I had just finished a short read about the Theory of Change in a PDF from the UN Development Group. To simplify, a Theory of Change (TOC) says “if we do X, then we change Y”.
Let’s say you want to reduce the number of people end up back in jail after being released. Your TOC might be “if we find jobs for people who were incarcerated, then we can reduce the number of them that go back to jail”
What was fascinating to me was how close to the scientific method this was. You have a theory, you try something, and based on the results, you prove or disprove the theory. And since this was to be applied to non-profits, it would work the same way that market forces work in the for-profit world.
But I chalked the whole idea of a theory of change into the “nice idea in a perfect world” and tossed it into my backlog of ideas.
Then, one day, in the middle of a shower, it bubbled back into my head. After slip-sliding my way to my notebook and furiously drying my hands, I managed to sketch out an idea.
So here it is, a Theory of Change based business idea.
Value Proposition / Business Model
Donating to a non-profit is seen as a ‘good thing’. But sometimes, doing a ‘good thing’ doesn’t result in a better world. And other times, the smallest act can have massive ramifications.
But what if you could say, within a reasonable margin of error, that the things you’ve done have caused a measurable change in the world?
By using public data sources, TOChanger can empirically demonstrate the impact of a non-profit. TOChanger will use machine learning to identify not just the intended effects, but the also the unintended ones.
Non-profits will be ranked and scored based on their verifiable impact. You can choose to donate directly to one of them, choose which group you want to help, or leave it up to us to direct the funds where we think you’ll make the highest impact.
To ensure impartiality, TOChanger will be funded by a fixed percentage of the incoming donations. TOChanger will also avoid any external funding that may impact it’s impartiality or potentially endanger the data of those being served by the NGOs and non-profits.
People, Process, and Technology
Let’s do this in reverse order.
Using the MVC framework, the tech stack might look like:
Model — Big, unstructured, non-RDBMS, secure, cloud based
View — Data input for orgs seeking funding, and another for funders, dashboards for orgs and funders, and an analysis suite for the data.
Controller — CRUD for batch and individual records (heavily permissioned). Many cloud DBs have these as a feature.
The core of the stack will use machine learning (ML) to find the correlations between actions and results. To start, this stack will need as much base-level demographic information as possible. Taking the pre-existing public databases of populations will provide base-level demographic data. The US Census is a good example of this kind of public data, but other data sources exist. Note that the main reason for using unstructured data is because many of these data sources are not standardized.
With that base level data, subdivisions can be constructed by the variables in the TOC. If the TOC focuses on a cohort that isn’t covered by this base-level data, the information must be provided by the funded orgs. For example, if the TOC is about low income children, the base-level data from the US census should suffice. But if the TOC focuses on students with learning disabilities, the funded org will need to provide that information or it will need to be gathered from alternate sources. The ML analysis will have to factor in the size of the target cohort vs the total cohort to prove/disprove the TOC.
Finally, after the inital TOC is tested, other variables of interest will be tested. It may be that the non-profits impact is greater in a different area than intended. This allows the discovery and highlighting of outsized impacts.
- Org applies for funding, clearly outlining their TOC
- Their TOC is tested against the base-level data. If it is validated, they are listed as a potential target. If not, they are informed what interventions are more tightly correlated with their TOC.
- If the TOC is inconclusive, the org is given an interface into the tech stack to provide their data. There will be requirements for the kind of data that must be provided, but the stack will allow for various forms of ingestion via API or other options.
- After one funding cycle, the inconclusive orgs are then reevaluated. If they remain inconclusive, they will receive guidance and coaching towards changing their interventions based on current findings.
- At the end of each funding cycle, all the orgs are reevaluated, scored, and the most impactful orgs are highlighted to the funders.
- Create a profile and connect a payment method
- Identify which areas you want to make an impact
- Define what percentage of your donations should be dedicated to each area
- The platform will provide suggested orgs and their information
- You can then accept or reject the suggestions
- Every fundraising cycle, the most impactful funders will be highlighted alongside the most impactful orgs
The workforce will need the following skills
- Data analysis for the core of the stack
- Software development for the interface into the stack. This will include three views: funders, funded, and inconclusive.
- Soft skills, especially coaching and empathy for non-profits
Minimizing the customer support roles can be done by improving the interface, but staff in this role can be cross-trained to provide support to inconclusive orgs. If the staff are also cross trained in data analysis and UI/UX, then headcount can shift between roles. This keeps operational costs down and improves employee engagement, simultaneously.
For example, a customer support rep could raise the issue of a particular data entry field being problematic for that type of NGO. That rep, leveraging their experience, weighs in on the UX and then the UI. With a little additional training, that same rep could develop and test the new solution. Finally, that rep could reach out to the NGO that triggered this issue, and work with them to perform user tests.
Scale and Scope
So far this model has focuses on crowdfunding, but there exists the opportunity to work with local and state governments on this sort of work. That said, this sort of scaling effort will require substantial effort (as anyone who’s working in fundraising will tell you)
Further, there are international aid organizations that would benefit from the ML core of the stack for their own internal audits.
So, that’s my idea. There are a lot of gaps and holes and missing details but nothing that can’t be addressed with a little time and focus. If you’d like to run with this, be my guest and reach out if you have any questions.