The unsexy and essential work of Building an Evidence-Based Org Four lessons we learned while building an evidence-based organization for migrant labor market integration in Spain
TL;DR
Eight months ago, we set out to build PROPIA, an organization to help Latin American migrants in Spain break out of precarious, low-quality jobs. But we made a deliberate choice: we would not launch until we had done the unsexy, foundational work.
No glossy website. No public programs. No premature scaling.
Instead, we buried ourselves in data, built a rigorous Theory of Change, mapped every possible failure point, and ran prototypes with the people we hoped to serve. We were incubating within the Theory of Change Makers program, and we forced ourselves to follow a disciplined sequence: first understand the problem, then find an intervention with evidence, then test our assumptions before committing.
Today, after eight months of that work, PROPIA is finally ready to launch. We’re not starting from zero, we’re starting from a place of clarity.
Looking back, four lessons stand out. They aren’t groundbreaking, but they were surprisingly easy to neglect in practice.
1. Start with a problem that is tractable, not just urgent
The labor exclusion of migrants in Spain is severe. In Catalonia, 1 in 5 residents is foreign-born, and the wage gap reaches 29%. Over half of foreign-born university graduates work in jobs that are below their qualifications. The precariousness is real.
But urgency alone doesn’t justify a new organization. Our first real challenge was moving past the emotional pull of “helping migrants” and defining a problem we could actually solve.
We narrowed our focus to a specific population: Latin American migrants in Spain whose skills are systematically underused due to structural barriers—non-recognition of credentials, employer distrust, lack of local networks, and language hurdles. That gave us a tractable question: How can we create a credible signal of their capabilities and reduce the risk employers perceive?
It felt slow. But it meant we entered the design phase with clarity, not just good intentions.
2. Borrow mechanisms, not just inspiration
Our natural instinct was to invent a brilliant new program. Instead, we anchored ourselves in what already had empirical support.
We found a rigorous RCT from Sweden (Dahlberg et al., 2023) testing an early labor-market integration program for refugees. The result: it nearly doubled employment rates in the first year. The mechanisms weren’t exotic: intensive language training, supervised work experience, targeted job placement support.
We also drew inspiration from Malengo, GiveWell and Coefficient Giving funded, a program that demonstrates how structured mentorship, financial support, and a clear pathway to economic mobility can create durable outcomes, particularly in contexts where migrants face systemic barriers. Their model reinforced the importance of combining high-touch support with a focus on long-term income gains.
Our “innovation” wasn’t inventing from scratch; it was adapting proven mechanisms (from both Sweden and Malengo) to a new context (Spain), a new population (Latin American migrants), and a new set of constraints (a complex administrative system, the added layer of Catalan language, and the prevalence of informal job networks). This shifted our mindset from “what sounds compelling?” to “what mechanisms are proven, and how do we preserve their integrity here?”
3. A Theory of Change is a tool to find failure, not only to signal coherence
We built a detailed Theory of Change, a beautiful flowchart of inputs, activities, outputs, and outcomes. It looked serious and coherent.
Its real value came when we used it to break the model. We listed dozens of critical assumptions and turned them into learning questions:
Would employers actually offer real, supervised placements, or just token tasks that don’t build credible experience?
Would the program’s intensity cause dropout among people juggling survival work and family care?
Would our contingent reimbursement model feel enabling or like a debt trap?
This exercise was uncomfortable. It made our elegant model feel fragile. But it also turned the Theory of Change from a static document into a roadmap for our prototyping phase.
4. Prototype to expose frictions, not to validate the concept
Our prototyping phase (surveys, focus groups with potential participants, and interviews with employers) was humbling. The results weren’t a clean “go” or “no-go.” They were far more useful.
We confirmed the core value proposition: access to real companies, a supervised work experience, and close support was powerfully attractive to potential participants. Employers also saw value, but for pragmatic reasons: they wanted to reduce hiring risk and coordination burden, not just do social good.
But we also found frictions. For participants, “interest” was different from “readiness”: real-life constraints like income pressure, childcare, and transportation would dictate who could actually enroll. For employers, simplicity was everything; any administrative friction would kill the collaboration.
That led us to a “PIVOT with signals of GO” decision. The model had real traction, but it needed to be simplified, segmented, and tested again with a more focused pilot before we could responsibly launch.
What the numbers tell us
As we refined the intervention, we also modeled a cost-effectiveness analysis (CEA) to ground our decisions in more than intuition. PROPIA’s CEA indicates that, under reasonable assumptions and with explicit uncertainty, the program has the potential to be a cost-effective livelihood intervention for its context. The base case (6 years of persistence, years 1–12) produces a cost of $0.34 for every $1 of additional income generated for beneficiaries, better than direct cash transfers and comparable to global benchmark programs.
Cost per beneficiary:
Year 1 (pilot) $5,024.40
Year 12 (scale-up) $2,529.48
Uncertainty remains high, especially regarding the duration of the effect. That’s why PROPIA’s internal evaluation will prioritize monitoring employment and income at 13, 36, and 60 months, replacing assumptions with our own evidence. The model is designed to be updatable: each new piece of field data can be directly incorporated into the conversion chain.
Today, PROPIA launches with that learning built in: not the original complex model, but a simpler, segmented version we’ll test with a small cohort and iterate from there.
But the problem we’re tackling is anything but small. Across Europe, millions of migrants from the Global South remain trapped in precarious work, and these flows will only grow. If we can make an evidence-based, cost-effective model work in Spain (one of Europe’s most complex contexts) it has the potential to travel.
Eight months of groundwork felt slow. But it gave us something more valuable than speed: a clear-eyed understanding of where our model can deliver impact, and where it can fail.
If you’re building in complex areas like migration, workforce development, labor mobility, or any field where implementation complexity is high, we’d love to connect. Visit our website, read about our process, or just reach out. We’re excited to finally step into the world.
PROPIA
The unsexy and essential work of Building an Evidence-Based Org
Four lessons we learned while building an evidence-based organization for migrant labor market integration in Spain
TL;DR
Eight months ago, we set out to build PROPIA, an organization to help Latin American migrants in Spain break out of precarious, low-quality jobs. But we made a deliberate choice: we would not launch until we had done the unsexy, foundational work.
No glossy website. No public programs. No premature scaling.
Instead, we buried ourselves in data, built a rigorous Theory of Change, mapped every possible failure point, and ran prototypes with the people we hoped to serve. We were incubating within the Theory of Change Makers program, and we forced ourselves to follow a disciplined sequence: first understand the problem, then find an intervention with evidence, then test our assumptions before committing.
Today, after eight months of that work, PROPIA is finally ready to launch. We’re not starting from zero, we’re starting from a place of clarity.
Looking back, four lessons stand out. They aren’t groundbreaking, but they were surprisingly easy to neglect in practice.
1. Start with a problem that is tractable, not just urgent
The labor exclusion of migrants in Spain is severe. In Catalonia, 1 in 5 residents is foreign-born, and the wage gap reaches 29%. Over half of foreign-born university graduates work in jobs that are below their qualifications. The precariousness is real.
But urgency alone doesn’t justify a new organization. Our first real challenge was moving past the emotional pull of “helping migrants” and defining a problem we could actually solve.
We narrowed our focus to a specific population: Latin American migrants in Spain whose skills are systematically underused due to structural barriers—non-recognition of credentials, employer distrust, lack of local networks, and language hurdles. That gave us a tractable question: How can we create a credible signal of their capabilities and reduce the risk employers perceive?
It felt slow. But it meant we entered the design phase with clarity, not just good intentions.
2. Borrow mechanisms, not just inspiration
Our natural instinct was to invent a brilliant new program. Instead, we anchored ourselves in what already had empirical support.
We found a rigorous RCT from Sweden (Dahlberg et al., 2023) testing an early labor-market integration program for refugees. The result: it nearly doubled employment rates in the first year. The mechanisms weren’t exotic: intensive language training, supervised work experience, targeted job placement support.
We also drew inspiration from Malengo, GiveWell and Coefficient Giving funded, a program that demonstrates how structured mentorship, financial support, and a clear pathway to economic mobility can create durable outcomes, particularly in contexts where migrants face systemic barriers. Their model reinforced the importance of combining high-touch support with a focus on long-term income gains.
Our “innovation” wasn’t inventing from scratch; it was adapting proven mechanisms (from both Sweden and Malengo) to a new context (Spain), a new population (Latin American migrants), and a new set of constraints (a complex administrative system, the added layer of Catalan language, and the prevalence of informal job networks). This shifted our mindset from “what sounds compelling?” to “what mechanisms are proven, and how do we preserve their integrity here?”
3. A Theory of Change is a tool to find failure, not only to signal coherence
We built a detailed Theory of Change, a beautiful flowchart of inputs, activities, outputs, and outcomes. It looked serious and coherent.
Its real value came when we used it to break the model. We listed dozens of critical assumptions and turned them into learning questions:
Would employers actually offer real, supervised placements, or just token tasks that don’t build credible experience?
Would the program’s intensity cause dropout among people juggling survival work and family care?
Would our contingent reimbursement model feel enabling or like a debt trap?
This exercise was uncomfortable. It made our elegant model feel fragile. But it also turned the Theory of Change from a static document into a roadmap for our prototyping phase.
4. Prototype to expose frictions, not to validate the concept
Our prototyping phase (surveys, focus groups with potential participants, and interviews with employers) was humbling. The results weren’t a clean “go” or “no-go.” They were far more useful.
We confirmed the core value proposition: access to real companies, a supervised work experience, and close support was powerfully attractive to potential participants. Employers also saw value, but for pragmatic reasons: they wanted to reduce hiring risk and coordination burden, not just do social good.
But we also found frictions. For participants, “interest” was different from “readiness”: real-life constraints like income pressure, childcare, and transportation would dictate who could actually enroll. For employers, simplicity was everything; any administrative friction would kill the collaboration.
That led us to a “PIVOT with signals of GO” decision. The model had real traction, but it needed to be simplified, segmented, and tested again with a more focused pilot before we could responsibly launch.
What the numbers tell us
As we refined the intervention, we also modeled a cost-effectiveness analysis (CEA) to ground our decisions in more than intuition. PROPIA’s CEA indicates that, under reasonable assumptions and with explicit uncertainty, the program has the potential to be a cost-effective livelihood intervention for its context. The base case (6 years of persistence, years 1–12) produces a cost of $0.34 for every $1 of additional income generated for beneficiaries, better than direct cash transfers and comparable to global benchmark programs.
Cost per beneficiary:
Year 1 (pilot) $5,024.40
Year 12 (scale-up) $2,529.48
Uncertainty remains high, especially regarding the duration of the effect. That’s why PROPIA’s internal evaluation will prioritize monitoring employment and income at 13, 36, and 60 months, replacing assumptions with our own evidence. The model is designed to be updatable: each new piece of field data can be directly incorporated into the conversion chain.
Research report: [PUBLIC] Reports—PROPIA
Cost-effectiveness model: 10.2 [PUBLIC] Cost-Effectiveness Model - (PROPIA) - [Updated 03/2026]
What comes next
Today, PROPIA launches with that learning built in: not the original complex model, but a simpler, segmented version we’ll test with a small cohort and iterate from there.
But the problem we’re tackling is anything but small. Across Europe, millions of migrants from the Global South remain trapped in precarious work, and these flows will only grow. If we can make an evidence-based, cost-effective model work in Spain (one of Europe’s most complex contexts) it has the potential to travel.
Eight months of groundwork felt slow. But it gave us something more valuable than speed: a clear-eyed understanding of where our model can deliver impact, and where it can fail.
If you’re building in complex areas like migration, workforce development, labor mobility, or any field where implementation complexity is high, we’d love to connect. Visit our website, read about our process, or just reach out. We’re excited to finally step into the world.
Daniela Betancourth—betancourthruizd@gmail.com
Alejandra Conto alejandracontoa@gmail.com