The Metis Method

Fixing how research data teams work, from the inside.

The Metis Method is a Sydney consultancy for market research agencies and insight teams. Systems design, workflow mapping and automation for the work that sits between a fielded survey and a delivered answer.

What we do

Three kinds of work. Each stands alone, and each leads naturally to the next.

Review the function

How a job moves from scope to delivered data, where it breaks, and why. You get a flow map, a register of pain points traced to cause, and a roadmap sorted by horizon and effort.

Build the system

Automation for the routine: data pipelines in R and Python, tracker reporting, dashboards in Displayr, Crunch or Shiny, and the data layer behind a live poll or app. Built on a written spec, so it doesn't live in anyone's head.

Run it each wave

Managed tracker reporting. The Metis Method handles the setup, metadata and data processing every wave. Your team gets a clean, reusable interface for analysis, reporting and AI-assisted outputs.

How it works

Bottom-up first, then structure built on what the weeds revealed. The view from above and the view from the desk are mapped together, because most problems are only visible from one of them.

  • In the weeds first. Reading the code and the syntax, reviewing outputs, following live jobs end to end. That's where the pain points are, and where they're usually invisible from above.
  • Cause before fix. Each problem is traced to its cause (platform, data structure, metadata, handover) before anyone acts on it. The same cause tends to surface in several places.
  • Prototypes as bridges. Where a fix can be built small and run on a live job, it is. A working script is a better argument than a recommendation.
  • Tool-agnostic. Displayr, Q, Crunch, R, Python, SPSS, Decipher, Shiny. The method is the constant; the runtime is whatever fits the team.
  • AI with a person in the loop. Models take on more of the production work. QA, judgement and the client conversation stay with people.
  1. Discoveryharvest and map
  2. Diagnosistrace to cause
  3. Blueprintroadmap by horizon
  4. Playbackworking session
A review runs in four phases, which overlap: gathering tapers as designing takes over.

Work

Two examples from work Matt has led over the past six years. Client names are left out here; the detail is available in conversation.

A global shopper insights programme, more than ten markets

From 20,000 lines of SPSS to a metadata-driven system

The programme had to change data platforms at short notice while still delivering to clients worldwide. A decade of SPSS logic, different in every market and held mostly in people's heads, was untangled into recode tables, naming conventions and documented derivations for the variables that matter most: category, brand and retailer.

Then came the automation. An R library turns those computations into functions the data operations team runs on its own. Weighting came in-house. Dashboards are now built at scale from structured metadata rather than by hand.

What changed. Stable datasets across markets, less load on analysts, faster and more consistent reporting, and room to build new products.

An analytics software platform, enterprise research accounts

Making dataset setup repeatable for large research clients

Each large client arrived with its own SPSS legacy and its own internal language, so setup was slow and every update carried risk. Datasets were rebuilt around clear derivations and naming conventions. The unwritten rules of data preparation became a stepwise guide new DP staff could follow. Success plans set out milestones, risks and ownership across DP, product, engineering and the client.

A set of diagnostic steps separated platform issues from data-structure issues, so the product team saw themes rather than the loudest client. A small R tool for cleaning metadata grew into a script builder that DP teams could run themselves.

What changed. Steadier high-risk accounts, more consistent DP output, a clearer product roadmap, and tools that are still in use.

What we've learned

Across analytics software and insight platforms, the same patterns repeat.

  • Coherence over chaos. When collection, operations, engineering and delivery run on different models, effort multiplies and quality drops. Sequencing, ownership and handover discipline matter as much as the product.
  • Discipline over drift. Variable structures, naming conventions and routing logic aren't glamorous, but they decide whether a system holds or buckles.
  • Pragmatism over purity. Platforms can lean so far into rigour that they lose usability, or simplify until they limit insight. The job is statistical rigour inside a workflow people will use.
  • Governance beats heroics. "Just get it over the line" works until it doesn't. Shared processes, artefacts and clear ownership speed a team up more than raw effort.
  • Design for real-world mess. Inputs are never clean. Workflows that tolerate incomplete logic, inconsistent metadata and human error are the ones that scale.
  • Flow comes from structure. When workflow discipline and clean handovers line up, data moves predictably: fewer escalations, clearer expectations, faster answers.

Metis is the Greek word for practical wisdom: knowing how to get things done, not abstract knowledge.

It's the kind of knowledge a good tradesperson has, and the kind most data teams run on without writing down. The method is to find it, document it and build the system around it.

About

The Metis Method is run by Matt Steele. Matt has spent twenty years in market research: as an agency researcher in Sydney, New York and London, working across qualitative and quantitative; then more than eight years inside analytics software at Displayr and Crunch.io, building customer success and data solutions teams; and most recently as research director of a global shopper insights programme running in more than ten markets.

He works in R and Python, builds automated data pipelines and LLM-assisted workflows, and is strongest on segmentation, survey analytics and data structure. He holds an honours degree in psychology from UNSW and a graduate diploma in directing from NIDA.

Matt on LinkedIn

Contact

The first conversation is about how your work flows now and where it gets stuck.

Sydney, Australia