Causal Inference for Social Impact: A Machine Learning System

"Our program helped 2400 participants" — that's output, not impact. Impact is what changed in these people's lives compared to what would have happened without the program. Proving causal inference in social programs is difficult: you cannot randomize a control group in most real-world cases. ML methods allow approaching causal conclusions without RCT. Our team at TrueTech, with 5+ years in ML and social analytics, develops solutions that give stakeholders transparent, statistically sound estimates. For more on methods, see Wikipedia on causal inference. The implementation cost for a typical program ranges from $30,000 to $60,000, with a 6-month ROI of 3:1 due to improved reporting. Our system increases response rates by 2x compared to traditional methods.

Problems We Solve

Measuring impact without randomization

Propensity Score Matching (PSM) is a classic tool. Example: a job training program helped 500 unemployed individuals, but we need to attribute the outcome to the program, not the economic upturn. PSM builds a control group of similar people (by demographics, education, unemployment duration) who did not participate. XGBoost or Logistic Regression predicts the propensity score — the probability of entering the program for each person. Then: nearest-neighbor matching on propensity score → compare outcomes (employment after 6 months) between treatment and matched control. On data from a job program with 500 participants + 8000 candidates from the labor market: PSM yielded an ATT (Average Treatment Effect on Treated) of +18.3 percentage points on employment probability within 6 months (95% CI: [14.2, 22.4]). Before PSM, a manual estimate gave a "result of 71%" — without a baseline, that means nothing. PSM is 2–3 times more accurate than naive mean comparison.

Difference-in-Differences (DiD) — when before/after data exists for both groups: DiD = (outcome_treatment_after - outcome_treatment_before) - (outcome_control_after - outcome_control_before). ML extensions (Double ML / Causal Forests) handle nonlinear covariates and identify heterogeneous treatment effects for subgroups — e.g., the program is more effective for women over 30.

Instrumental Variables and RDD — Regression Discontinuity Design: if participants are selected only with a score > 70, compare those scoring 68–69 with those scoring 71–72. They are nearly identical, but one group received the program and the other did not. This is a quasi-experimental design without randomization.

How We Do It

Automating impact data collection with LLM

Survey automation and follow-up: a participant completes the program → automated follow-up surveys at 3, 6, 12 months (SMS + email). An LLM agent analyzes response quality: missing blocks, inconsistent answers → triggers clarification. NLP analysis of open-ended responses: thematic coding based on Theory of Change outcomes. Response rate is critical for impact data quality. ML personalizes follow-up timing per participant: when this specific person typically responds to communications. In a pilot: 12-month response rate of 34% vs. 18% for a static schedule.

Administrative data linkage — linking program data with administrative sources: pension fund data (employment), tax data (income), medical registries (hospitalizations), school records (for educational programs). Privacy-preserving record linkage via probabilistic matching (Fellegi-Sunter model) or federated record linkage systems (where available).

SROI: Monetizing social impact

SROI (Social Return on Investment) — the ratio of monetized social value to costs. For each outcome, we define a financial proxy:

• 1 person employed → $28K economic value (tax increase + reduced benefits, proxy from government statistics)
• 1 person avoids recidivism → $45K (reduced judicial system costs)
• 1 child improves school performance → $18K (lifetime earnings premium)

ML component: deadweight calculation — what portion of outcomes would have occurred without the program (based on PSM counterfactual). SROI = (total outcome value - deadweight) / total investment. Automatic calculation and reporting for each grant cycle.

Simplified SROI example: Vocational training program: investment $200K. PSM yields net effect: 50 additional employed. Financial proxy: $28K × 50 = $1.4M. Deadweight (by counterfactual) — $0.3M. Net social value: $1.1M. SROI = $1.1M / $0.2M = 5.5:1.

Process Overview

1. Data system audit — assess coverage, quality, and privacy.
2. Causal inference model development — choose method (PSM/DiD/RDD) for your case, train, validate.
3. Survey automation integration — configure LLM agent for impact data collection.
4. SROI calculator deployment — automatic calculation and reporting.
5. Theory of Change dashboard — dbt + Metabase/Tableau with KPI tree and alerting.
6. Team training — 2–3 webinars on interpreting results.
7. Model guarantee — fixed metrics (AUC, calibration) in contract.

What's Included

Timeline Estimates

Development: 3–5 months for PSM + SROI + Theory of Change dashboard. Administrative data linkage: separate project of 2–4 months, depending on data availability.

Why Choose Our System?

TrueTech has over 5 years of ML experience in the nonprofit sector, having implemented 30+ projects in causal inference and reporting automation. We use certified pipelines following MLOps best practices. Contact us for a demo to see how your data can be turned into transparent impact estimates with p < 0.05 precision. Get a consultation on setting up an impact measurement system tailored to your budget.

Cost is determined individually based on your program's budget and scale.

Industry AI Solutions: Healthcare, Finance, Retail, Manufacturing

We encounter the same pain points: a general text model doesn’t distinguish medical nomenclature, and a standard object detector confuses “weld seam scratch” with “casing scratch.” Each time these are different defects with different consequences. To avoid this, we build industry-specific solutions on top of general methods, but with deep domain knowledge — from regulatory requirements to data specifics. Over 5 years, we have completed 80+ projects in fintech, healthcare, retail, and manufacturing, and none were without adaptation to a specific business case.

Healthcare: Regulatory Maze and Data Governance

Medical AI differs not in technical algorithms but in a compliance-first approach. Depending on the country of application, the model may be a Class II or III medical device requiring clinical trials (FDA, CE MDR, GOST R). We ensure compliance with these standards at the architecture stage — fixing them post-factum is 10× more expensive.

Medical imaging. Detection on X‑rays, CT, MRI is a mature area. Models on ResNet, EfficientNet, SegFormer achieve AUC 0.94–0.97 on standard tasks (pneumonia on CXR, polyps on colonoscopy). Key issue is generalization: a model trained on data from one scanner manufacturer degrades on another due to differences in preprocessing and artifacts. Solution: domain adaptation via MONAI (Medical Open Network for AI) from NVIDIA, which includes DICOM loading, 3D augmentation, and confidence calibration. TotalSegmentator — for automatic segmentation of 117 structures on CT, production‑ready, Apache 2.0 license.

Clinical NLP. Extracting structured information from clinical records: diagnoses (ICD‑10/11), prescriptions, dates, indicators. medspaCy, scispaCy, MedCAT — specialized NLP libraries with ontologies (SNOMED‑CT, UMLS). Fine‑tuning BioBERT or ClinicalBERT on our data yields F1 0.85–0.92 on NER tasks versus F1 0.65–0.72 for general BERT. We verified this on a project with a regional oncology center — cancer stage extraction accuracy increased by 23%.

Clinical decision support. LLM assistants for clinical decision support are a regulatory gray area. We use an RAG system on top of clinical guidelines (UpToDate, local protocols) with explicit citation for each statement. The model does not diagnose but helps find relevant protocols. Stack: LlamaIndex + pgvector + pubmedbert-base-embeddings + Llama Guard for safety. Data in DICOM/HL7 FHIR, on‑premise deployment mandatory.

Finance: How to Ensure Interpretability of a Scoring Model under Basel IV?

The financial sector is one of the most mature in applying ML, but regulation is maximal. Every model affecting credit decisions falls under Basel IV, EU AI Act, GDPR Article 22. We deliver AI solutions for fintech that satisfy these requirements — in a project for a top‑10 bank we deployed a scoring model where each record required SHAP explanations.

Credit scoring. Gradient boosting (LightGBM, XGBoost) dominates. Neural networks yield +0.5–2% AUC but lose interpretability. Standard: LightGBM + SHAP to explain each decision. Fairness checking is mandatory: Fairlearn or aif360 for auditing disparate impact on protected attributes (age, gender). The default class is 1–5% — with an imbalance of 1:30, a model with 97% accuracy may have recall 0.2. Solution: focal loss, class_weight='balanced', SMOTE + careful validation. In one fintech scoring project, the model reduced credit losses by $2.1 million annually.

Algorithmic trading and risk management. LSTM and Transformer for price forecasting are popular but unstable in production due to non‑stationarity of financial series. A more robust approach: ML for signal generation (classification: up/down over horizon N) with traditional portfolio optimization on top. Backtesting via Zipline‑Reloaded, vectorbt, QuantLib. Proper backtesting is critical — look‑ahead bias kills results. We guarantee a clean experiment: all data at signal time is available in real time.

AML (Anti‑Money Laundering). Graph Neural Networks for analyzing transaction networks is an actively developing area. PyG, DGL for GNN. Task: detect suspicious patterns in transaction graphs (layering, structuring). Recall is more critical than precision — better 10 false alarms than miss one money laundering. In a project for a large payment service, we increased recall by 18% without increasing false positive rate.

Retail and e‑commerce: Recommendation Systems and Demand Forecasting

Recommendation systems. Current architectural standard: two‑tower model for retrieval + ranking with cross‑features. TensorFlow Recommenders or Merlin from NVIDIA for GPU‑accelerated feature processing. For small catalogs (<100k items), LightFM is sufficient. A common mistake is training on implicit feedback without accounting for position bias. Solution: IPW (Inverse Propensity Weighting) or randomized logging on a portion of traffic. Development time for a basic recommendation system is 4–8 weeks, including A/B test.

Demand forecasting and inventory optimization. Hierarchical forecasting: SKU → category → store → region. HierarchicalForecast from Nixtla automatically reconciles forecasts across levels. TFT or N‑HiTS for base forecast, gradient boosting for adjustment on exogenous factors (promotions, weather, events). One retail project led to a 15% reduction in stock‑outs due to precise promotion calibration.

Visual search and size compatibility. CLIP embeddings for image search — deploy in 2–3 weeks: clip‑ViT‑B‑32 or clip‑ViT‑L‑14, Faiss or Qdrant index, REST API. For size recommendation — specific models on return data and reviews with fit indication.

Manufacturing: Quality Inspection and Predictive Maintenance

Quality control and defect detection. CV models for product inspection are one of the most mature industry tasks. YOLOv10 for defect detection, SegFormer for segmentation. Specifics: class imbalance (defects are rare), high recall requirement (missing a defect is worse than false alarm). Typical dataset: 500–2000 defect images + 500–1000 normal. Few‑shot learning via DINO or SAM 2 works with 50–100 annotated examples. We gained experience on an electronics production line — recall 0.95 at FPR 0.03. A predictive maintenance deployment saved a manufacturing client $500,000 per year in unplanned downtime.

Predictive maintenance. Vibration sensors, current sensors, thermocouples → feature extraction → anomaly or mode classification. Models: LSTM‑AE for unsupervised, LightGBM for supervised (if failure history is available). Integration with SCADA/OPC‑UA via opcua-asyncio or MQTT. Key metric: False Negative Rate — a missed pre‑failure is more costly than a false alarm. Threshold tuned to business cost of each error type. Timeline: 3 to 6 months to production.

Digital twin and simulation. Surrogate models — ML models replacing expensive physical simulation. If a CFD simulation takes 6 hours and a surrogate (trained on 10,000 simulations) takes 0.01 seconds, that's 2,000,000× speedup for optimization. SALib for sensitivity analysis, botorch for Bayesian optimization on top of surrogate.

General Principles of Industry AI

Regardless of industry, there are patterns that work everywhere. Data matters more than architecture. In healthcare, 1000 quality labeled images are better than 100,000 poor ones. In manufacturing, 200 real defect examples are more valuable than 10,000 synthetic ones. Compliance‑first design — regulatory requirements are easier to embed into architecture from the start than to add later. Logging, explainability, versioning from day one. Domain expert on the team — an ML engineer without domain knowledge does slowly and error‑prone what an ML engineer plus a doctor/financier/technologist does quickly and correctly.

We guarantee certification to customer requirements (ISO 13485, SOC 2, GDPR) and provide full model documentation (model card, datasheet, compliance report). Our experience: 10,000+ engineering hours and 80+ projects.

Work Process for an Industry AI Solution

1. Domain immersion (2–3 days) — interviews with experts, studying regulatory requirements, auditing available data.
2. MVP design (1–2 weeks) — stack and architecture selection, feasibility assessment.
3. Development and validation (from 4 weeks to 6 months depending on industry) — model training, testing, compliance.
4. Integration and deployment (1–4 weeks) — on‑premise / cloud / edge, documentation, staff training.
5. Support and monitoring — model drift, retraining, SLA.

Estimated timelines:

Cost is calculated individually for each project. Get a consultation — we will evaluate your dataset, regulatory map, and business goals.

Why Choose Our Industry AI Solutions?

• 80+ completed projects in fintech, healthcare, retail, and manufacturing.
• 5 years on the market — proven experience with compliance and deployment.
• Quality guarantee: we ensure target metrics (AUC, recall, latency p99) and provide full documentation.
• Licensed technologies: PyTorch, MONAI, LightGBM, Qdrant — we use open‑source with commercially safe licenses.
• Flexibility: we work as a contractor or as an extension of your team.

Contact us for a free data audit and consultation. Request a proposal with a detailed work plan. We will discuss your task and prepare a commercial proposal.