Our services
Delivered on their own or together, our services cover how data is collected, stored, analysed, turned into evidence and acted on.
Data strategy
A data strategy is the plan that connects your business objectives to the data, systems, people and governance needed to meet them. We assess where your organisation stands today, define the target state, and deliver a costed, sequenced roadmap your team can execute.
What the engagement covers
- Maturity assessment across data, technology, people, process and governance, scored on a five-level scale you can re-measure next year.
- Target data architecture covering systems, flows and platforms.
- Prioritised roadmap with cost and effort estimates.
- Operating model setting out who owns, builds and approves.
- Governance and data protection readiness, including the Personal Data Protection Act.
- AI readiness, and which use cases are viable now.
The problem it solves
A core system, a mobile money platform and three spreadsheets disagree on how many active customers there are. Board packs take a fortnight to assemble. The warehouse bought three years ago is still half populated.
Tools and platforms
Structured maturity assessment across five dimensions, data architecture blueprints, DAMA-aligned governance and stewardship models, cost and effort estimation, and a readiness review against the Personal Data Protection Act.
What makes us different
We build what we recommend, so the roadmap is written by the engineers who would implement it and the costs reflect delivery reality. We favour open source and modular architecture, so the stack can grow a piece at a time.
Data Science Academy
Cohort-based programmes in data analytics, business intelligence, machine learning and generative AI for working professionals, corporate teams and graduates.
How a cohort runs
- Real problems and real data drawn from live client work, not textbook exercises.
- One-to-one mentorship from a practising engineer for the length of the cohort.
- Weekly review of work in progress, so nobody discovers they are lost in week nine.
- A capstone project defended before a panel that includes practitioners from industry.
The problem it solves
Adverts ask for SQL and Power BI; applicants arrive with certificates and no portfolio. Analysts who can build a chart cannot build a data model. Staff are sent on a five-day course and nothing changes when they get back to their desks.
Tools and platforms
Python, pandas and scikit-learn, SQL, Power BI and DAX, Git, Jupyter, Docker and open-source LLM tooling. Participants work in the same stack our engineers use on client projects, not a simplified teaching environment.
What makes us different
Every cohort works on real problems with real, messy data, is mentored one to one by practising engineers, is reviewed weekly rather than tested once, and finishes with a capstone project defended in front of a panel.
Applied AI & machine learning
We build and deploy machine learning models that run inside your operations: credit scoring, churn prediction, fraud detection, forecasting and document processing. Each use case is taken from problem framing through to monitoring and retraining.
Financial services
- Credit scoring on mobile money and alternative data
- Fraud and anomaly detection
- Early-warning and collections prioritisation
- Segmentation and next-best-action
Telecommunications
- Churn prediction and retention triggers
- Network and service quality optimisation
- Revenue forecasting and ARPU modelling
- Segment-level experience analytics
Public & social impact
- Targeting and beneficiary prioritisation
- Demand and caseload forecasting
- Anomaly detection in programme data
- Automated coding of open-ended responses
How we deliver
- Frame the decision, the user and the measure of success
- Ready the data and build the pipelines the model depends on
- Model, baseline first, on the metric the business cares about
- Deploy into your systems, on-premise or cloud
- Monitor for drift and retrain on a schedule your team can run
Responsible deployment, and a clean handover
We document what a model was trained on, test for disparate impact across the groups you serve, and keep a human in the loop for consequential decisions. You receive the pipeline, the trained model with its lineage, an evaluation harness, and a runbook covering retraining and monitoring.
The problem it solves
Credit decisions still rest on payslips and collateral, so salaried customers borrow and traders do not. Fraud is found after the money has gone, churn after the SIM goes quiet, and years of transaction history sit in a warehouse nobody has modelled.
Tools and platforms
Python, scikit-learn, PyTorch and XGBoost, SQL and dbt, Spark, MLflow, Airflow, Docker and Kubernetes. Deployed to cloud, on-premise GPU or fully air-gapped infrastructure.
What makes us different
We are open-source first and deployment first. No licence cost inside the business case, the model can move between environments, and you receive the pipeline, evaluation harness and runbook so your own team can operate it.
Powerful large models on your server
We deploy open-source large language models, and fine-tuned versions of them, on hardware you control. The same class of model behind commercial AI assistants runs inside your own environment, trained on your documents, terminology and languages.
What we build with it
- Internal knowledge assistants grounded in your own policy, product and procedure documents.
- Document processing covering extraction, summarisation and review.
- Verbatim and call analysis at scale, in Swahili and English.
- Drafting assistants for analysts, credit teams and programme staff.
- SQL assistants that let non-technical staff query governed data safely.
How a deployment is built
- Model selection from current open-weight families, sized to your hardware.
- Fine-tuning on your corpus, with a held-out evaluation set agreed first.
- Retrieval pipeline with permissions, citation and PII redaction built in.
- Evaluation and assurance covering accuracy, safety and audit logging.
- Handover of the runbook, monitoring and update procedure.
The problem it solves
Regulators and the Personal Data Protection Act rule out sending customer files to an overseas AI service. Foreign-currency subscriptions priced per user are hard to budget. Branch connectivity drops. And a model trained elsewhere cannot read a Swahili complaint or quote your credit policy.
Tools and platforms
Open-weight model families, vLLM and Ollama serving, LoRA and full fine-tuning, vector databases and retrieval pipelines, evaluation harnesses, Docker and Kubernetes, and affordable GPU servers sized to your workload.
What makes us different
We specify, install and tune the whole stack on affordable, high-performing servers. Everything is open weights and open tooling, so there is no per-token bill, no lock-in, and a system you can keep running without us.
Data analytics & business intelligence
We build the reporting layer of an organisation: the data warehouse underneath, the semantic model that defines the numbers, and the dashboards people actually open.
Platform agnostic by design
We do not resell a BI product, so we have no reason to push one. If you have standardised on Power BI, we build in Power BI. If it is Tableau, Qlik or Looker, we build there. If licence cost is the binding constraint, we build the same thing on open-source tools. The value sits in the data model underneath, which is portable between all of them.
- Data warehouses and pipelines that consolidate fragmented source systems into one governed store.
- Semantic models giving the organisation one definition of a customer, a branch and a month.
- Executive and operational dashboards on mobile and desktop, tailored by department and role.
- Real-time reporting for field operations, research programmes and service delivery.
- Migration between platforms when a licence decision changes, without rebuilding from scratch.
The problem it solves
Finance, operations and the branches each report a different figure for the same month, so meetings open with an argument about whose number is right. Month end is rebuilt by hand in Excel. Ten dashboards were built by a consultant and nobody opens them.
Tools and platforms
Any major BI platform: Power BI, Tableau, Qlik, Looker, Superset, Metabase. SQL Server, PostgreSQL and cloud warehouses, dbt, Python, API and core-banking integration.
What makes us different
We build the data model underneath, not just the visuals on top. That is what makes the numbers agree, and what makes the work portable if you ever change platform. Training is included in every build.
Customer experience research
We design and run customer experience measurement programmes for organisations with large branch, agent and channel networks, from instrument design and fieldwork through to analysis and live reporting.
What we measure
- Net Promoter Score (NPS), relationship and transactional, with driver analysis explaining what moves the score.
- Customer Satisfaction (CSAT) across channels, products, branches and journeys.
- Customer Effort Score (CES) for the journeys where friction costs most.
- Mystery shopping across branches, agents, contact centres and digital channels.
- Service and compliance audits of branch and agency networks.
- Institutional assessments of internal stakeholders and external clients, with stratified sampling by sector and role.
The technology behind the fieldwork
Smart glasses
Discreet point-of-view capture for mystery shopping, so a visit is evidenced rather than recalled.
Offline-first tablets
Full questionnaires with skip logic and validation, working with no network and syncing when there is one.
Automated checks
GPS, timestamps, duration outliers, straight-lining and duplicates flagged as the data arrives.
Secure sync
Encrypted transfer into a governed store, with personal data separated and access controlled by role.
Live dashboards
Scores by branch, region, product and agent, filtered to each manager's own patch.
The problem it solves
The score moves four points and nobody can say which branch caused it. The mystery shopping report lands six weeks after the visits, by which time the branch manager has moved. Enumerators paid per interview return excellent completion rates and poor back-checks.
Tools and platforms
Offline-capable CAPI tablets, smart glasses for point-of-view capture, GPS and timestamp validation, automated quality checks and back-checking, live BI dashboards, and LLM-assisted analysis of open-ended responses in Swahili and English.
What makes us different
We are a data company that runs research, so the instrument, the quality checks, the analysis and the dashboard are one connected system. Branch managers can see their own scores while fieldwork is still running.
Results-based MEL & impact harvesting
Monitoring, evaluation and learning that funders trust. We design the framework with you, capture evidence in the field, harvest and verify the outcomes that matter, and put live results in front of managers, boards and donors, not a year later but the same week.
Results-based MEL
Theory of change, results frameworks and indicators built with your team. Baseline, midline and endline, designed to answer the questions funders will actually ask.
Impact harvesting
Working backwards from change observed in the field, then verifying it with the people involved. It captures the results a logframe never predicted.
Real-time dashboards
Offline tablets feeding a live dashboard, so programme managers, country directors and donors see the same figures on the same day.
Assessments and evaluations
Baseline, process and outcome evaluations, root cause analyses and institutional assessments, with stratified sampling and CAPI fieldwork.
Success stories and film
Evidence turned into stories people remember: case studies, infographics and short captioned films of change, told by the people who lived it.
Data & Impact Partnership
Data and insights as a service: Silver, Gold and Platinum packages covering one to three projects. See the packages.
Collect
Offline tablets and basic phones, in Swahili and English, GPS stamped.
Verify
Automated checks, back-checks and triangulation against trusted sources.
Harvest
Outcomes identified, substantiated with the people involved, and coded.
Visualise
Live dashboards tracking progress against your results framework.
Tell
Reports, stories and film that funders and boards will actually read.
The problem it solves
The annual report counts trainings held and kits distributed, and the funder asks what changed. Field forms come back weeks late and half complete. The most important results of the year were never in the logframe, so nobody wrote them down.
Tools and platforms
Theory of change and results frameworks, outcome harvesting, offline mobile collection and CAPI tablets, automated quality checks, live dashboards in Power BI or open source, and AI-assisted analysis of open-ended responses in Swahili and English.
What makes us different
We are a data company that does MEL, so the framework, the collection tools, the dashboard and the story are built as one system. Evidence reaches the boardroom the week it is gathered, and your own staff are trained to keep it running.
Your own AI, inside your headquarters
A secure server that sits in your office and runs open-source AI on site. It works without the internet, so there are no data bundle costs, and your community’s information stays with you instead of going to the cloud.
It is built with Tanzania’s Personal Data Protection Act in mind, and it is not a single-purpose tool. It takes on new tasks as your programmes grow, from field reports to finance to leadership briefings.
How it works day to day
1 Capture in the field
Field staff record requests, supply updates and voice reports on a basic phone, even with no signal.
2 Sync at the office
Back in range, everything uploads to your own server, and nowhere else.
3 Just ask
Your team types or speaks a question, the way they would ask a colleague. Transcribe a voice report, draft a dispatch list, or find out how many requests came from a ward this month.
What supports every service
Data and insights as a service
You pay for what you need, when you need it. Choose the package that fits, move between packages as your needs change, and add other services only when required.
Silver
Local NGOs, single-project organisations and small donor-funded programmes
- Data oversight for one project
- Data quality assurance through verification and cross-checking
- Quarterly programme analysis
- Quarterly lessons learned reports
- Knowledge products from annual programme data
- Access to our multidisciplinary data team
Gold
Growing NGOs, multi-project organisations and national-scale programmes
- Data oversight for up to two projects
- Automated data processing
- Advanced analytics, with machine learning where it adds value
- Enhanced impact measurement framework
- Quarterly programme analysis and knowledge products
- One audio-visual data story a year
- Two capacity-building workshops a year
Platinum
Large NGOs, INGOs, UN programmes and multi-country donor-funded initiatives
- Data oversight for up to three projects
- Full monitoring system and governance support
- Real-time executive dashboards
- Systems interoperability support
- Monthly analytics and learning reports
- Two audio-visual data stories a year
- Up to three capacity-building workshops a year
Additional services, booked when you need them
Not included in the packages; quoted separately, with 10% off for partnership clients: baseline, midline and endline evaluations, outcome harvesting, data quality assessments, GIS mapping, survey programming, enumerator teams, impact stories and documentary film, and an AI server for your office.
Prices exclude VAT. Pay a quarter in advance for 3% off, or the full year for 6% off. Travel outside Dar es Salaam, workshop venues and field data collection are budgeted separately. Shilling prices at TZS 2,650 to the US dollar.
Tell us what you are trying to get right.
Most of our work begins with a short scoping call about a decision someone is tired of guessing at. There is no charge for it, and no obligation at the end.