Definitions

• Open Data is data that is freely available to everyone to use and republish without legal or other restrictions. The most important sources of open data are open science data connected to scientific activities that allow the replication of scientific achievements. In Europe, the re-use of public sector information, in other jurisdictions, freedom of information regulations make various public institutions’ and taxpayer funded datasets available for reuse. Open data is a very important source of information for business, scientific and policy uses.
• Reproducible research: The quality control of open data is focusing on reviewable, reproducible and confirmable findings. Auditability is a requirement in most high-level business, scientific or policy applications.
• Open Source: In most cases, when the data processing code and procedure is not a well-documented, open-source algorithm, reproducibility and confirmability is limited, or impossible.
• Press ⯅ to go back to the opening slide or ⯈ to jump to the first topic.

Incentives

• Data analytics usually requires one-time or flat investment and has exponentially growing benefits vs Policy/management consulting is a linear business and there is drive to sell more expert hours / days.
• Even very experienced (and expensive!) senior data scientist and engineers cannot exactly forecast the number of hours needed to solve particular problems in the production.
• Data scientists do not create data tables, visualizations and models for a project, but code, that can be re-used any number of times. Their value added is not related to a single project, but over many repeated projects.
• Because data science cannot be treated as a profit center in the management consulting, billable hours model, it is either a cost-center or an outsourced activity.

Monetary incentives for open-source software development

• Connected to a ⯈ ⯈research product for the UvA IVIR we created an ↗ open-source software that handles this problem. The labor input in hours was more than the entire research projec, but the results can be used in many-many further research products.
• Our solution, that left kept the software open-source allowed very high quality - it is a peer-reviewed scientific project that meets the highest documentation standards -, but its development costs is not covered by any consulting project fee or scientific grant. It found about 800 users globally in the first month, and we are just planning a new release with a PR campaign via ↗ rOpenGov that will bring thousands of new users.
• We would like to pilot a hybrid consulting-open source model where we are not selling software, but data-as-service, where we make sure that the software is faultless, and gains new features on demand, and this work is remunerated from the use of the data. For example, ↗ Github, where we store the software code and documentation offers donated funding option for software products. Ecorys could be a sponsor of this software development.
• The approach can be fine-tuned with less or more permissive software licenses, and exclusivity of certain exploitation as long as it bridges the needs of remuneration for software development and documentation and the business case for exploitation - for example, exclusive client support for Ecorys, first exploitation of new uses for new research products, etc.

Monetary incentives for data-as-service

• The creation of an application like our example requires very significant investment, however, this application can be very easily modified, adopted to new data sources.
• A project-based remuneration would lead to under-investment. Various options: a framework contract, intent for next project, fix retainers, or some claw-in to later uses should be explored.

Monetary incentives for reproducible research products

Pre-sales and after-sales can be incentivized with success fees, or fixed fees for pre-defined efforts. The highest ROI is expected on this element, because reproducible research techniques aim for effortless re-use.

• After the competition of a complex research product, we can make sure that all its data tables, visualization, int-text references for calculated model values, and citations are constantly maintained.

• We can do this against a small maintenance fee, or an appropriate success fee in case a re-cast or re-use of the research product or its components is achieved with a paying client.

• We can even create a small subset of research products on a monthly or quarterly basis in the form of a policy journal, newsletter or corporate blogposts to maintain interest in the topic that was already covered in a paid project.

Data Observatories: Big Data & Small Organizations

• Permanent data collection: in social and natural sciences alike, many scientific discoveries, hypothesis testing, or scientific proofs require consistent data collected over a longer period. Only the largest, usually public institutions have the organizational capacity and budget to organize such a data collection program.

• Funding cooperation: a long-term data collection program has many advantages for all scientific, policy or business uses, and offers many cost savings, but requires eventually some basic funding. Almost all the data observatories that we have reviewed receive some sort of public funding, and the ones that ceased to exist usually stopped their data collection program because of the availability of further public funds. Nevertheless, some sort of co-funding from participants or users is usually present.

• Better value for money: the 17-years old European open data regime recognized that the value for money in most of the data investments can be significantly improved by data sharing and reuse. Experience tells that most publicly collected datasets are only exploited in a small fraction for the primary, first use, but they can provide value for businesses, researchers or the public sector when reused.

• Organization: In social or economics sciences, often an hoc large cross-sectional data collection, such as an international comparative data collection, requires significant organizational investments, and this is even more the invest into longitudinal data collection that repeats regularly or irregularly over time. A permanent observatory structure, as an institution or a partnership of institutions is necessary for complex, longitudinal data collection. This can be provide by a multi-disciplinary team of domain experts, statisticians, data scientists, engineers and computer scientists.

Cooperation Proposal

The open collaboration leaves project sponsors with a flexible reserve of developers and researchers, as well as data capacity that they do not have pay for, even on a retainer basis. Financing the core of the collaboration, for example, in the form of sponsoring open-source specialist software or the maintenance of permanent data collection platforms, a high level of service level and exclusivity can be achieved without bringing basic research costs in house.

1. Open-source software development: ↗ Github Sponsorship or other forms of support for maintenance and further development. Available for sponsoring: ↗ regions - sub-national statistical indicators, 342 average monthly downloads, ↗ retroharmonize - individual survey data harmonization, 309 monthly downloads or ↗ iotables economic and environmental impact assessment, multipliers, direct and indirect effects for all EU countries, 482 monthly users. software packages; atypical licensing/support agreements for first exploitation, exclusive servicing.
2. Open data observatories: Co-founding or sponsoring open data observatories, with opportunities for early, embargoed exploitation of new, valuable data assets. We have many data assets for a ↗ Music, ↗ Climate Policy Data Observatory and other data observatories, and planned Digital Media, Culture Heritage EU projects, among others. Participation in observatories gives access to more data, and opens data sharing and consultancy sales opportunities within the domain-specific data ecoysytem of the observatory.
3. Research contributions: on a project-basis, with proportional consulting day basis.
4. Pre- and after sales: success fee or flat fee basis.

Dilemmas of in-house, exclusive or open collaboration ⯆ below

Open Collaboration & Exclusive Services

Open collaboration is an agile project management method originating in open-source software development and reproducible scientific practice. The open collaboration leaves the project sponsors with a flexible reserve developer/researcher and data capacity that they do not have pay for even on a retainer basis. It is built on - modularization: split up the tasks into independent components, - information commons that are well structured and make contributions easy, - incentives to participate, even with small contributions on an individual level.

• While more and more activities are coded and automated, software development is an activity that is best kept outside of a consultancy due to product liability and business model consideration. This element is best kept open-source.

• Data collection is ongoing and expensive, and a hybrid model that encourages exploitation of open data, data sharing, data altruism is best combined with exclusive, early access rights. Our triangular observatory concept allows the utilization of scientific, policy and business research assets to be combined, because these are usually these are non-competing interests.

• Consultancy is not about data but insight. Any work that requires exclusive access to the insights of data analysts, statisticians, geographers, data scientists, can be internalized on a project (external consultant) or internal (part- or full-time employment) basis. Probably the best value for money is an external service contract with certain levels of non-competition clauses.

• Pre-sales and after-sales can be incentivized with success fees, or fixed fees for pre-defined efforts. The highest ROI is expected on this element, because reproducible research techniques aim for effortless re-use.