Marianne Costa Avalone
Marianne Costa AvaloneData & Analytics at SEBRAE-RS, Brazil

I design analytics solutions that reach the decision.

I investigate the problem before answering it, design the method, bring in the people who need to build it, and follow up on whether the solution was actually used. I came to data from architecture and eight years of scientific research, and I use both.

Seven-step work cycle, from investigating to applying, looping back to the start 1 Investigate 2 Frame 3 Design 4 Mobilize 5 Build 6 Translate 7 Apply from problemto application

The cycle that repeats across my projects. The dashed arrow is the part I care about most: going from application back to a better question.

Three ways of looking at a problem

They are not stages of my career. They are lenses I use at the same time.

Researcher

I question assumptions and separate what the evidence supports from what remains uncertain. I review the method before trusting the result.

7 journal articles and 23 conference papers; a review of an impact evaluation that informed a leadership decision.

Architect

I read the context, what already exists and the constraints before designing. A solution has to fit the system it will live in.

Geoprocessing to measure business exposure to the 2024 floods in Rio Grande do Sul.

Solution builder

I turn knowledge into methods, models and tools that other people can apply and improve without me.

A governance framework applied by colleagues in other areas; a portfolio model built into routine management.

Case studies

Told with what worked, what didn't and what is still open. Program names, vendors and amounts are omitted.

  • Impact evaluation, Statistics, Executive communication

    Reviewing an impact evaluation before acting on it

    Asked to validate a vendor's report, I found methodological inconsistencies and, instead of only flagging them, built an alternative analysis with the program team that leadership could actually use.

    Read the case study

  • Applied AI, LLM, Output evaluation

    Strategic interviews analyzed with an LLM

    A Python workflow on Azure OpenAI that turns transcripts of leadership interviews into summaries, a strategy matrix and tabulated themes. I evaluated the outputs and refined prompts and code until the tabulation was reliable enough for analysis.

  • Geoprocessing, Data integration, Crisis

    Business exposure to the 2024 floods in southern Brazil

    Geocoded businesses crossed with satellite-mapped flood and landslide areas, to guide where support went first during the crisis.

  • Data governance, Method

    Data governance that starts from decisions

    A framework that starts from business questions and works back to sources and the views each decision needs. I applied it in the first area; colleagues applied and improved it in the next ones.

  • Modeling, Automation, Portfolio management

    Scoring the analytics solution portfolio

    A model that ranks solutions by usage and recency and recommends what to migrate, keep or review, built into the management process.

Background

Changing fields was never a break. In data I found what research was missing for me: seeing knowledge put to use.

  1. Architecture and Urban Planning

    Degree and practice at a firm whose design approach valued evidence and open building systems.

  2. Master's and PhD in Japan

    Six years of research, with statistical analysis and projects funded by the Japanese government.

  3. Postdoc and teaching

    Research with industrial engineering groups and university teaching in Brazil.

  4. Data & Analytics

    SEBRAE-RS, a business support organization, in the strategy office's data team: analysis, program evaluation, governance and spatial intelligence.

Research

7 journal articles, including one in Energy Policy, and 23 conference papers. The full list is on ResearchGate.

Contact

To talk about a problem, a role or a project.

LinkedIn GitHub ResearchGate Résumé (PDF, in Portuguese)