The Planner: From the First List to the Deposit

Being redesigned (2026-10-09): the owner's review of the build moved the plan to the main navigation, cut the six stages to three tabs, removed the ranking, and added numbers-first sorting, a scores coach, and a family calendar. See redesign/; the stage specs below describe what was built in PR #103.

Overview of the planner specs (not a work item itself). Written 2026-10-07 from the owner's brief ("help students get organized and stay on track, with parents able to see how they're tracking"), the built accounts, household, and list work, the planned planning tools, and new research. Each spec below is a separate work item on the roadmap. The planner is meant to be the site's primary paid feature.

Why

The site can tell a family a great deal about any college. It can't yet help them do the eighteen months of work between "we should start looking" and "the deposit is paid": build a list, sort it honestly, decide who gets the one binding early application, follow and visit the colleges, keep a hundred dates straight, apply, read the decisions, compare the offers, and choose. Families do this today in a spreadsheet, a group chat, and a parent's memory. The planner is that spreadsheet, filled in from what the site already knows about every college, with the parent's view built in rather than bolted on.

Three things make it worth paying for and hard to copy:

• It starts from data the family already trusts here. Every date, admit rate, early-round rule, aid form, visit link, and social account on the plan comes from the college's own documents or the federal files, cited with its edition like everything else on the site. Scoir and Naviance list deadlines; neither says why a step matters at this college ("Tufts says interest is important; Harvard says it isn't considered").
• The household already exists. Parents and students are in one household with the privacy rules a planner needs: a parent sees the list and the plan and never the student's private notes; a student never sees the parent's finances (accounts.md). Common App has no parent view at all.
• One flow, start to finish. List → rounds → actions → timeline → applications → offers → the choice, each stage writing facts the next one reads. The award-letter work that was planned on its own now sits where it belongs, at the end of that flow.

The journey

The planner is one page per student (the Plan tab beside List and Numbers on /household/[person], model.md) that walks six stages. A family can jump around; the stages are a map, not a gate. Stage 0, the student's numbers, is already built (student-profile.md).

Research (2026-10-07)

• The season in numbers. Common App's 2025–26 end-of-season report: 1.53 million first-year applicants and 10.8 million applications across 1,146 members, 7.06 applications each, up 6%; through December 1 (the early rounds), 1.16 million applicants, up 4%, with growth fastest at colleges admitting 25–49%. The Princeton Review's 2025 College Hopes & Worries survey (9,317 respondents, a quarter of them parents): 73% report high stress about applications. The planner's job is to take the stress out of remembering, which is the part software can do.
• Early rounds are where the plan matters most. For the class of 2030, early decision admit rates ran two to six times the regular rate at selective colleges (Brown 14.8% ED against 3.5% RD; Dartmouth about 4.6×; Vanderbilt about 14% against 3%; Northwestern about 20% against 6.5%; from the colleges' announcements compiled by IvyWise, Oriel Admissions, and Dewey Smart). About 30 colleges offer ED II with January 1–15 deadlines and mid-February decisions, after ED I results arrive, so a student gets one more binding try; Washington University reported 27% across its two binding rounds against 9.3% outside them. The caveats the site's early-decision strategy already states still hold: hooked applicants are concentrated in early pools, ED is binding, and an early application doesn't turn a Reach into a Target.
• Demonstrated interest. Roughly 40–45% of colleges consider it, concentrated at small and mid-sized private colleges managing yield; the most selective mostly don't (Harvard, Princeton, Yale, Columbia, MIT, Caltech, UCLA say "not considered"). It is disclosed per college in CDS C7, which the site stores (cds-admissions.md), so the planner can say per college whether a visit, a follow, or an information request counts, instead of telling everyone to do everything.
• One-click actions, honestly. X documents a follow Web Intent (x.com/intent/follow?screen_name=) with an inline sign-in; YouTube's ?sub_confirmation=1 opens a subscribe prompt for a signed-in viewer. Instagram, TikTok, Facebook (its follow button was retired in 2018), and LinkedIn offer no follow intent: the best available is opening the profile, in the app on a phone, with a follow button one tap away. The planner does the real one-click where it exists and the one-tap-then-confirm elsewhere, and never pretends otherwise (actions.md).
• Reminders work when they're concrete. Castleman and Page's randomized trials (Dallas, Philadelphia, Boston; 4,882 graduates) sent texts about specific, dated tasks (file the FAFSA, take the placement test, register for orientation) and raised fall enrollment 3.1 points (68.0% against 64.9%), 5.7 points for the lowest-income students; messaging parents as well as students did not measurably add to messaging students alone. The lesson the planner takes: one task, one date, one link, a few at a time; parents get visibility and a way to help, not a second stream of the same reminders (parents.md).
• May 1 isn't the end. Since NACAC's 2019 consent decree with the Department of Justice removed the rules that made May 1 final, colleges may recruit students who have deposited elsewhere, and wait lists move through the summer. The plan therefore has a real "committed" step with deposit and withdrawal tasks, and a summer list (orientation, housing, final transcript, aid verification), rather than ending at the decision.
• Comparable tools. Scoir (parents see the list, suggest colleges, see upcoming visits), Naviance (structured tasks for transcripts and recommendations, counselor-driven), Common App (each college's requirements and deadlines once added; no parent accounts), College Kickstart (balance grades, an early-admission engine, an action plan; no family accounts), Appily (deadline management, scholarships, virtual tours). None combines a family account, a cited reason per step, the early-round decision with its money question, and offers at the end.

What exists, what moves, what's new

New: model.md, list-building.md, early-rounds.md, actions.md, applications.md, parents.md. Early Decision and Early Action: Does Applying Early Help? stays in the planning tools as the data and the profile section; stage 2 is the planner built on it. Chances and Fit: Where You Stand stays too; stage 1 reads its standing when it exists and uses the admit rate alone until then.

Decisions in short

Each is argued in its spec.

1. One list, seen through six lenses. The planner adds columns and tables around list_items; it never copies the list. Ticking "applied" on the timeline, in the applications stage, and on the list row are one fact.
2. Dream is a flag, not a category. A Dream can be a Reach or a Target; it is the college the student would choose over all others today, one per list, and it is what stage 2 looks at first.
3. Rounds are proposed, never assigned. The site shows availability, dates, advantage, conflicts, and the money question, proposes a plan, and the student edits it. No sentence says "apply ED here".
4. Tasks are generated, idempotent, and owned by a person. Each task has a source (the college's date, the cycle, the stage, the family), an assignee (student, parent, either), and a key, so regenerating after a data publish moves a date without losing a tick. Parent tasks (FAFSA, CSS Profile, deposits) show on the parent's own page.
5. Do the action where the network allows; record it everywhere. A real follow intent on X and YouTube; a deep link plus "did you follow?" elsewhere. What's stored is the fact and the date, never a token or a login.
6. Offers start as a form; letters are read later. The family enters each offer in the College Financing Plan layout and is asked whether they'd share the letter itself. The shared letters become the set that a later upload-and-read path is built and measured on; until then nothing is extracted by a model.
7. Parents see, nudge, and own their tasks; they don't grade. No scores, no "behind", no comparisons between siblings. A nudge is one line about one task, rate-limited, and the student sees who sent it.
8. Reminders reach the phone. Email and text, each with its own switch, texts with explicit consent; one task, one date, one link.
9. The commercial model comes after the build. The planner is meant to be the primary paid feature, but these specs name no tier and gate nothing; once the stages exist the owner draws the paid line in Commercialization: Free, Plus, and Pro and the stages pick it up through one hook (model.md).

Owner decisions (2026-10-07)

From the owner's review of the first draft:

• No commercial model in the specs. Decide tiers once things are built; the tier table that was here is gone.
• Texts are in scope. Reminders and nudges by SMS as well as email (timeline.md).
• No photos on visits.
• Letters: form first. The owner has no letters to pilot a reader on; the form is the first release and asks families whether they'd share the letter, which builds the set for the reader later (offers.md).
• Common App screenshot import: later, not this pass.
• Twilio is the text provider (timeline.md); number registration starts before the timeline unit is built.
• Thirty shared letters across several colleges, admission letters included, is the threshold for building the letter reader as its own roadmap item (offers.md).

Build order

model ──► list-building ──► early-rounds (after early-decision-strategy for the advantage data)
   │           │
   │           ├──► actions
   │           └──► timeline ──► applications ──► offers (after net-price-estimator for the four-year math)
   │                    │
   └────────────────────┴──► parents

The model first: tables, the stage machine, the Plan tab's frame, and entitlement hooks. Then the list stage, which is small and makes the tab useful on day one. Early rounds and actions are independent of each other; the timeline needs the rounds (its dates follow the chosen round) and feeds applications and offers. Parents last, because it is a view over everything else, though its summary line should ship with the model so the hub shows the planner from the first release. Build with the build-roadmap-section skill, one unit per spec.

Shared rules

In addition to the product rules for user data:

• Every date names its source and edition; a date a person typed says "your date". No invented dates: a college that hasn't published one gets "the college hasn't published this; add your own".
• Nothing in the planner is a verdict. Categories are the student's choice with the site's suggestion beside it; rounds are proposed; standing is a classification with reasons; offers are compared on facts. No composite score, no letter grade, no probability (Ideas: Fresh Directions From the Competitive and Cross-Vertical Research).
• Interpret a step only where the college's data allows it. "Interest is considered here" comes from C7; where the college says it isn't, the step is optional and says why.
• A parent sees what the privacy model allows and nothing more: lists, plans, tasks, visits, offers; never private notes, hooks, or the finances of another guardian.
• Minors and messaging. Reminders go to the account's email and, with explicit consent (a guardian's for a student under 18), by text to the phone on the household record. No reminder or text carries a personal number (GPA, scores, income, award amounts): "Michigan: essay due Nov 1".
• Telemetry records that a stage was used and which college, never the content (telemetry.md).

Open questions for the owner

None at the overview level as of 2026-10-07; each spec keeps its own small ones. The two that were here (the text provider, the letter threshold) were decided the same day and are recorded above.