SourceFrame
ProductHow it WorksBrief TypesUse CasesFAQ

Brief typeResearch

Research Brief

A snapshot of a question that is still open — and a deliberately unfinished one.

It records what the evidence supports so far, which research questions are answered and which aren’t, what this round actually investigated, and what to research next. Uncertainty stays in the brief as state.

Free plan · No card required

Research Brief

Directional findings

On-call setups for small product teams without a platform team

Current read

Small product teams mostly run one shared rotation, and manage how many alerts reach it rather than how fast anyone responds source 2 source 5.

Research questions

  • Which rotation shape is most common at this size?

    Answered

  • Does tooling choice affect burnout, or only reporting?

    Partial

  • How do rotations work below four engineers?

    Open

Next research moves

RESEARCH STILL OPEN

Establish how follow-the-sun splits are staffed

Research this
Answers
“What do we know so far, what is still open, and what should we look at next?”
Appears when
You ask an open question in Auto on the web — no options named, or options SourceFrame has to find. There is no Research mode to pick.
Never does
Picks between options, claims the landscape is complete, or makes a recommendation. It reports where the research stands, not a consensus.

Why this type exists

Two halves of the same artifact, weighted the same

A finished-looking report would have to drop the right-hand half. A Research Brief keeps it, with the same standing as the read.

Current read

What the evidence supports right now, stated as a position rather than a conclusion.

  • One shared weekly rotation is the norm below about fifteen engineers source 2 source 4
  • Alert volume, not response time, is what teams actively manage source 5

Every point carries its sources, so the read can be re-checked when a later round changes it.

Alongside

Still unresolved

What keeps it from being settled — named, not implied by leaving it out.

  • Nothing in this round covers rotations below four engineers.
  • Whether quieter on-call came from better alerting or from fixing the services underneath.

This is where the next research moves point. An unresolved item is somewhere to look next, not a disclaimer.

Not a comparison — not yet

A comparison needs options. Research is what finds them.

Research Brief · this page

starts from an open question

Starts from

An open question — the options aren’t settled yet

Research questions

Answered, partial or open — within this round

Landscape

What this round investigated, and what it only surfaced

Still unresolved · Next research moves

What remains open, and where to look next

No criteria yet: there is no fixed set of options to hold them against.

Versus

Comparison Brief

starts from known options

Starts from

Named options — yours, or the ones research settled

How they compare

Shared criteria, coverage per option, a takeaway per criterion

Tradeoffs

Where doing well on one criterion costs another

A matrix assumes the options are the right ones. Research earns that assumption first.

Comparison Brief

Inside the brief

A snapshot that keeps its open parts in it

Research BriefV1 · Primary · 6 sourcesDirectional findings

On-call setups for small product teams without a platform team

Current read

Small product teams mostly run one shared rotation, and manage how many alerts reach it rather than how fast anyone responds source 2 source 5.

Teams under about fifteen engineers overwhelmingly describe a single weekly rotation with no formal severity tiers source 2 source 4. What they work on is how many alerts reach the person on call; response-time targets appear only where a customer contract requires them source 5.

Follow-the-sun splits come up as the way a small team spreads that load across time zones, but only one account says anything about running one source 4.

Research questions

  • Which rotation shape is most common at this size?

    Answered

  • What compensation or time off comes with a rotation?

    Partial

  • Does tooling choice affect burnout, or only reporting?

    Partial

  • How do rotations work below four engineers?

    Open

  • Which options deserve a focused comparison

    Partial

Key findings

Rotation frequency, not tooling, tracks complaints

Across the write-ups, dissatisfaction follows how often an engineer is on call rather than which paging product the team uses source 1 source 2 source 4.

Evidence gap:

No account of rotations below four engineers

No source in this round describes an on-call rotation with fewer than four people.

Landscape

Options surfaced in this research pass — not a complete market list.

Investigated

  • Shared weekly rotation

    practice
    Good evidence · 3 sources

    The default at this size: one rotation across the whole team, described with team sizes and week lengths source 2 source 4 source 5.

  • Follow-the-sun split

    practice
    Partial evidence · 1 source

    Relevant where a team spans time zones; one write-up describes running it, without saying how the split was staffed source 4.

  • Outsourced first response

    service
    Thin evidence · 1 source

    Evidence gap:A way to take overnight pages off the team; only the provider’s own page describes it source 6.

Also surfaced

  • Business-hours-only support adjacent: a support policy rather than an on-call setup
  • Managed SRE retainers mentioned once; not pursued

Worth deeper investigation

Where the current evidence points — not a recommendation.

  • Shared weekly rotation

    The setup the sources describe most fully at this size source 2 source 4.

  • Follow-the-sun split

    The only setup here that might spread load for a very small team — on the strength of one account source 4.

Compensation and time off

How teams pay for being on call, or offset it.

Time off after a rough week is the usual offset

Where compensation is described at all, it is time off in lieu rather than extra pay source 1.

Evidence gap:

Nothing outside the US

No source in this round describes compensation under EU working-time rules.

Still unresolved

4 OPEN
  • Evidence gap:How rotations work below four engineers — no source in this round describes one.
  • Evidence gap:Whether quieter on-call came from better alerting or from fixing the services underneath source 3.
  • Evidence gap:Five of the six sources describe US teams.
  • Evidence gap:Two sources are provider-published and marked single-perspective.

Next research moves

RESEARCH STILL OPEN

Establish how follow-the-sun splits are staffed

It is the one setup that might suit a team below four engineers, and a single account is not enough to compare it source 4.

Research this

  • Research rotations below four engineers — no source in this round describes one

    Research this

  • Look further into outsourced first response — one provider page so far

    Research this

Sources rail

  • source 2

    Incident-response survey write-up

    Evidence · data-heavy
  • source 4

    Startup write-up on a follow-the-sun split

    Evidence · example-based
  • source 6

    Outsourced support provider page

    Perspective · single perspective

Every claim carries the sources it came from — including the thin row, whose only source is the provider’s own page.

Source quality6 processed

Section order is fixed

  1. Current read
  2. Research questions
  3. Key findings
  4. Landscape
  5. Worth deeper investigation
  6. Sections titled by the research
  7. Still unresolved
  8. Next research moves

Landscape and Worth deeper investigation appear only when the research has options to show. There is no Evidence gaps & caveats or Suggested next steps: limitations fold into Still unresolved, and Next research moves carries the continuation.

Reading the brief

Research question · state

Answered
The evidence in this round answers it.
Partial
Some evidence, not enough to settle it.
Open
Nothing in this round answers it yet.

Answered means answered by this round’s evidence, not proven for good — a later round’s variant states every question afresh. In the brief itself, the state is a small label after the question.

Landscape · evidence strength

Good evidence
This round established a lot about it.
Partial evidence
Some of it is established.
Thin evidence
Little beyond a mention or a single account.

Strength describes how well this round investigated an option — not how good the option is. A thinly researched option may still turn out to be the best one.

Maturity

One line telling you what this research is currently good for

Maturity describes where the research stands — not a score, not a queue, and not a series of stages to finish. SourceFrame computes it from what the round established; the brief doesn’t get to choose it, and a later round’s variant computes it again.

  • Directional findings

    Some options are established well enough to point in a direction, but not yet a set you could line up against criteria. The example above is here.

  • Evidence still developing

    No option is established yet — or the question has no options to establish. The read and the open questions still stand on their own.

  • Ready to compare

    At least two options have evidence gathered for them, enough to compare them against criteria honestly. A Comparison Brief is the natural next artifact.

  • Ready to decide

    Only when your question asked for a conclusion and nothing in the research stands in the way. The Research Brief still contains no recommendation, and its next formal artifact is still a Comparison Brief.

In the briefDirectional findingsOne token in the header, beside the brief’s confidence — never a pill or a progress bar.

How you get one

Start with the research question. The structure follows the problem.

From a prompt · Web app

“What on-call setups work for a small product team without a platform team?”

Find sources & createMode · Auto
Route in
  1. Open question
  2. Discovery
  3. Research Brief

There is no Research mode to choose. Write the open question in Auto; when SourceFrame has to find the options — or there are none to find — it builds this structure instead of an answer.

What decides it

In Auto, your wording does:

“What setups work for X?” — no options named
Research Brief
“Best tools for X” — options you want found for you
Research Brief
“Compare A, B and C” — options named
Comparison Brief
“Should we pick A or B?” — options named, a decision asked for
Recommendation Brief

If you already know the options, name them and you’ll get a comparison. If you don’t, don’t invent them — the open wording is what gets you the research structure.

What you can do next

Research continues in rounds — when you want it to

  1. Stage 01
    Open question
  2. Stage 02
    Discovery
  3. Stage 03
    Research Brief
  4. Stage 04
    Research this

Stage 04Returns to stage 03 as a new variant

You can stop at 03

A Research Brief is a complete snapshot. Ending with questions still open is a legitimate place to stop, and nothing moves on by itself.

  • A bounded round

    Research this sends one move — a question or an option — out for one more round of discovery. It doesn’t re-run the whole question, and it appears only when the brief’s evidence can support another round.

  • A new variant

    A round that finds something produces a new Research Brief variant beside the original, which stays PRIMARY until you choose otherwise. A round that finds nothing new leaves your brief unchanged. A variant is a different research outcome; a version is a revision inside one.

  • Not Add sources

    Research Briefs don’t use the Add sources path. They continue through their research moves, which stay tied to the question or option they came from.

Where the work goes afterwards

  1. Research Brief
  2. Create comparison
  3. Comparison Brief
  4. Turn into…
  5. Recommendation Brief

The next formal artifact after a Research Brief is a Comparison Brief, created when you choose, from the research’s PRIMARY variant. A Recommendation Brief can follow from that comparison if you need a decision. Neither is required, and no stage advances on its own.

From Research Brief

  • Continue to

    Comparison Brief

    The next artifact once the options stop moving. It compares what this research surfaced against criteria — and still doesn’t pick one.

  • If the sources are chosen

    Source Brief

    When you already have the tabs, pages or AI chat, a question about them gets a Source Brief — an answer, not a map of what’s open.

  • Two stages away

    Recommendation Brief

    Follows a comparison, not the research directly — and the only type that states a recommendation status.

  • Use it for

    Research an open question

    The job this type was built for, including how a second round changes the picture.

Ask the question you can’t answer yet.

You’ll get a position, where each research question stands, and a next move — not a confident answer the evidence can’t support.

Free plan · No card required