of talking about the “Wisdom of the Crowd”, it would be more accurate to speak about the “Wisdom of the expert group”. This fact is amplified by the requirement for efficiency in business tasks. Therefore it is logically consistent to restrict the access to business tasks to a group of experts. In the consumer world user contributions are often not paid or the monetary compensation is marginal.
Quality Aspects in Sustainable Business Networks
Quality, in any established definition, has many different aspects. The GHG Protocol, the most applied guideline for measuring carbon emissions, explicitly mentions the aspects relevance, completeness, consistency, and accuracy. Most relevant in the case of Sustainable Business Networks is the dimension of accuracy. In corporate reporting, the determination of an EPI follows a 5-step approach . In a first step the relevant sources have to be identified. Second, a calculation approach has to be determined. Third, the activity data (the data of the corresponding processes etc.) has to be collected. Fourth, the EPI can be calculated, and in a last fifth step the data can be rolled-up to corporate level. Consequently, information about each step should be provided so a significant judgment about the quality can be made. Most quality problems are related to the definition and collection of activity data. Specifically, activity data can be primary (collected from the company’s operation or supply chain), secondary (such as industry-averages or environmental databases), extrapolated (adapted from similar process) or proxy (directly transferred from similar process) data. These aspects should also be transparent when providing it to the SBN.
Connecting the Pieces
Now, how do you solve a problem with an expert team, in particular in the case of data quality in Sustainable Business Networks? Again, the approach can be divided into 5 steps, of which the steps transparency and aggregation methodology are particular important in the case of SBNs:
- Formulate the problem and problem parameters: In our case, the main problem is to ensure the accuracy of the EPI data within the SBN – parameters of the accuracy are the data source, the calculation approach, the activity data collection process, the EPI calculation, and the roll-up to corporate level.
- Select the experts: The selection of the experts is key in Sustainable Business Networks and it can be done straightforward: transaction-based, meaning that the sustainability experts consuming the data rate it afterwards – this guarantees that the voting entity also has an actual reference to the particular content.
- Make the problem and its parameters transparent: In order to make all parameters of the problem transparent, all 5 steps of the EPI calculation process should be documented in the system and all activity data should have a “tag” describing the data type.
- Let experts apply their own methodology and calculation: The experts can now easily explore the data and its parameters, and there should be a “Web2.0-style” easy and fast possibility to rate the quality of the data.
- Aggregate results: Aggregation is a complex topic of its own. The aggregation inherits many elements of reputation systems. The overall score of the rating should lead to a “reliability” score for the data owner, while the degree of accordance with other votings should be used to calculate a “credibility” score. Suitable algorithms can be based on Bayesian statistics or the Page-rank algorithm . The reliability score can later on be used in supplier evaluations, in order to provide suppliers with a suitable incentivation to submit high quality data.
In our research we show how to work out such a mechanism for data quality in SBNs in detail – and discuss other related aspects such as the matter of critical mass in SBNs.
Hans Thies, SAP Research St. Gallen