The tutorial is finished and the bookmarks have multiplied. Next weekend, you cannot decide whether to read an annual report, change a backtest, or investigate the newest model. After three busy hours, a few summaries remain, and next week begins with the same choice.
A research system connects the next session to the previous one: question, evidence, reason to continue, and reason to stop. Arrange existing work across ninety days, then deliberately halve the weekly time allowance. Which tasks must disappear?
Define what this round will leave behind
You will select one T12 priority, write a complete time-and-cost plan, connect evidence and judgments with periodic and event-driven review, and revise or stop the plan when conditions change. Ninety days is an optional planning span for sourcing, review, and reflection—not a universal learning prescription.
A plan specifies when, where, and what checkable action will occur. Research on implementation intentions examines linking a situation with a goal-directed response.1 That supports concrete planning here, without establishing that this ninety-day template improves investment returns or that a fixed workload works for everyone.
Continue with fictional Jianning's capability archive, without adding real investment results. She has completed a given company-update exercise and a small fee-inclusive ledger; independent sourcing, recipient review, and operation still have gaps. She chooses company research for this round rather than launching a new quantitative model too.
Narrow the question: Can a chosen ordinary nonfinancial company's revenue change be explained alongside operating cash changes using consistent periods, currency, and scope? Independently locate at least one relevant original disclosure absent from the exercise packet. Preserve identity and publication time, two key fields, an opposing explanation, and an update. The assignment does not require proving the stock is a buy.
The quantitative route can substitute G63's fixed question, lawful inputs, rules, ledger, and negative results, with only one limited variation. Insufficient real data or a tool path requiring unprovided source-code changes remains a blocker. An author-supplied mechanism or issuer-identity page cannot stand in for independent research.
Budget the scope, without filling every day
This planning example gives Jianning three hours per complete week: two for materials and one for checking and records. The first 84 days are twelve weeks, or 36 hours. The last six days receive two hours for archiving, bringing the total cap to 38 hours. Reviews are included, not added as an invisible extra evening.
| Relative days | Time cap | Task | Expected file or decision |
|---|---|---|---|
| D1–7 | 3 hours | Review T12, choose one question, list materials and limits | One-page T03, old-work index, excluded scope |
| D8–21 | 6 hours | Locate original disclosures; verify publication time and two fields | T02 source record and locations, or a data-blockage note |
| D22–35 | 6 hours | First explanation, one key calculation, opposing explanation | V1, inputs/formulas, unknowns, review triggers |
| D36–49 | 6 hours | Four hours for counterevidence and one variation; at most two for optional new-candidate reading, canceled if the main question is blocked | Counterevidence and version differences; separate proposal or explicit noncompletion of optional work |
| D50–63 | 6 hours | Four hours for recipient checks; at most two for tool migration after core work is complete | Reviewer's source/calculation differences and rerun notes; migration may be canceled |
| D64–77 | 6 hours | Check new material or gaps; update the same question and key assumptions | V2/V1 change record; no new disclosure means record its absence |
| D78–84 | 3 hours | Review process evidence, actual resources, and T12; choose the next stage | Continue, narrow, consider a passive approach, or pause, with reasons |
| D85–90 | 2 hours | Archive, check readability and backups, arrange the next review | Complete index and limits; one priority for the next round |
Any phase can end the question. If D8–21 reveals that public materials lack comparable fields, use the six hours to document searched entry points, gaps, and scope, then narrow the task. Do not write the explanation first and search for evidence to fit it.
D64–77 does not guarantee that the company releases a report then. Recheck registered disclosure sources and outstanding questions, or use another already-public document for a clearly labeled transfer exercise. That is not a new prospective update for the target company. A deadline cannot create missing evidence.
For an illustrative start, D1 = November 2, 2026 makes D30 = December 1, 2026, a possible first stage-review date. Convert D60 and D90 from your own start. These dates create no actual reminder. Review work stays within the weekly checking hour.
Set a cost ceiling without obligating yourself to spend it
The separate ninety-day budget is RMB 600: prospective data 200, tools 100, review arrangements 200, and contingency 100. These are synthetic allocations, not service quotes or purchase requirements. With lawful public data, existing tools, and free reciprocal review, actual spending may be zero.
Before paying, identify the missing field or task, acceptance evidence, permitted use, and access to necessary records after cancellation. Better features or a discount do not automatically justify taking the review allocation. The contingency is not money that must be spent on the final day. Subscription obligations beyond ninety days belong in the plan too.
Research expense is not household strategy capital or a promised investment to be recovered through returns. Assess whether it obtained missing lawful material, corrected an error, or reduced maintenance demands before renewing. A short-term market gain or loss cannot establish that the RMB 600 was well spent.
Keep one entry point and retain earlier versions
Give each project a stable ID. Its entry page identifies the question, scope, state, last review, and next action. Organize original-source locations, processed fields, calculations, judgment versions, and events beneath it. Retain retrieval time and old inputs when a disclosure changes, and identify affected outputs when tools or models change.
Folder count is not the objective. A conclusion should lead back to source locations and calculations; a revision should connect old judgment to new evidence. AI summaries can support working notes without replacing source identity, numbers, or gaps. Apply G59's minimum-necessary treatment to sensitive information.
CFA Institute's record-retention standard includes supporting research and decisions, including reviews that do not change a holding.2 We borrow traceability, not a universal personal legal retention period or a claim that complete files satisfy every professional standard.
Attempt tool migration only after core outputs can be exported. Use the same inputs, compare output, timing, and cost, and retain the option of returning to the earlier version. Rebuilding the whole software environment while researching can leave a functioning launch screen and no research deliverable.
Review by date and by changed facts
Periodic review handles easily postponed questions: Is the question clearer? Are time or expenses over budget? Did missing material arrive? Is the next action worthwhile? D30/D60/D90 pages identify which T12 task gained evidence, rather than substituting greater confidence for demonstrated ability.
Event triggers include corrected disclosures, changed contract or cash assumptions, new necessary household spending, data/order exceptions, or costs beyond the researched range. Verify the source, then determine the affected layer. Do not wait for the calendar review to investigate a time-sensitive risk.
A fashionable headline does not automatically restart the whole project. Material unrelated to the current question enters a candidate list. When a new task takes priority, state which existing task is canceled or delayed; “just a quick look” cannot become an unbudgeted second project.
Research review also is not a trading instruction. G50 separates allocation drift from changed targets. Changed household purposes first return to policy; drift under an unchanged policy follows its registered rebalancing rule. Investor.gov connects horizon and loss conditions to allocation and explains how changing holdings affect weights.3 Opening the research record weekly does not require weekly trading.
Stopping can complete a judgment
Unavailable required public fields, irreconcilable definitions, implementation costs removing the original rationale, or an exhausted budget can end the round. Preserve searched sources, unsupported conclusions, and conditions for reopening. Ten hours already spent do not themselves justify ten more.
Evidence can justify stopping too. If opposing material defeats the hypothesis, continued searching may merely seek reassurance. A new question receives a new version and budget. Do not turn a failed prediction into a successful explanation by quietly changing the task.
Continue active research only if question, material, and maintenance remain compatible. Lower frequency or fewer objects can reduce demands while preserving checks inside the narrower scope. A passive approach still needs its scope, fees, and risks examined; lack of time does not justify blindly choosing a product.
Pausing new research and discretionary trading can retain necessary management of existing holdings, payments, and exceptions. A pause is not completed liquidation, a risk-free account, or the end of learning. Record review and restart conditions more specific than waiting for a favorable market.
Halve the time instead of moving every task into the night
Transfer. At the end of D28, the first four weeks have used 12 hours and no expense has actually occurred. From D29 onward, only 1.5 hours per week are available, and the final six days receive only one hour. Keep the original start and D90 endpoint. Calculate capacity before canceling tasks.
D29–84 contains eight full weeks, giving 12 hours, plus one final hour: 13 future hours. Add the 12 already used for 25 total, which is 13 fewer than 38. Relabeling the weekly allowance cannot preserve the original remaining 26-hour workload.
Use this checkable replacement:
| Revised phase | Time | Preserved work |
|---|---|---|
| D29–42 | 3 hours | Finish the same question's V1 and key calculation |
| D43–56 | 3 hours | Examine one consequential counterargument |
| D57–70 | 3 hours | Limited-scope recipient check |
| D71–84 | 2 + 1 hours | Update materials, then make the stage decision |
| D85–90 | 1 hour | Archive |
That totals 13 future hours, with no hidden evening. Reconcile the cancellations against the old remaining tasks: keep V1's remaining 3 hours; reduce counterevidence 6 → 3, saving 3; recipient work 6 → 3, saving 3; updating 6 → 2, saving 4; stage decision 3 → 1, saving 2; and archive 2 → 1, saving 1. Savings total 13.
The original two hours of optional reading and two hours of migration are canceled first. Those four hours are included in the thirteen, not additional savings. The other nine come from narrower counterevidence, review, updating, and archive coverage. Leave uncompleted coverage unverified in T12; the smaller output is not a pass for the original complete independent company project.
Cancel prospective data 200 and tools 100 too, retaining review 200 plus contingency 100: a new cap of RMB 300. Because the exercise says nothing has been spent, no refund is invented. Actual nonrefundable spending would remain in the ledger.
Explain. Why can canceling optional reading and migration represent good execution?
Reference answers · reveal
Answer. The revision responds to real capacity, protects core evidence and responsibility, and honestly narrows acceptance scope. More files, trades, or tool familiarity are not automatically more valuable than one completed consequential check.
Assess what changed during the ninety days
Return to the original gap. Can you locate the required disclosure, explain a previously misunderstood field, recalculate a key result, preserve opposing evidence, and let a recipient identify the first disagreement? Which tasks changed from unverified to supported, and which remain beyond your responsibility?
G58's four log entries may contain no real trades. With no applicable new information, record no applicable opportunity. Do not manufacture four transactions or count four revisions to one event as four independent forecasts. A loss does not prove learning failed, and a gain does not verify sources, implementation, and judgment.
You can describe a process change from missing to complete fields. Without suitable comparison and identification, do not attribute it causally to the tutorial, AI, or software, or translate it into a learning-investment return. The system makes changes visible and correctable; it does not promise more profit every round.
Save your own ninety-day plan
Choose one route from existing work. Record question, current evidence, excluded scope, phase hours, cost ceiling and actual spending, data/tool versions, outputs, periodic review dates, and event triggers. Predefine a stopping condition for an unproductive study and conditions for reconsidering continued research, a passive approach, or a pause.
Test it with three actions. Add the phase hours without looking at the attractive title. Assume a critical disclosure never appears and identify the phase allowed to end the question. Then choose one canceled task and state its saved time and remaining unverified scope. A plan that answers all three can adapt to reality instead of merely changing a date.
A sustainable practice preserves continuity: one worthwhile gap, evidence that another session can pick up, facts that can change the plan, and questions that can end. At D90, the useful result is knowing why the next hour deserves your time, not merely owning a longer task list.
References
- Gollwitzer: Implementation Intentions—Strong Effects of Simple Plans, American Psychologist 54(7), 1999, 493–503, original paper hosted by the NYU research group. Situation/action planning, not evidence of optimal ninety-day or three-hour schedules.
- CFA Institute: Standard V(C)—Record Retention, updated April 2024. Supporting records and decisions, without extending every professional requirement to individuals.
- Investor.gov: Asset Allocation and Diversification, horizon, loss conditions, and allocation changes; no supplied budget or product recommendation.
Editorial status
AI-assisted draft synchronized to the accepted second-round Chinese source. Jianning's time/cost schedules are synthetic editorial examples, without real reader results, reminders, or a universal investment plan. See the author review. Independent AI full-text and bilingual review is complete. Human financial review and target-reader testing have not been performed and are not scheduled for this edition.