Which category best describes your project?
- [x] Unleash data
Which Fundamental Metric best describes your project?
- Data Consume Volume
Proposal in one sentence: This project aims to interview prospective buyers of a crowdsourced exercise quality dataset in a market feasibility study to understand their needs and willingness to pay for datasets more deeply.
Description of the project and what problem is it solving:
Description of the project
Our first proposal focused on unlocking data publishers using a crowdsourcing app. This is proceeding ahead of schedule and on budget. Also, by mapping out the various customers and providers in the motion analysis community, we realised the potential for approaching intermediaries as data consumers. The goal of this follow-up proposal is to understand the data consumer more deeply in terms of their needs and willingness to pay for datasets. Stronger buyer interest and more dataset purchases will increase the value and trust in the crowdsourced dataset.
The actors in the ML applications-level ecosystem can be broken down into:
- Data providers (e.g. physios and sports clubs)
- Data customers/model providers (e.g. AI companies in the motion analysis space)
- Model customers/app providers (e.g. Emovi, a medical device company)
- App customers (e.g. physios and sports clubs).
This looks like a feedback loop (at the ML applications level of the Ocean ecosystem).
Previous market research illustrated the need for data privacy and GDPR compliance (aside from accuracy, usability and affordability), which is a primary focus of Ocean. Furthermore, it focused on (4.) the sale of a final consumer app to end users such as physios, sports clubs and medical practices, rather than (2.) the sale of data to other startups and companies. In future proposals, there is also the possibility of (3.) selling trained models to other startups and companies. While many of these actors would be direct competitors within a Web2 approach, we believe that a Web3 marketplace and infrastructure for tracking provenance of data, models and other IP can encourage collaboration and progress towards the shared goal of understanding human movement.
During our previous experience in the space, we have spoken to and developed relationships with a large network of companies developing AI products for human motion analysis. These include startups, well-established companies and consultancies and the methods they are currently using to acquire and label data vary. The AI products they are building include end user apps or model APIs and SDKs.
During this project, a number of interviews will be performed with these potential customers and they will be segmented according to their needs and willingness to pay for the dataset. The value proposition and market feasibility of the new Web3 approach will be further developed and evaluated, and the data accuracy and GDPR requirements will be analysed with a different group of customers. Deeper understanding of the buyer perspective will also help ongoing Token Engineering efforts to better model the OCEAN ecosystem. For example, the insights gained can be used to validate the characteristics of data buyers for future token simulations.
What problem is it solving
- Lack of customers for data
The demand-side for Ocean datasets is less well developed than the supply side. Previously, we aimed to understand the value proposition for data publishers and providers of motion datasets. This time, we will identify prospective data consumers from our network of startups and companies in the space.
- Lack of understanding about incentives for potential data consumers
Our previous market research focused on an end user app for physios. However, it’s currently unclear which businesses have the most to gain from consuming data in a data marketplace infrastructure. In the current proposal, we will perform an improved market feasibility study to speak with prospective customers and understand in this specific case why they would (or wouldn’t) purchase data from Ocean. This will activate a new community of customers into the data economy, unlocking customers of associated data and models.
- New startups in the space need to build a lot of the stack from scratch
The model and app providers would typically be considered direct competitors in a Web2 ecosystem. We hope that a marketplace and infrastructure for tracking provenance of data, models and other IP can encourage collaboration and progress towards the shared goal of understanding human movement.
Since tools for motion analysis using deep learning have recently reached maturity, there are many startups and established companies busy in the space. As is typical in AI, many startups aim to build the full stack, including acquiring the dataset, processing the data, developing models and designing an app for the end user. However, the skills needed for these steps differ greatly, meaning that a number of very different roles such as technician, machine learning engineer and back-end/front-end developers must be hired for every team. Assets may become siloed within entities that do not have the full set of skills (e.g. data silos in universities, hospitals). Also, work such as time-consuming data acquisition may be duplicated across competing startups. The reason for these inefficiencies is likely due to the lack of mechanisms for trust and collaboration. A more efficient approach might be to create increasingly-specialised companies, devoted to working on a single part of the stack. We believe that the Ocean marketplaces can facilitate this.
What is the final product?
- A shortlist of 10 prospective data buyers where we have confirmed interest during an interview
- 2 letters of intent from potential buyers
Funding Requested: 32,000 OCEAN
If your proposal is voted to receive a OceanDAO Grant, how would the proposal contribute a value greater than the grant amount back to the Ocean Ecosystem (best expressed as “Expected ROI”)?
From the shortlist of 10 prospective data buyers in the motion analysis space, we assume a 20% conversion rate. From the additional 2 LOIs, we assume a 50% conversion rate. This would lead to 2+1=3 new buyers.
Firstly, we consider the purchase value of the dataset. We assume that they buy initially for 10,000 $OCEAN and then update on a quarterly basis for 4,000 $OCEAN. This is 22,000 $OCEAN in one year per buyer, giving a total of 22,000 * 3 = 66,000 $OCEAN.
Secondly, we believe that activating buyers for our crowdsourced dataset will build trust in the quality of the dataset and attract new stakers. This will lead to higher amount of captured $OCEAN (hence an increase in OCEAN price value). We assumed the TVL for this dataset to be 100,000 $OCEAN over one year, which was already accounted for in the ROI of that proposal. We now assume that each buyer increases the number of stakers by 10%. This results in a new TVL of 194,871 $OCEAN, an increase of 94,871 $OCEAN.
bang 66,000 + 94,871 = 160,871 $OCEAN
buck = 32,000 $OCEAN
(% chance of success) = 80%
ROI = 204,871 $OCEAN / 32,000 $OCEAN * 0.8 = 4.0
This is above the expected ROI of 1.0.
Proposal Wallet Address (*mandatory): 0x36f741F4808a329C9E876F551Bcf337B7dDc54Ff
Team Website (if applicable): https://visiotherapy.netlify.app/
Project lead Contact Email: firstname.lastname@example.org
Twitter Handle (if applicable): richardblythman
Discord Handle (if applicable): richardblythman | VisioTherapy
Country of Residence: Ireland
Have you previously received an OceanDAO Grant (Y/N)? Yes, Round 6.
Project Deliverables – Category
- 1 blog post will be published illustrating the community opportunities to increase activation
- Shortlist of 10 prospective data buyers where we have confirmed interest during an interview
- 2 letters of intent (LOIs) with potential customers
- Input to the token engineering simulation (modelling the influence of buyers)
Project Deliverables - Roadmap
- Any prior work completed thus far?
This project builds on a number of previous OceanDAO-funded projects:
- VisioTherapy: Building an exercise quality dataset using a community of physiotherapists at professional rugby and sports clubs
- Go to Market Analysis (GTM)
The VisioTherapy R6 proposal focused on (i) building an app to crowdsource domain knowledge from physios and athletes (60%) with the rest for (ii) business model development to map out the various customers and providers in the motion analysis community (20%), (iii) blog and video based on the app targeted at physios, tutorials for uploading/labelling images (20%). Based on feedback from the OceanDAO voting, we also created a simple website (here, also see our Figma design here) . The proposal is proceeding ahead of schedule and on budget.
VisioTherapy are working with DataUnion to adapt their crowdsourcing app towards our use case. The first step was to modify the code to split the design and content of the app. We could then start to rebrand the app by applying the VisioTherapy colours and logo (see the website Figma design above). We also started work on the upload screen for video data, which is different to that used by DataUnion for uploading images. The APIs in the backend have been created to upload and download videos but in the frontend of the mobile app the screens are not finished. You can watch a simple video to demonstrate the work completed on the app (here).
Outreach with blog/video/tutorials
This component is yet to be completed when the app is ready and can be put into the hands of athletes and physios.
Business Model Development
We have been working with Mark to develop a Web3 business model for our application. For example, the various agents in the movement analysis community have been mapped out, which has been beneficial for the current proposal. I am also taking the Token Engineering Academy course where we have been implementing simulations using Balancer and TokenSPICE. These simulations are useful for simulating various business models in terms of data publishers, consumers, stakers etc.
The GTM project was the first to focus on identifying data buyers, understanding their needs and investigating the potential to increase $OCEAN and data value through buyer activity. A number of insights have been developed. For example, the importance of customer segmentation has been highlighted. Different categories of buyers have different needs and buy for different reasons. Also, the importance of targeting intermediaries has been pointed out. This fits well with the current proposal.
- What is the project roadmap? That is: what are key milestones, and the target date for each milestone.
Jul 12 – 19: Design interview questionnaire and finalise list of interviewees
Jul 19 – Sep 19: Reach out and perform interviews with prospective data buyers
Sep 19 – Oct 10: Analyse interviews: customer segmentation, determine market feasibility, assess data accuracy and GDPR requirements
Oct 10 – Oct 17: Finalise shortlist, acquire LOIs and write blog post
- Please include the milestone: publish an article/tutorial explaining your project as part of the grant (eg medium, etc).
A blog post will be published for this project describing our ecosystem and illustrating the community opportunities to increase activation.
- Please include the team’s future plans and intentions.
One of our long-term goals is to use knowledge of real-world buyers as an input to Token Engineering. While these simulations are useful for simulating the ecosystem in terms of data publishers, consumers, stakers etc. the results are dependent on the input variables to the model. Thus, the insights gained here can be used to validate the characteristics of data buyers for future token simulations. Furthermore, future data consumption statistics for our use case can be used to validate the simulations.
In addition to labelled videos, we also believe that siloed MoCap datasets are valuable. We have close relationships with a number of universities, sports clubs and hospitals that have large MoCap datasets of exercises for physiotherapy applications. While generally happy to share, there are a number of hurdles that slow down the process. Firstly, ethical approval needs to be sought from a panel to share the data with each new group of researchers. Secondly, licensing arrangements need to be discussed every time with each new customer. These can take a long time. In future, we plan to establish the value that the Ocean ecosystem can give to these data providers, as well as potential consumers of such data.
As described previously, there is also the possibility of (3.) selling trained models to other companies. Potential customers include startups creating apps for physiotherapy, fitness or sports as well as major fitness product companies such as Mirror. In future, we plan to perform another market feasibility study to establish the needs of these prospective model buyers and train a set of models on our crowdsourced dataset.
Dr. Richard Blythman
- Role: Machine Learning Engineer, Biomechanical Engineer at VisioTherapy
- Relevant Credentials (e.g.):
- Video Intelligence Researcher at Huawei Technologies
- Research Fellow (Computer Science), Trinity College Dublin
- Machine Learning R&D Engineer at FotoNation, Xperi
- Role: Full Stack Developer, Lead Physiotherapist at VisioTherapy
- Relevant Credentials (e.g.):
- CORU Registered Chartered Physiotherapist
- MSc Computer Science, University of Bath
- Founder Oak Digital Health
- UG Cert Innovation and Entrepreneurship(edited)
Dr. Mark Siebert
- Data publishing (10yrs)
- Business Development for Data Markets
- Owning and driving global executive engagements and partnerships, Data and Open Science
- Web3 experience: < 1 Year
- Positioning businesses in emerging markets or innovative fields with focus on data and AI-driven solutions
- Role: DataUnion/OceanDAO proposal onboarder
- Head of Ocean Protocol Ambassador program
- CEO at DataUnion
- ML/Web3/DataUnion strategy at deltaDAO
- Machine learning (10yrs)
- Web3 (4yrs)
- Working on a bottom up approach to bring DataUnions to as many verticals as possible to learn and adopt the concept to their needs