Guerrilla research on par with higher-order evidence

In my previous post, I briefly mused about how community teams can turn fast, local inquiry (which I am calling guerrilla research) into findings that carry weight beyond their own programs. The trick is to keep the speed and proximity of guerrilla methods while layering in habits that signal rigor, comparability, and fairness. What follows is a practical route that starts with small interviews and quick polls, then climbs toward designs and outputs that academics and policy people will recognize.

Start by upgrading the question

Frame each project with a short theory of change and one primary outcome. Write one sentence on who is affected, one sentence on what changes, and one sentence on how you will know. That tiny discipline prevents scope creep and makes every later decision easier to justify. Beware of loose questions inviting unhelpful answers.

Use “minimum viable rigor”

Like any good designer, a good researcher should aim to achieve some level of minimum viable rigor (MVR) that demonstrates the quality of their data and the care and robustness of their analysis. There are three core steps to build this MVR:

First, register your plan before you collect data. OSF or a simple timestamped document is enough. Add a paragraph on sampling, instruments, and the primary analysis. You are not locking yourself into a rigid script. You are showing intent and cutting hindsight bias.

Second, standardize measurement. Use the same two or three items every time you run a quick intercept interview. Keep a short codebook for qualitative tags. Reuse instruments across projects so findings line up over time.

Third, preserve lineage. Store raw data, cleaned data, and analysis outputs separately. Save the cleaning steps. Later readers care less about your software and more about whether a number can be traced from claim to source.

Move beyond one-off snapshots

One of the most common designs that community practitioners fall back on is the cross-sectional design – a single snapshot (usually after an intervention) that acts more like a debrief than a robust examination of a phenomenon or intervention. A simple step to layer onto one’s approach would be the pre-post or before-and-after follow-up. From the beginning, already think about how you will want to capture a before snapshot so that you will have at least one comparator for the after snapshot. Other possible designs include:

Interrupted time series uses repeated measures before and after a change. Even six to eight time points can show whether a shift coincides with an intervention or with ordinary ebb and flow. You can plot the line in Jamovi or R and estimate a level change with a simple model. The visual does half the persuasion.

Stepped adoption uses natural rollouts. When one site or team adopts a practice earlier than another, you can compare paths rather than single points. This needs careful logging of when changes happened and a shared outcome definition. It does not require randomization or large samples.

Borrow quasi-experimental habits without the pain

Simple matching reduces obvious confounding. If you are comparing people who got a new home-visit protocol with those who did not, match on age band and baseline risk. Exact matching in a spreadsheet is fine for small samples. Report how many cases were dropped and why.

Synthetic controls are sometimes possible with public time series. If a neighborhood clinic changes appointment reminders in March, build a comparator from similar clinics that did not change in March. You are still doing service improvement. You are also speaking a dialect that evaluators understand.

Triangulate

Triangulation is the process of drawing from more than one source of data that is independently generated from other sources, and then comparing them to see if they corroborate or contrast. Interviews tell you why. Counts tell you how much. Documents tell you what actually happened. Using two of the three is usually enough to firm up a claim. For example, pair a five-minute exit question with routine service metrics, or pair coded field notes with a weekly tally. You are not chasing methodological decoration. You are checking that a story holds under a second light.

Write like a careful researcher without the jargon

A short abstract with a question, setting, design, primary outcome, and main result builds trust. Report effect sizes and uncertainty rather than only p-values. State one limitation you believe could change the conclusion. Add a plain-language paragraph on what a thoughtful skeptic would still want to test.

Pre-plan a pathway from quick test to cumulative evidence

Think in arcs. Three or four tiny projects that use the same outcome can be meta-summarized. When sites differ, treat site as a variable and show the range rather than a single pooled number. Over a year, this produces a living evidence line rather than scattered anecdotes.

Pair with some basic economics

Pair outcome changes with a basic resource calculation. Sketch out rough pathways of how resources are used for an intervention; track how much money is lost due to some social ill. A two-row table that shows cost per avoided escalation or cost per additional completed referral speaks to managers and policy staff. You do not need a health economist to calculate exact dollar values; sometimes a simple account sheet can be compelling as a source of data.

Create shareable, reviewable packages

Archive every project as a small, complete bundle. Protocol, instruments, de-identified data, cleaning recipe, code, figures, and a one-page brief. Packages that travel are packages that get read. They also let academics replicate or extend without a long onboarding call.

Publish where practitioners and scholars both look

Three tiers help you punch above your weight.

• Post the full package in an open repository with a stable link. That is the canonical version.
• Write a practice note for a field outlet read by community providers. Short, concrete, and timely.
• Prepare a preprint or a short methods letter that highlights your design trick, your measure, or your stepped rollout. Scholars often welcome lean, well-documented natural experiments, especially if data and code are available.

Building the infrastructure and community

Picture the smallest version of a shared toolkit. A folder of templates. A few short guides that people actually read. A monthly call where someone kindly says, try this first before you run the numbers again. That is all “infrastructure” needs to mean in a community setting. Templates for protocols shrink the blank-page dread and nudge teams to define the question, the outcome, and the time window. A consent note that has already been checked for plain language stops last-minute scrambles. A tiny codebook keeps five people from inventing five names for the same idea.

With some careful curation and standardisation, a community should start to form around these small improvements. One person keeps the protocol template tidy. Another hosts open office hours once a month, camera on, ready to look at your spreadsheet without judgment. A third curates short examples of good practice from down the street rather than from a distant journal. Over a season, those rituals change how people work. New staff learn the house style. Old staff stop reinventing entry forms. Confidence grows because the basics are shared, predictable, and easy to reach. If your sector can manage one shared drive and one recurring call, you already have the bones of a research community.

An example ladder in practice

Week 1. A youth outreach team pilots a new text reminder. They register a two-paragraph plan and define one outcome. They clean four weeks of historic attendance in Power Query and set up a weekly export.

Weeks 2 to 6. They run an interrupted time series with six pre and four post points. Attendance jumps, then drifts. In parallel, five exit interviews get coded in QualCoder. The theme that shows up is timing rather than wording.

Week 7. Reminder time is shifted by one hour for late-shift youth. Another four points confirm a sustained rise in that subgroup. A simple match on baseline attendance shows similar youth without the timing change did not move. The package goes to OSF. A one-page brief lands with the city youth office alongside a two-row cost table. A methods note on “micro ITS in outreach programs” goes to a practice journal.

What makes outsiders take this seriously

People look for a few cues. Consistent outcomes across projects. A trail from figure to raw file without detective work. A design that fits the question you actually asked. Two names who were not on the team giving light review. A clear line from effect to effort so managers can see the trade. None of this needs lavish budgets. All of it rewards care.

That is the aim. Keep the speed of everyday inquiry and add enough scaffolding to turn local learning into public evidence. If we share the bones, speak the same tags, and archive complete little bundles, the sector starts to sound coherent. Funders and academics hear a chorus. And you feel less alone when the next question arrives.