Skip to content

User Research Plan

beta  ·  Family research-docs  ·  Phase discover  ·  Sizes lean, full  ·  ~4,900 tokens

The document written before one study of the people a product serves: why the study runs, which decision its findings will inform, what the team must learn, from whom, how, and when. For a usability test it also carries the tasks and the measures, which is why one source calls it a “usability test plan”.

This document plans one study of how people use a product, whether nothing is built yet or the product already ships. It is not a test plan, the test-plan bundle that verifies a product increment against a specification. Nor is it an A/B test plan, the experiment-design-doc bundle that allocates participants between variants. The words “test plan” overlap only because a usability test’s plan is sometimes called that - see Picking a variant, below. The deep explainer behind every rule here is user-research-plan_companion.md; a fully worked instance is user-research-plan_example.md.

Before you start: is this the right document at all?

Section titled “Before you start: is this the right document at all?”

Write a user research plan when one study of the people a product serves is about to run, and you need to agree, before it starts, why it runs, which decision its findings will inform, what the sessions must learn, from whom, and when. Nielsen Norman Group states the identity directly: “A research plan is a document that outlines the goals, methods, and logistical details of a research study.” When the study is a usability test, the plan also carries the tasks and the measures, which is why the alias “usability test plan” exists at all.

Write something else if:

You actually need Because
a user research report the not-yet-built sibling to this bundle; it answers, after the study ends, the same questions this plan sets out before the study runs
a test plan (test-plan) that document verifies a product increment against an agreed specification and grades it; this document studies how people use a product and records observations, never a verdict
an experiment design document Itamar Gilad draws the line this library relies on: “Per this definition an A/B test is an experiment, but a usability test is not,” with his own qualifier attached, “I realize that this is not the common definition used in the industry - in many cases the words test and experiment are used interchangeably.” A study that allocates participants between variants and applies a decision rule is an experiment, whatever it is called
a problem statement or a design brief this family’s own contract draws the line: those documents take the findings and frame or decide; this plan plans the study that produces the findings, and frames no problem and chooses no concept
a project brief Digital NSW places this plan “After you have been given an initial project brief, but before you commence any official research activities.” A project brief asks for authority to start; this plan sets out one study inside the authority the project brief already granted
an interview guide or a screener those are the session script, catalog entry 25, an unbuilt candidate in this family. This plan links to it under Related Documents (full) or in Method (lean); it never holds the moderator’s opening, the consent wording, or the probes itself
a standing service-level research plan, or a research roadmap GOV.UK’s round-planning page presupposes exactly that standing document: “You can only plan specific rounds of research when you’ve created a research plan for your service.” A roadmap tracking many studies at once is the same kind of standing artifact from a different angle. Both are maintained over time; this plan is written for, and finished with, one study

Write a lighter version, or none at all, if your team already meets customers every week. Teresa Torres argues that project-based research, plan and all, “it’s not sufficient for teams who continuously ship value to customers,” because “Good product discovery teams engage with customers at least weekly.” Against her, User Interviews holds that even a quick study “still needs a plan, but it can be a brief one,” and Paul Boag, who argues for scoping research down to a specific question, hedges his own recommendation: “The results you get back will be far from perfect and could possibly lead to false conclusions.” Treat this as a real dispute over how light a plan can be, not a rule either way: a team on a weekly cadence may write no separate plan per interview, but that is a judgment call for that team, not a default this guide sets for every reader.

One more sense to rule out. The word “brief” is not this document’s name under any sense this research found. Dovetail uses it for the stakeholder’s request that comes before the plan: “A brief is the request - what a stakeholder wants to learn and why it matters to them. A plan is the response - how you’ll actually study it, with method, participants, and timeline.” Maze uses it for the session guide. A third source uses it for the study’s whole intake document. Whichever sense your own team already uses “brief” for, do not call this plan a “research brief” and expect a reader to know which of those three things you mean.

Who writes it, and who agrees it before recruiting starts. No source read for this bundle names one role as the plan’s only legitimate author; a researcher, a product manager, or the pair writing together all appear in practice, and this library lets either role write it. Agreement before fieldwork is better supported: Defra’s own sequence runs from a scoping workshop to “Submit the research plan to your lead user researcher (if your delivery group has one) for review or approval,” then to “Share the approved plan with your team so everyone is aligned.” Koji is blunter still: “Get sign-off on the research questions from the stakeholders who will act on findings.” This template’s frontmatter carries that agreement as the approver field, filled in with a name, not a team, before the dates in Schedule and Logistics begin.

If the study changes while it runs, this plan records the change. Nielsen Norman Group: “Update the plan after the research study is over. In moderated studies, it’s common to make adjustments during the study to get better data.” This family’s contract fixes the plan’s finish line at the report, as this library’s own position. Until the findings are reported, revise this plan in place to say what changed. Once the report is written, this plan stops changing.

Lean carries the seven sections every size ships: Why This Study and the Decision It Informs, Goals and Research Questions, Method, Participants and Recruiting, Tasks and Measures, Schedule and Logistics, and What Happens to the Findings. Full adds four, in this order: What We Already Know, Roles and Observers, and Consent, Privacy and Ethics, each inserted among the lean seven, then Related Documents, appended after them.

Use full when at least one of these is true: earlier research or assumptions are worth summarizing before the sessions run; more than one person will moderate, take notes, or observe, and that needs writing down rather than assumed; consent, data handling, or retention needs its own section rather than one line inside Participants and Recruiting; or related materials, the session guide, the screener, the consent form, the prototype, are worth linking in their own place rather than folded into Method. Use lean when none of those is true: one moderator, consent handled the way the team always handles it, and nothing to link beyond the session guide itself.

Tasks and Measures ships at every size, filled in only for a usability test. For any other method, the section is one line naming the method actually used, with the task table and the two measure lines marked N/A. That is a complete answer, not a placeholder waiting to be filled. The usability-test-plan identity rests on one publisher’s sentence: “When planning a usability test, the research plan is often referred to as a test plan.” So this block is the clause that makes the alias true when it applies, never an optional extra once it does.

The session script stays out of this plan at every size, and that exclusion is this library’s own position, not a documented convention. Most sources this bundle’s research read put the moderator’s opening, the consent wording, and the follow-up probes inside the plan itself; this library links them instead, because the interview guide is a separate, unbuilt candidate in this family’s own catalog. The line this plan draws is: task scenarios, their success criteria, and the measures stay here; the moderator’s opening, the consent script, and the probes go to the linked guide. No source draws that exact line; it is this library’s own arrangement, stated here so a reader does not mistake it for received practice.

Below 16 out of 22 on a full plan, a stakeholder reading it cannot tell what decision the study serves, whether the sessions will actually answer the questions set for them, or who is accountable for what happens once the findings come in, without asking the author questions this document should have answered on its own. Below 10 out of 14 on a lean plan, the same gap opens with four fewer sections standing between a weak answer and the people the plan is meant to align.

# Criterion 0 1 2
1 A falsifiable decision No decision named, or the goal is to “understand” or “explore” A decision is named, but nothing says what result would change it You can name the decision and the specific result that would change it
2 Agreed before recruiting starts No approver named anywhere in the document An approver is named, but the status does not read as agreed, or the plan’s date falls after sessions in Schedule and Logistics begin A named approver appears, the status reads as agreed, and the plan’s date comes before the first recruiting or session date
3 Answerable research questions A question restates the goal (“understand the experience”) A question is phrased as something to learn, but reads like something a moderator would say aloud to a participant Every question is phrased as something the study must learn, none could be read aloud to a participant as it stands, and the questions a moderator would ask are left to the linked session guide
4 Method fits the questions A method is named with no reason it fits these questions Method and a reason are both named, but a reader cannot tell whether it is generative or evaluative Method, fit, and a stated generative-or-evaluative (or qualitative-or-quantitative) label together tell a reader what the findings can and cannot settle
5 Screenable participant criteria “Some users,” with no criterion for who qualifies A criterion is named but vague enough that two readers would screen different people A reader who has never met the team could use the qualifying and disqualifying lines alone to screen a real person in or out, and the chosen count carries a reason stated for this study
6 Tasks set before fieldwork The study is a usability test and the table is blank, or the N/A line names no actual method Tasks exist, but a success criterion is missing, vague, or set after the sessions ran For a usability test, every task is a scenario with a success criterion set now and a named measure; for any other method, one line names the method actually used and the table and measure lines read N/A, which scores the same as a full usability block
7 Findings feed a decision The section names a build, iterate, or drop decision, or any other verdict Output and audience are named, but nothing ties back to the decision named in row 1 Output, audience, rough timing, and the decision it feeds are all named, and nothing in the section reads as the study’s own conclusion
8 A named earlier signal (full only) “No prior research exists,” or the section is blank Something earlier is named, but no assumption is stated for this study to test A specific earlier signal is named with where it came from, and the assumption this study tests is distinguishable from something the team already knows
9 Roles are named people (full only) “The researcher will handle it,” with no note taker or observer named A moderator is named, but note-taking or observing is unaccounted for Moderator and note taker are named (even as the same person, stated as such), and observers are named with how they watch, open to anyone willing and informed
10 Retention points to policy (full only) “Participants will sign something,” with no mention of how long anything is kept Consent is described, but a retention period is stated as this study’s own figure rather than a pointer to policy Consent and what participants are told are both concrete, and retention points to the organization’s standing policy rather than restating a number as this study’s own
11 Linked documents resolve (full only) “See the shared drive,” or similar Items are named, but not as something a reader could click through to The session guide, screener, consent form, and test environment each resolve to a real link or an explicit “not applicable,” and no moderator script is pasted in here instead of linked

Which rows apply to what. Full ships all eleven rows, because a reader of a full plan can ask about earlier knowledge, named roles, consent and retention, and linked materials, as well as the seven every size must get right. Lean ships the first seven, every row the lean variant has a section to support.

Document Rows Maximum Score against
lean 1-7 14 10
full all 11 22 16

The test behind every row above: could someone satisfy the cell by adding words without the plan actually telling a reader more? Row 5’s count, for instance, is never graded by how many participants are listed; the sources disagree too widely for any one number to be a standard, so the row grades whether the chosen number carries its own stated reason, not what that number is.

Treating “test plan” as this document’s own name, unqualified. Nielsen Norman Group’s own usability checklist uses the phrase for this document too, which is exactly how the collision happens. The tell: the word “test” appears with no “research,” “usability,” or “user” beside it. Fix by naming the document fully, a usability test plan or a research plan, and never a bare “test plan.”

Calling a usability test an experiment. The tell: a decision rule compares two variants, or participants are assigned to a condition. Fix by reserving “experiment” for a study that allocates participants between variants and applies a decision rule, and calling a usability test a test, as Gilad’s own definition does.

Writing a recommendation or a build-or-drop decision into the plan. The tell: a sentence in What Happens to the Findings reads like a conclusion rather than a plan for producing one. Fix by naming which decision the findings will inform and stopping there; the recommendation itself is the report’s job, once the sessions have run.

Carrying a programme budget, a reserve, or a standing retention policy as though it were this study’s own figure. The tell: a specific number of months, or a dollar reserve, that reads like it was copied from somewhere else. Fix by naming how this study’s own participants are thanked or paid, and pointing to the organization’s retention policy rather than restating one program’s figures as this plan’s own.

Repeating a participant count, an analysis-to-fieldwork ratio, or an aggregation threshold as though it were a standard. The tell: a bare number with no reason attached to this study. Fix by stating the number chosen, with the reason it was chosen for this study, and leaving someone else’s figure as background rather than a rule.

Using the words “research brief” as though they meant this document. The tell: “brief” appears with no qualifier, anywhere in the plan or in conversation about it. Fix by saying plainly which study informs which decision, and leaving “brief” for whichever of its several senses your own team already uses it for.

Presenting the usability block as an optional extra for a usability test. The tell: Tasks and Measures is filled in last, or thinly, while every other section is detailed. Fix by filling Tasks and Measures before most of the rest of the plan when the study is a usability test; it is the heart of the plan at that point, not an appendix to it.

Pasting the session script into this plan instead of linking it. The tell: a moderator’s opening, a consent script, or a list of probe questions sitting inside Method or Related Documents rather than behind a link. Fix by linking the session guide and the screener, and keeping only the task scenarios, their success criteria, and the measures inside this document.

When the decision the study serves is named with what would change it, when every research question reads as something the sessions can answer rather than something a moderator would say aloud, when the participants named are falsifiable by a stranger reading the plan, and when a named approver has agreed the plan before recruiting began, by a real date.

Then self-grade against the rubric above, and delete every HTML comment. A plan that reads well but names no decision, or that quietly carries a recommendation before the sessions have run, has not finished the job it exists to do, whatever else it contains.

user-research-plan_template-lean.md · ~4,900 tokens

---
title: "{{study_name}} Research Plan"
study_name: "{{study_name}}"
author: "{{author}}"
approver: "{{approver}}"
status: "{{status}}"
last_updated: "{{date}}"
doc_type: user-research-plan
size: lean
source_template: user-research-plan
source_template_version: 0.1.0
---
<!--
LEAN USER RESEARCH PLAN. The shortest form that still says why one study runs, what it must learn,
who takes part, and what the sessions will do, with the usability block (Tasks and Measures) carried
at every size because the alias "usability test plan" promises a reader it is there. Use it for a
study with no need to document earlier assumptions, roles, consent detail, or related links in their
own sections. To add What We Already Know, Roles and Observers, Consent, Privacy and Ethics, and
Related Documents, see user-research-plan_template-full.md. Never rename or reorder the sections
below; the full variant is a strict ordered superset of this one.
WHAT A USER RESEARCH PLAN IS, AND IS NOT
It is written before one study of the people a product serves. It says why the study runs, which
decision its findings will inform, what the team needs to learn, from whom, how, and when. Nielsen
Norman Group: "A research plan is a document that outlines the goals, methods, and logistical details
of a research study. (When planning a usability test, the research plan is often referred to as a
test plan.)" No second source states that identity in a sentence; three more give it structural
support without naming it. IT IS NOT a bare "test plan": this document and a separate bundle,
`test-plan`, share those two words, but `test-plan` verifies a product increment against a
specification, and this document studies how people use a product. IT IS NOT an experiment: Itamar
Gilad draws the line this library relies on, "Per this definition an A/B test is an experiment, but a
usability test is not," and his own next sentence must travel with that line wherever it is cited: "I
realize that this is not the common definition used in the industry - in many cases the words test and
experiment are used interchangeably. I chose this definition because it’s closer to how scientists
think of experiments." IT IS NOT the report: this plan names which decision the findings will inform
and stops; the recommendation itself is the not-yet-built `user-research-report`'s job. IT IS NOT the
session script: the moderator's opening, consent wording, and probes live in a linked session guide,
never pasted into this plan (catalog entry 25, an unbuilt candidate). IT IS NOT a "research brief":
that word carries at least four senses across the sources this bundle's research read, and none of
them is a synonym for this document. IT USUALLY COVERS ONE STUDY: Nielsen Norman Group states the
common case and its own exception together, "Typically, a research plan is constructed for each
study," though "sometimes, a research plan could include numerous studies or phases of research,"
naming interviews and a diary study as one pairing. IT STILL DECLARES DISCOVERY WORK FOR A SHIPPED
PRODUCT: a study that asks where people lose time or need help with something they already have is
still discovery work, by this library's own position; no source names a study of a shipped product as
discovery work in those words. See user-research-plan_companion.md sections 1, 6 and 7.
HOW TO FILL THIS IN
1. Read the comment under each heading: WHAT it wants, WHY it matters (with a pointer into
user-research-plan_companion.md), guiding questions to ASK, a GOOD and a WEAK example, and the TRAP
to avoid. For the Tasks and Measures table, PRIORITY explains the ordering rule and ROW HINT says
what a good row contains.
2. Replace each {{placeholder}} with your content.
3. If a section does not apply, including Tasks and Measures when this study is not a usability test,
write "N/A" and one line of why, rather than deleting the section.
4. Before you share this plan with stakeholders: self-grade against user-research-plan_guide.md, then
DELETE every HTML comment. They are guidance, not content.
-->
# {{study_name}} Research Plan
## Why This Study, and the Decision It Informs
<!-- WHAT Why the study runs now, and which decision its findings will inform. If this studies a
product people already use, says why that is still discovery work.
WHY Dovetail ties the two together: "Background explains why the study exists and what
decision it informs." Coleman sharpens it: "Start with the decision, not the question," and
asks "What would change their choice." unitQ opens the same way: "What decision will this
inform?" This library holds, as its own position, that a study of a product people already
use is still discovery work when it looks for an opportunity rather than grading the
product against a specification; no source names a shipped-product study as discovery work
in those words. Deep dive: user-research-plan_companion.md section 3 (Anatomy > Why This
Study, and the Decision It Informs), section 8.
ASK Why does this study run now? What decision will the findings inform, and what result would
change that decision? If this studies something people already use, why is this still
discovery work rather than a check against a specification?
GOOD "This study runs because the delivery-slot step has been proposed for redesign twice and
deferred twice for lack of evidence. Decision it informs: whether to fund a redesign next
quarter. If shoppers complete the step without confusion, the redesign drops off the
roadmap."
WEAK "We want to understand how people feel about checkout." (no decision named, no reason the
study runs now)
TRAP Naming a goal to understand or explore without naming the decision the findings would
change. If nothing the study finds would change a decision, the study has no reason to
run. -->
- Why this study runs now: {{why_now}}
- Decision the findings will inform, and what would change it: {{decision_informed}}
- If this studies something people already use, why it is still discovery work: {{shipped_product_note}}
## Goals and Research Questions
<!-- WHAT One or two goals, and the three to five research questions the sessions must answer, stated
as what the team needs to learn, not what a moderator will ask. A hypothesis may stand in
for a question.
WHY Nielsen Norman Group: "Purpose and goals of the research study." Dovetail draws the
load-bearing line: "Research questions describe what you need to learn, not what you'll ask
participants," limited to "three to five per study." 18F agrees: "Research questions are
different from interview questions." unitQ frames a hypothesis as a belief with a
confirming condition. Deep dive: user-research-plan_companion.md section 3 (Anatomy > Goals
and Research Questions).
ASK What is the one or two goals of this study? What three to five questions must the sessions
answer? Could a hypothesis, a belief with a confirming condition, stand in for one of them?
GOOD "Goal: find where shoppers lose time choosing a delivery slot. Question: what makes a
shopper hesitate before picking a slot? Hypothesis: we believe shoppers skip the cheapest
slot because they distrust it will hold; we will know we are right if more than half
mention price when they hesitate."
WEAK "Understand the delivery-slot experience." (a vague verb, not a question the sessions can
answer)
TRAP Writing an interview question instead of a research question. "Walk me through picking a
slot" belongs to the session guide; "What makes a shopper hesitate" is what the plan needs
to learn. -->
- Goal(s): {{goals}}
- Research questions (three to five): {{research_questions}}
- Hypothesis, if one stands in for a question: {{hypothesis}}
## Method
<!-- WHAT The method or methods, why they fit the questions, the session format and length, what
participants will see, whether the study is generative or evaluative, and where the
session guide and screener live.
WHY User Interviews: "How: Methodology." Nielsen Norman Group: "Method and procedure."
digital.gov: "Note the method or methods of research you will be using, such as contextual
inquiry or usability testing." Torres names the distinction this section carries:
"Generative research helps teams discover opportunities... Evaluative research helps teams
test which solutions will best address those opportunities." Nielsen Norman Group lists
interview materials inside its own method section while also naming relevant documents
separately; this plan follows the same pattern, naming the session guide here and, in the
full variant, listing it again under Related Documents. Deep dive:
user-research-plan_companion.md section 3 (Anatomy > Method).
ASK Which method fits these questions, and why? Is this generative or evaluative, qualitative
or quantitative? What will participants see or use? Where do the session guide and screener
live?
GOOD "One-to-one moderated sessions, about 30 minutes, on the live delivery-slot screen with a
sample cart. Generative: we are looking for where shoppers lose time, not testing a
specific redesign. Session guide: linked below."
WEAK "We'll talk to some users." (no method named, no format, no sense of generative or
evaluative)
TRAP Naming a method without saying whether it is generative or evaluative. A reader cannot tell
what the findings can and cannot settle without that distinction. -->
- Method(s), and why they fit the questions: {{method_and_fit}}
- Session format and length: {{session_format}}
- What participants will see or use: {{what_participants_see}}
- Generative or evaluative, qualitative or quantitative: {{study_kind}}
- Where the session guide and screener live: {{session_guide_link}}
## Participants and Recruiting
<!-- WHAT Who qualifies and who does not, how many (with a backup count), how they are recruited and
thanked or paid, what access needs you ask about, and how consent is taken.
WHY Nielsen Norman Group: "Participants." Dovetail: "Your plan should state who qualifies, who
doesn't, and how many sessions you need." GOV.UK adds "how you'll cater for any access
needs" and "how you'll find them within your budget." Maze names paying participants its
own step, "Determine how you'll pay them." Pernice names a planned fallback, "intentional
overrecruiting," because "each lost participant represents a fairly substantial proportion
of your potential insights." Participant counts are contested across the sources, from
NN/g's five for a qualitative study to its own 20 to 30 for a quantitative one; state the
number chosen for this study, not a source's figure as a rule. AnswerLab argues for a
demographic mix by default; DHS argues against worrying about it for a usability test.
Deep dive: user-research-plan_companion.md section 3 (Anatomy > Participants and
Recruiting), section 6.
ASK Who qualifies, and who explicitly does not? How many participants, and how many will you
recruit as backup? How are they recruited and thanked or paid? What access needs will you
ask about? How is consent taken (full detail lives in Consent, Privacy and Ethics, where
this plan uses it)?
GOOD "Qualifies: placed 3 or more grocery orders this month, uses delivery. Does not qualify:
anyone on the design team. 6 sessions, recruiting 8 for drop-outs. Recruited through the
in-app panel, thanked with a 40-dollar gift card. Every participant is asked about access
needs before scheduling."
WEAK "We need about 5 to 10 users, TBD." (no qualifying or disqualifying criteria, no recruiting
channel, no incentive)
TRAP Quoting a participant count from a guide as a rule. The sources disagree widely; state the
number chosen for this study and why. -->
- Who qualifies: {{who_qualifies}}
- Who does not qualify: {{who_does_not_qualify}}
- How many participants, and how many recruited as backup: {{participant_count}}
- How participants are recruited: {{recruiting_channel}}
- How participants are thanked or paid: {{incentive}}
- Access needs you ask about: {{access_needs}}
- How consent is taken: {{consent_summary}}
## Tasks and Measures
<!-- WHAT For a usability test: each task as a scenario, a success criterion set before the sessions
run, the measures the sessions will collect, and when questions run. For any other method,
one line saying so, with the task table and the two lines under it marked N/A.
WHY The usability-test-plan identity is stated in a sentence by one publisher, Nielsen Norman
Group: "A research plan is a document that outlines the goals, methods, and logistical
details of a research study. (When planning a usability test, the research plan is often
referred to as a test plan.)" For a usability test this is the heart of the plan, not a
supplement: NN/g's own task guidance, "Strong tasks are concrete and free from clues that
might prime people's behavior," and MeasuringU's, "Determine an objective success criteria
for each task," both put it among their named items. Deep dive:
user-research-plan_companion.md section 3 (Anatomy > Tasks and Measures), section 1.
ASK Is this a usability test? If not, name the method actually used and mark the rest N/A. If so:
what is each task, as a scenario free of priming clues? What is its success criterion, set
now? Which measures (time on task, success rate, error rate, satisfaction) will the
sessions collect? When do questions run, before, between tasks, or after?
PRIORITY List tasks in the order participants attempt them. A success criterion is set now,
before the sessions run, never scored after the fact.
ROW HINT A good row states the task as a scenario a participant could act on, with no clue to the
right path, and a success criterion stated as an observable outcome. A weak row names a
feature instead of a scenario, or leaves the criterion blank.
GOOD | 1 | You want produce delivered tomorrow evening. Find a delivery slot and confirm it. |
Confirms a slot without asking for help, in under 90 seconds |
WEAK | 1 | Test the delivery-slot picker. | (blank) |
TRAP Treating this block as optional detail. For a usability test it is the core of the plan,
not an appendix to it. -->
If this study is not a usability test, say so, name the method, and mark the table and the two lines below N/A: {{not_usability_test_note}}
| # | Task (as a scenario) | Success criterion |
|---|---|---|
| {{task_number}} | {{task_scenario}} | {{success_criterion}} |
- Measures this session will collect: {{measures}}
- When questions run (before, between tasks, after, or both): {{questionnaire_timing}}
## Schedule and Logistics
<!-- WHAT Dates and session times, place or tool, how sessions are recorded and noted, the pilot
session, and time scheduled for analysis.
WHY User Interviews: "When and where: Research schedule and logistics." GOV.UK: "Plan the
session times" and "Arrange recording and note-taking." A pilot is strongly supported:
GOV.UK's own "Run a practice session," Michigan Tech's "the protocols are pilot-tested
before the actual tests are conducted," and Maze's caution that "your pilot participant has
had no part in the usability testing process so far." Analysis time is contested:
digital.gov suggests "twice as long as you give to data collection," GOV.UK "For every 2
hours of research, you should do 1 hour of analysis"; the two figures may not measure the
same thing, so state what you scheduled rather than either ratio as a rule. Deep dive:
user-research-plan_companion.md section 3 (Anatomy > Schedule and Logistics), section 6.
ASK When and where do sessions run? How are they recorded and noted? Is a pilot scheduled, with
someone who has had no part in the study so far? How much time is scheduled for analysis?
GOOD "Sessions run 2026-03-10 to 2026-03-14, 30 minutes each, over video, recorded with consent.
Pilot: 2026-03-07, with a teammate from a different team who has not seen this work.
Analysis: two days blocked after the last session."
WEAK "We'll schedule sessions sometime next month." (no dates, no recording plan, no pilot, no
analysis time)
TRAP Skipping the pilot, or running it with someone already close to the work. -->
- Dates and session times: {{schedule}}
- Place or tool: {{location_or_tool}}
- How sessions are recorded and noted: {{recording_plan}}
- Pilot session, and with whom: {{pilot_plan}}
- Time scheduled for analysis: {{analysis_time}}
## What Happens to the Findings
<!-- WHAT Which report or readout the study produces, for whom and by roughly when, how it will be
structured if known, and which decisions the findings feed. Never a recommendation or a
decision.
WHY User Interviews: "Next steps." Dovetail's "share-out plan." MeasuringU: "If possible, try to
outline how the report will be structured in the test plan." Two qualifiers travel with
this section: 18F warns, "Avoid over-specifying your outputs, because you don't know what
you'll find until the research is underway," and DHS notes one of its own approaches "may
or may not result in a formal report." What this section does not carry: unitQ's own next
step, "a clear recommendation: build, iterate, keep exploring, or drop," belongs to the
report, not this plan. Deep dive: user-research-plan_companion.md section 3 (Anatomy > What
Happens to the Findings).
ASK What report will this produce, for whom, and roughly when? How will it be structured, if
you can say now? Which decision named above will the findings feed? (Name no recommendation
here; that is the report's job.)
GOOD "A short readout to product and design within a week of the last session, one section per
research question. Feeds the go or no-go decision on funding the redesign. The exact shape
is not locked, since we do not know what we will find."
WEAK "We'll write something up and recommend building the redesign." (names a recommendation
before the study runs)
TRAP Writing the recommendation or the build-or-drop decision into this section. That is the
report's job, once the sessions have run. -->
- Report or readout, for whom, and roughly when: {{findings_output}}
- How it will be structured, if known yet: {{findings_structure}}
- Which decision(s) named above the findings will feed: {{decisions_fed}}

The reasoning, the history and every source, in the repository:

  • Companion - the long-form argument: why these sections, where the sources disagree, and what the bundle refuses to claim
  • History - what changed in this bundle, and when
  • Research log - every source consulted, with what each one actually supports
  • Catalog metadata - the machine-readable record this page is generated from

Catalog record: 11 sections across 1 format(s), methodology UX research / user-centered design, typically owned by UXR / PM.