Lock a research plan's scope before Deep Research burns its run on the wrong question

You are helping me write a research plan for a decision I actually have to make, not a general topic to explore. Before this plan goes into a Deep Research run, it needs to be scoped tightly enough that the run doesn't wander into interesting-but-irrelevant territory.

DECISION THIS RESEARCH FEEDS
Whether to add a self-serve annual plan alongside our existing monthly-only pricing before Q1.

CURRENT BEST GUESS
I think annual pricing would raise ARPU but I'm worried it would tank our trial-to-paid conversion rate.

WHAT WOULD CHANGE MY MIND
If two or more comparable SaaS tools our size saw conversion drop more than 15% after adding an annual-only upsell path.

OUT OF SCOPE
Not researching enterprise contract pricing or multi-year deals — this is only about the self-serve tier.

DEADLINE AND DEPTH
One afternoon, roughly 90 minutes of active research, not a multi-week study.

RULES
Start from the decision, not the topic — every section of the plan should trace back to how it would change what I do next. If a sub-question wouldn't change the decision either way, cut it, even if it's genuinely interesting. Write the current best guess as a real position, not a neutral placeholder, because a research plan built against "I don't know yet" produces a survey of the whole topic instead of a targeted check of one belief. State the disconfirming evidence as a concrete, checkable observation — the plan should include an explicit instruction to actively look for that observation, not just accumulate confirming detail. Take the exclusions seriously: name adjacent topics people commonly conflate with this one and mark them as out of scope so the eventual research run doesn't drift into them just because they're nearby. Size the plan to the time budget — a two-hour research pass and a two-week one should produce structurally different plans, not the same outline with more or fewer bullets.

WHAT NOT TO DO
Do not produce a generic research outline (background, analysis, findings, conclusion) — that shape belongs to a report, not a plan for gathering evidence. Do not hedge the current best guess into vagueness to sound neutral.

OUTPUT FORMAT
1. The single decision this research must inform, restated in one sentence.
2. 3-5 sub-questions, each tagged with which way an answer would push the decision.
3. The specific disconfirming check to run.
4. Explicit exclusions list.
5. A recommended source mix (primary data, expert commentary, competitor filings, etc.) sized to the time budget.