PlanRetirement.ai

1. Executive Summary

PlanRetirement.ai is a complete lifecycle financial planning platform — career through retirement, tax-aware, AI-assisted, free — built on a privacy architecture that makes it impossible for the operator to read individual user data. Its foundation is a purpose-built retirement spending index that replaces the inflation assumption every other tool gets wrong.

The product spans the entire financial life of a household. On the accumulation side, it models industry-specific wage growth, equity compensation, account-aware contributions across every major tax-advantaged vehicle, and career-interruption scenarios from sabbatical to caregiving break. On the decumulation side, it handles Social Security optimization with full spousal, survivor, and divorced-spouse modeling; withdrawal strategies from the 4% rule through Guyton-Klinger guardrails and floor-and-upside; Monte Carlo simulation with regime-switching; and pension treatment including the often-overlooked purchasing-power erosion of non-COLA private pensions. A unified tax engine handles Roth conversion optimization, IRMAA bracket cliffs, ACA subsidy thresholds, the Net Investment Income Tax, AMT for ISO exercises, RMDs under SECURE 2.0, and income tax across all 50 states plus DC — modeled as a single network of interacting decisions, not isolated calculators. An AI conversational layer fields plain-English scenarios and personal explanations of projections without ever generating a number itself.

What makes this rigor possible is the inflation engine underneath. The Consumer Price Index — the figure that anchors virtually every retirement projection in the United States, and the single most consequential assumption in any retirement plan — was designed to measure the cost of a representative urban worker's basket of goods. It was not designed to measure what retirees actually spend on, and its methodology systematically understates the inflation experienced by people who are no longer working. A 1% underestimation compounded over 30 years produces a 35% real wealth gap; CPI's methodological bias against retirees often runs two or three percentage points across the categories most relevant to late retirement.

The Personalized Retirement Spending Index addresses this directly: a pre-1980 fixed-basket methodology that does not substitute cheaper goods, does not apply hedonic quality adjustments, and does not treat housing through the abstraction of Owner Equivalent Rent. It models three distinct retirement phases — Active (65–75), Slower (75–85), and Care (85+) — with category weights that shift as households age. It is personalized to geography, housing status, and Medicare flavor, with optional further refinement for lifestyle, health, and anticipated moves. The measured base — the category price growth itself — comes from the FixedBasket index, an independent, open-methodology fixed-basket index that PlanRetirement also operates; PlanRetirement adds the lifecycle re-weighting and personalization on top.

The product is free. There is no advertising, no lead generation to financial advisors, no data sale, and no monetization mechanism that competes with the user's interest. The zero-knowledge encryption architecture means the operator cannot read any individual user's salary, balances, projections, or scenario data — not as a policy choice, but as an architectural impossibility. Communications with users are limited to broadcast newsletters and product updates; nothing about any individual is visible to the operator. The methodology is documented publicly, citable, and open to outside scrutiny.

This paper explains the foundation (why CPI fails for retirees), the methodology (how the index works), the lifecycle engine and tax-aware optimization layered on top of it, the privacy architecture, and the principles that govern how the product is operated.

2. The Problem with CPI for Retirement

The Consumer Price Index has three properties that make it broadly defensible as a macroeconomic statistic, and three properties that make it deeply unsuitable as a planning input for retirees. They are the same three properties.

2.1 Substitution

When steak becomes expensive, the modern CPI assumes households switch to chicken. This is the substitution principle, and it has been a core feature of CPI methodology since the early 1980s. It reflects a real behavioral pattern, and as a measure of actual consumer spending it is defensible.

It is indefensible as a measure of purchasing power. The retiree who used to eat steak twice a week and now eats chicken has not been compensated for inflation. They have been forced to accept a lower standard of living. The substitution itself is the loss. By assuming away that loss, CPI hides it.

For a worker still earning, substitution may be a choice. For a retiree on a fixed income, substitution is a constraint. A planning methodology that uses substitution-adjusted inflation tells the retiree the cost of accepting the downgrade, not the cost of avoiding it. This is precisely the wrong number for someone trying to plan a retirement they actually want to live.

2.2 Hedonic Adjustment

When a new computer offers twice the processing power at the same price, the modern CPI records a price decrease. The product is treated as having improved in quality, so the per-unit-of-quality cost has fallen. This is hedonic adjustment, applied to many categories of goods since the late 1990s.

For a household replacing a working computer with a new working computer, hedonic adjustment is reasonable. For a retiree who simply needed the old computer to work, the new computer is more expensive in dollars and they receive no benefit from the additional processing power. The "quality improvement" is invisible to them; the higher dollar cost is not.

Hedonic adjustments are now applied to electronics, appliances, vehicles, healthcare equipment, and a growing list of other categories. In aggregate, they suppress measured inflation by several tenths of a percentage point per year — a small-sounding figure that compounds into a meaningful gap over a thirty-year retirement.

2.3 Owner Equivalent Rent

Housing is the single largest category in most household budgets and the single largest distortion in CPI. Rather than measure the actual cost of buying or owning a home, CPI uses Owner Equivalent Rent — an estimate of what homeowners would pay if they were renting their own home. This abstraction means that during the housing booms of the 2000s and 2020s, when home prices rose dramatically, the housing component of CPI rose modestly.

For renters, this understates the actual rental inflation they face. For prospective buyers — including pre-retirees planning to downsize or relocate — it dramatically understates the actual market reality. For homeowners who pay property taxes, insurance, and maintenance on the real value of their property, none of which scale to "equivalent rent," it also misses the actual cost trajectory.

2.4 The Cumulative Effect

None of these adjustments is unreasonable in isolation; each was introduced for defensible technical reasons. But their cumulative effect on a retirement plan is significant and one-directional. They all suppress measured inflation. They all make planning tools that use CPI more optimistic than they should be.

The Bureau of Labor Statistics publishes its own experimental index, CPI-E, intended to measure inflation as experienced by Americans 62 and older. CPI-E has historically run about 0.2 percentage points above CPI-U. But CPI-E uses the same methodology — substitution, hedonic adjustments, OER. It changes only the basket weights to reflect what retirees buy. It does not address the methodological bias.

2.5 The Shape Problem

Even setting methodology aside, every existing planning tool we have studied applies a single, flat inflation rate across a user's entire retirement — typically 2.5% or 3%. This is wrong in a second, distinct way: retirees do not have a single spending pattern across retirement. They have at least three.

A 67-year-old and an 87-year-old face different inflation rates because they buy different things. The categories that dominate the "Care" phase — long-term care, skilled nursing, specialized healthcare — are precisely the categories with the highest sustained inflation in the U.S. economy. A flat 3% assumption is wrong by a small amount in the Active phase and by a catastrophic amount in the Care phase.

3. A Purpose-Built Retirement Spending Index

The PlanRetirement.ai inflation engine is built in two layers. The measured base is the FixedBasket index — an independent, open-methodology fixed-basket index that measures the price growth of a constant bundle of real goods directly from primary sources, with no CPI input, no substitution, and no hedonic adjustment. FixedBasket is a sister project that PlanRetirement also operates; its full methodology is published openly and is reproducible from frozen inputs. On top of that base, PlanRetirement adds the structural recognition that retirees pass through distinct spending phases — re-weighting the measured categories across the lifecycle and personalizing them to the household. Together these define what we call the Personalized Retirement Spending Index, or PRSI: FixedBasket measures the rates; PlanRetirement applies them to a life.

3.1 The Fixed-Basket Principle

The PRSI measures what it costs to maintain a household's original basket of goods. When prices in a category rise, the basket does not change — the measured cost of the basket rises. This is the methodology BLS used before 1980, before substitution and geometric weighting were introduced, and it is the methodology used by serious historical inflation researchers who want to measure purchasing power directly.

This is not a rejection of substitution as a real-world behavior. Households do substitute. But the substitution is the cost. By measuring the un-substituted basket, the PRSI tells the user the true scope of the inflation they face, not the smaller number that represents the inflation they could not avoid by downgrading.

3.2 No Hedonic Quality Adjustment

If a category's prices rise in dollars, the PRSI records the dollar increase. It does not adjust for changes in product quality. A retiree replacing a refrigerator pays whatever the current refrigerator costs; they do not receive a discount for the fact that today's refrigerators are more energy efficient than the one they bought in 1995.

This is the principle that most clearly separates the PRSI from official inflation measures. Hedonic adjustments lower official CPI inflation by approximately 0.3 to 0.6 percentage points per year, depending on the categories most affected. Removing them brings measured inflation closer to the dollars-out-of-pocket reality the household actually experiences.

3.3 Direct Housing Cost

The PRSI uses direct housing cost measures rather than Owner Equivalent Rent. For owners with a fixed mortgage, the relevant cost is property taxes, insurance, maintenance, and major capital expenditures; for owners with a paid-off home, taxes and maintenance only; for renters, actual market rent. The FixedBasket shelter strata capture this directly — market-rent series and property-tax-based owned-operating costs, not the OER abstraction. For households planning a future home purchase or downsize, the engine applies the projected market at the planned location.

This produces a much higher housing inflation number than CPI's OER-based figure in most years and most geographies. It is also a much more accurate predictor of what a retiree will actually pay.

3.4 The Three-Phase Lifecycle Model

A retiree at 67 and the same retiree at 87 face two materially different inflation environments because they buy different things. The PRSI explicitly models three phases, with category weights that shift across the retirement lifecycle:

Phase 1 — Active Retirement (ages 65–75)

Phase 2 — Slower Retirement (ages 75–85)

Phase 3 — Care Retirement (ages 85+)

The weights are derived from the Consumer Expenditure Survey, the Health and Retirement Study, and Medicare claim data, cross-validated against actuarial industry assumptions used in retirement income product pricing. They are revised on a documented cadence with public version history.

Note: in the current engine these per-category rates are measured by the FixedBasket index (see §3.5); the reference proxies above indicate the kind of data each category reflects, not the live source. Long-term care, assisted-living, and facility costs are not part of the fixed-basket index and remain PlanRetirement overlays sourced from Genworth Cost of Care data.

3.5 The Methodology Overlay

Earlier versions of the PRSI estimated category inflation by applying a methodology-correction overlay to CPI-derived proxies — adjusting each category for the difference between BLS's pre-1980 fixed-basket methodology and current methodology. That step was the most contested in the entire approach, and it is no longer used.

The current engine sources its measured base from the FixedBasket index, which constructs a true fixed-basket Laspeyres index directly from primary price data and uses no CPI series at any point. In FixedBasket, no-substitution, no-hedonic-adjustment, and market-rent-over-OER are structural properties of the measurement — not corrections layered onto an official index. PlanRetirement consumes FixedBasket's per-category (per-stratum) rates and re-weights them across the three lifecycle phases; there is no correction overlay left to calibrate.

One overlay remains: long-term care. The fixed-basket index does not measure long-term care, so PlanRetirement supplies it separately from Genworth Cost of Care data (with the recency adjustment discussed in Section 11). This is the single component of the blended PRSI that is a PlanRetirement estimate rather than a measured FixedBasket rate.

Because the measured base is now a direct fixed-basket measurement rather than a corrected proxy, the blended PRSI rates are lower than the earlier corrected estimates and vary by household; healthcare and long-term care remain the largest drivers of the gap versus CPI-U. FixedBasket is an open-methodology index that PlanRetirement also operates, so the base rates are independently published, auditable, and reproducible from frozen inputs — the same standard of scrutiny we hold the rest of this work to.

4. Personalization Framework

A national average inflation rate is still wrong for any specific household. A retiree in a paid-off home in Tampa faces a fundamentally different inflation environment than a renter in San Francisco or a homeowner relocating to Manhattan for proximity to grandchildren. The PRSI is designed from the start to produce a household-specific rate, not a national one.

Personalization is layered, with each tier increasing precision at the cost of additional input from the user.

4.1 Tier One — Geography, Housing, Medicare

Three inputs capture roughly 80% of the variance between retiree households:

• Geography — state and zip code, determining the housing market, healthcare cost environment, state tax exposure, and energy prices. Housing, healthcare, and energy are localized through the FixedBasket state-level series; state tax exposure is sourced from the Tax Foundation and state revenue departments.
• Housing status — owner free-and-clear, owner with mortgage, renter, anticipating-purchase. Determines whether the housing inflation rate reflects taxes-and-maintenance, mortgage-plus-taxes, market rent, or projected purchase prices.
• Medicare flavor — Original Medicare with Medigap, Medicare Advantage, or not-yet-decided. Determines the out-of-pocket cost trajectory and the IRMAA exposure structure.

With these three inputs alone, the PRSI produces a meaningfully personalized rate that can differ from another household's rate by several percentage points.

4.2 Tier Two — Lifestyle Modifiers

Optional adjustments to the standard phase weights based on user-declared lifestyle preferences:

• Travel-heavy retirement (raises travel weight in Phase 1, extends into Phase 2)
• Homebody retirement (lowers travel, raises home-related)
• Urban-versus-suburban-versus-rural retirement (affects transportation, energy, healthcare access)
• Philanthropic spending tier (adjusts the "other" allocation)
• Anticipated LTC strategy (self-insure, LTC insurance, family caregiving, Medicaid planning)

4.3 Tier Three — Full Per-Household

For users who want maximum precision, the PRSI accepts:

• Health status (broad buckets — affects expected utilization and life expectancy)
• Family longevity (affects expected duration in each phase)
• Specific anticipated geographic moves (re-projects phase weights through those transitions)
• Anticipated changes in housing status (renting in early retirement, buying in mid-retirement, downsizing in late retirement)
• Detailed Medicare and supplemental plan selection
• Whether a long-term care insurance policy is in force, and its terms

The principle: defaults are reasonable, personalization is opt-in, and every increment of additional input meaningfully refines the projection. The user is never asked for information that does not change the answer.

5. From Index to Engine: Lifecycle Modeling

The PRSI is the centerpiece of the product but not the entire product. It is the connective tissue that allows PlanRetirement.ai to model a household's complete financial life — accumulation and decumulation — under a single, coherent, retiree-specific inflation lens.

5.1 The Career Side (Accumulation)

Standard career calculators use a flat 3% inflation assumption to project both salary growth and retirement target. This is wrong on both sides of the equation. Wage growth varies by industry and occupation in ways that BLS publishes annually but no consumer tool consumes. And the "retirement number" — the savings required to fund a desired retirement — depends entirely on a retiree's actual cost basket, not on the urban worker's basket used for CPI.

PlanRetirement.ai models the accumulation phase with:

• Industry-specific wage inflation pulled from BLS Occupational Employment Statistics by SOC code
• Account-aware contribution modeling across 401(k), Roth 401(k), Traditional IRA, Roth IRA, HSA, taxable brokerage, with employer match, vesting schedules, and contribution limits including catch-up provisions
• Equity compensation modeling for RSUs, ISOs, NQSOs, ESPP, including vesting cliffs and AMT exposure
• Career interruption scenarios — sabbatical, parental leave, caregiving breaks, layoffs, industry pivots — with downstream effects on Social Security earnings record and tax-advantaged account contributions
• Tax projections under current law and under user-toggled scenarios (TCJA sunset, hypothetical Social Security reforms, hypothetical Medicare benefit changes)

The retirement number generated by the accumulation engine is anchored to the PRSI. As the user's location, housing status, or Medicare flavor changes, the target updates in real time. A user moving from San Francisco to Tampa at age 70 sees their required savings drop. A user staying in San Francisco as a renter sees it climb. No other career calculator surfaces this.

5.2 The Retirement Side (Decumulation)

The accumulation-to-decumulation transition is the most mishandled phase in every existing tool, and the one where high-value decisions concentrate:

• The Social Security claiming decision (62 vs. 67 vs. 70)
• The Roth conversion ladder window, often the lowest-tax-rate years of a household's life
• The healthcare bridge before Medicare eligibility
• The tax-efficient withdrawal sequencing decision (which account to draw from when)
• The Medicare flavor selection at 65
• The IRMAA bracket management implied by withdrawal and conversion choices
• Required Minimum Distributions starting at 73 (and moving to 75 under SECURE 2.0)
• The phase-specific shape of spending under the three-phase model

Each of these decisions is modeled explicitly, with their interactions visible to the user. A Roth conversion in 2030 raises MAGI in 2032 which raises Medicare premiums via IRMAA — a two-year lookback that most tools either ignore or model only partially. PlanRetirement.ai surfaces these interactions as a unified joint-optimization view, not as separate isolated decisions.

5.3 The Hero Visualization

The unifying surface is a single horizontal timeline spanning the user's age from today through approximately 95. On it, multiple layers are visible:

• Wealth trajectory — net worth rising through accumulation, peaking, declining through decumulation
• Income trajectory — salary in working years, Social Security and portfolio withdrawals in retirement
• Cost-of-life trajectory — projected annual cost using the PRSI
• A ghost line showing what the same projection looks like under CPI assumptions — the gap between the two cost lines, especially at age 85, is the invisible inflation tax that every other tool hides
• Phase bands (background tints) marking the four spans: accumulation, Active retirement, Slower retirement, Care retirement
• Draggable waypoints for major decisions ("retire at 67," "claim SS at 70," "move to Florida at 72," "LTC need at 86")

Dragging any waypoint re-projects the entire chart in real time. This is the surface that holds the user's attention, because it is the surface on which the planning decisions actually happen.

6. The Product: What We Build and Where We Lead

PlanRetirement.ai is a comprehensive lifecycle planning tool, free to use, with the depth of feature set that serious users require — plus a set of capabilities no existing competitor offers.

6.1 Capabilities We Match

The product matches the table-stakes feature depth users have come to expect from the most capable existing tools (Boldin, ProjectionLab, MaxiFi, WealthTrace):

• Account-level modeling across every account type a retiree might hold
• Federal and state tax modeling for all 50 states plus DC, with IRMAA, ACA subsidy cliffs, NIIT, AMT, and Social Security taxation
• Roth conversion optimization with multiple goal frameworks (lifetime tax, estate value, bracket-filling)
• Social Security with full optimization — claiming age, spousal, survivor, divorced-spouse, WEP, GPO, earnings test
• Pension modeling with COLA-vs-non-COLA, joint vs. single life, lump sum vs. annuity
• Healthcare modeling — ACA marketplace, all Medicare parts, IRMAA with two-year lookback, Part D out-of-pocket cap
• Long-term care modeling with insurance vs. self-insure decision tools
• Monte Carlo simulation with multiple modes (historical sequences, parametric, block bootstrap, regime-switching)
• Multiple withdrawal strategies (4% rule, Guyton-Klinger, bond tent, bucket, VPW, dynamic, floor-and-upside)
• Scenario modeling with side-by-side comparison and AI-driven plain-English scenario creation
• Estate planning basics including the SECURE Act 10-year rule for inherited IRAs
• Account aggregation via Plaid, with client-side encryption preserving zero-knowledge
• Conversational AI assistant for explanation, scenario simulation, and education

6.2 Capabilities Where We Lead

The capabilities below are not currently offered by any consumer retirement planning tool we have studied. Each represents a meaningful gap in the market and a defensible competitive position.

• Personalized retirement spending index — geography, housing, Medicare-flavor-adjusted. No competitor has anything comparable.
• CPI ghost-line visualization — every projection is shown against the projection a CPI-based tool would produce. The gap is the entire thesis of the product.
• SS COLA versus real basket gap — Social Security uses CPI-W for its annual COLA. The real purchasing power of a Social Security check, measured against an actual retiree's basket, declines steadily through retirement even after COLA. PlanRetirement.ai is the only tool that can show this gap because it is the only tool with a personalized retiree inflation rate to measure against.
• Non-COLA pension erosion visualization — same treatment for users with corporate defined-benefit pensions.
• Three-phase lifecycle inflation — Active, Slower, Care, with shifting category weights. Some tools allow per-category inflation overrides; none publish a structural model.
• Zero-knowledge encryption — the operator literally cannot read individual user financial data. Incumbents structurally cannot match this because their business models depend on aggregating user data.
• AMT-aware ISO exercise modeling — for tech and finance workers, often a critical and badly-handled decision.
• Industry-specific wage inflation from BLS Occupational Employment Statistics, replacing the universal 3% assumption.
• Five-year Roth conversion clock modeling — separately for each conversion. Notably absent from Boldin's well-regarded Roth Conversion Explorer.
• Asset location optimization — which assets belong in which account types for tax efficiency.
• MSA-level long-term care cost data with recency adjustment over the standard Genworth survey, which has lagged post-2021 reality.
• Joint Social Security and Roth optimization against IRMAA and ACA cliffs — a single unified optimization rather than isolated decisions.
• Caregiving career interruption modeling — the multi-year financial impact of leaving the workforce to care for family, with the long-tail effects on retirement savings and Social Security earnings record.
• Federal law uncertainty toggle — switch between projections under current law (e.g., TCJA sunset) and current policy.
• Calm-first design — no fear tactics, no urgency manipulation, honest uncertainty over false precision.

The combination is the defensible position. No existing tool can match all of these without rebuilding from a fundamentally different foundation.

7. Tax-Aware Planning

Tax modeling has become the primary competitive battleground in retirement planning software through 2025–2026. Every serious tool is racing to add depth. PlanRetirement.ai approaches tax planning from a different angle than most competitors: we treat taxes as a network of interacting constraints, not a set of independent calculations.

7.1 The Interaction Map

A single decision in a retiree's plan typically cascades through six or more tax surfaces:

A Roth conversion in 2030 raises taxable income in 2030, which determines MAGI for 2032 Medicare premiums via IRMAA's two-year lookback. The same conversion, if it pushes MAGI over the Net Investment Income Tax threshold, applies a 3.8% surtax on investment income. If the household is also receiving ACA marketplace subsidies, the same MAGI determines subsidy clawback. If state of residence imposes income tax, the conversion is taxed there too. If the conversion is large enough, it pushes ordinary income into a higher federal bracket, which may also push long-term capital gains from the 0% bracket into the 15% bracket. And the timing of the conversion relative to the five-year clock determines whether Roth withdrawals are penalty-free.

This is one decision with seven tax interactions. The accumulation of these decisions across a retirement is the substance of tax-aware planning. Most existing tools handle each surface in isolation; the joint optimization is where retiree households leave the most money on the table.

7.2 What We Model

• Federal income tax — bracket-by-bracket, current and projected, with TCJA sunset awareness
• State income tax for all 50 states plus DC, including state-specific retirement income exemptions (PA exempts retirement income entirely, IL exempts pension and IRA distributions, several states exempt Social Security)
• FICA, including Additional Medicare Tax on earned income above thresholds
• Capital gains, short-term vs. long-term, with qualified dividend treatment
• NIIT (3.8% surtax) with proper MAGI threshold modeling
• AMT with explicit handling of ISO exercise events
• IRMAA brackets for Medicare Parts B and D, with two-year MAGI lookback
• ACA subsidy cliff with current law (including OBBBA repayment cap changes)
• Social Security taxation with provisional income calculation
• Roth conversion optimization across multiple goal frameworks
• RMD scheduling under SECURE 2.0 (age 73 now, 75 starting 2033)
• QCD (Qualified Charitable Distribution) modeling for RMD satisfaction
• State estate tax for the 15+ states that have one
• SECURE Act 10-year rule for inherited IRAs

7.3 Where We Lead

Three areas where our approach goes beyond what existing tools offer:

• The bracket-cliff visualization. Every relevant cliff — IRMAA, ACA, NIIT, capital gains bracket transitions — is shown on a single chart with the user's projected MAGI overlaid. The user can see immediately when they will cross which cliff and what change in plan would prevent it.
• The five-year Roth clock. Each conversion has its own five-year clock. Most tools treat Roth as a single bucket; we model the multi-clock complexity that determines whether early withdrawals will be penalty-free.
• Federal law uncertainty as a first-class scenario. Users can toggle between "current law" (e.g., TCJA sunset, Social Security trust fund depletion) and "current policy" (extensions assumed) projections. Neither is a prediction; both are honest framings of the uncertainty.

8. Privacy and Data Stewardship

Most financial planning tools have business models that depend on monetizing user data — through advertising, lead generation to financial advisors, sale to third parties, or aggregation for institutional resale. PlanRetirement.ai's commitment is the opposite: the data belongs to the user, the operator cannot read it, and there is no provision in the architecture for changing this later.

8.1 Zero-Knowledge Architecture

User financial data is encrypted client-side with keys derived from the user's password via the Argon2id key derivation function. The server stores only ciphertext. The encryption library and key handling are architected so that the operator literally cannot read any individual user's financial profile, even with full administrative access to the database.

This is not a promise to behave well. It is an architectural constraint on what the operator is capable of doing. Promises can be broken; this capability does not exist to be broken.

8.2 What We Can and Cannot See

8.3 Communications and the User Base

Email addresses are server-readable because users need to be able to log back in. This permits broadcast communications: announcements, product updates, educational newsletters, surveys, and invitations. It does not permit anything that would constitute investment advice without proper regulatory standing.

This distinction matters because the product may one day grow beyond a free planning tool into something with adjacent services. The boundaries are clear: broadcast is permissible, harvesting individual data is not, and any future services will require explicit informed consent through a separate, regulated onboarding — not silent repurposing of the existing relationship.

8.4 The Promise as Architecture

Other tools claim privacy; PlanRetirement.ai architects it. The "How We Protect Your Data" page on the product explains, in plain language, what the encryption does and does not do, and what the operator can and cannot see. The promise is not a marketing claim with a backdoor. The capability to break it does not exist.

9. Mission and Principles

PlanRetirement.ai exists because the most important number in retirement planning — inflation — is the one every existing tool gets wrong. We believe people deserve a financial plan grounded in what their life actually costs, modeled with the same rigor a pension fund applies to its liabilities, and built by a company whose only customer is them.

• 1. Your data is yours. We do not sell, license, or share it. Ever. This is enforced architecturally, not just contractually.
• 2. We earn no revenue from your data. The product is free. There are no ads, no lead generation, no third-party trackers.
• 3. Our numbers are explainable. Every projection drills down to its inputs. Every category in the inflation engine has a documented source. The methodology is public.
• 4. We will tell you what we don't know. Honest uncertainty over confident illusions. The future is uncertain; pretending otherwise is the problem we are trying to solve.
• 5. The plan adapts to you. Defaults are reasonable. Personalization is opt-in. You should never be asked for information that does not change the answer.
• 6. Calm over urgency. Most financial software weaponizes anxiety. We commit to honest framing, calm presentation, and design that respects your time and your peace of mind.

10. Roadmap

The product is built in three stages, with the deliberate constraint that each stage is independently valuable. Users get a useful tool at every step; no functionality is gated behind future-version dependencies.

The product caps at V3. We will not add portfolio recommendations or investment execution. Those are different products operated under different regulatory frameworks, by different companies, and offering them through PlanRetirement.ai would compromise the trust and independence that are the entire point.

11. Open Questions and Honest Uncertainties

It would be easy to present this work as more complete than it is. We are committed to the opposite: open disclosure of where the methodology is contested, where the data is imperfect, and where future work is needed.

11.1 The Measured Base and Its Proxies

Earlier versions applied a contested methodology-correction overlay to CPI-derived proxies. That overlay has been retired: the measured base now comes directly from the FixedBasket index, a true fixed-basket index built from primary price data with no CPI input. The open questions move accordingly — to FixedBasket's own source choices, where a category is tracked by a single representative item or a proxy series rather than exhaustive coverage (apparel via a single denim SKU, food-away omitted, a thin out-of-pocket healthcare basket). These are documented in FixedBasket's published methodology. Because PlanRetirement also operates FixedBasket, we hold the base index to the same open, reproducible, outside-scrutiny standard as the rest of this work — and disclose the shared ownership rather than present it as independent third-party validation.

11.2 The Long-Term Care Data Question

Genworth's Cost of Care Survey is the industry standard for long-term care cost trends. It has also been criticized as conservative relative to post-2021 reality, particularly in skilled nursing and memory care, particularly in high-cost metropolitan areas. We apply a recency adjustment based on supplementary state Medicaid payment data and CMS facility cost reports, but this is an active area of refinement.

11.3 Validation

The strongest validation of an inflation index is back-testing against actual retiree expenditure data. We have begun this work using the Consumer Expenditure Survey and the Health and Retirement Study, but a fully-published back-test with confidence intervals is still in development. We do not consider the methodology validated until this work is complete and externally reviewed.

11.4 What CPI Will Print

The PRSI is not a forecast of what CPI will report. It is an estimate of what a retiree's actual basket will cost. The two numbers will diverge — that is the point. We are not claiming that BLS's reported CPI numbers are wrong as measures of what they purport to measure. We are claiming they are the wrong measure to use for retirement planning.

11.5 The Trust Fund

The Social Security trust fund is currently projected to require a 17–23% benefit cut around 2033–2035 unless legislation intervenes. PlanRetirement.ai models this explicitly as a user-toggleable scenario. We do not predict whether Congress will act. We give users the tools to plan for either outcome.

11.6 Federal Tax Law

Many of the federal tax provisions most relevant to retirement planning — TCJA tax brackets, the estate tax exemption, qualified business income deductions — sunset or change on known dates. We model both "current law" and "current policy" scenarios honestly. Anyone who tells you they know what Congress will do is selling something.

12. References & Sources

Inflation index — the measured base

• FixedBasket index (fixedbasket.org) — the open-methodology fixed-basket index that supplies PlanRetirement's per-category inflation rates. PlanRetirement also operates FixedBasket; the relationship is disclosed rather than presented as independent validation. FixedBasket draws on observed retail prices (BLS Average Price Data), EIA energy, KFF employer health premiums, NADAC and CMS out-of-pocket, market rents (Zillow / Census), IPEDS tuition, and other primary sources — and uses no CPI series at any point. Its full methodology is published and reproducible from frozen inputs.
• Genworth, Cost of Care Survey — long-term-care cost trend. A PlanRetirement overlay, not part of the fixed-basket index (the index does not cover long-term care).

Inflation comparison benchmarks (not used as PRSI inputs)

• Bureau of Labor Statistics, Consumer Price Index program (CPI-U, CPI-W, CPI-E) — shown only as the comparison "ghost line" against the PRSI
• Bureau of Economic Analysis, Personal Consumption Expenditures inflation
• Federal Reserve Economic Data (FRED), Federal Reserve Bank of St. Louis

Housing data

• Federal Housing Finance Agency, House Price Index
• U.S. Department of Housing and Urban Development, Fair Market Rents
• U.S. Census Bureau, American Community Survey

Healthcare data

• Centers for Medicare & Medicaid Services, National Health Expenditure Accounts
• Centers for Medicare & Medicaid Services, Medicare premium and IRMAA bracket data
• HealthCare.gov developer data (ACA marketplace)
• Genworth, Cost of Care Survey (public summary)

Social Security data

• Social Security Administration, Office of the Chief Actuary (bend points, average wage index, mortality tables)
• Social Security Administration, Trustees Report
• Social Security Administration, COLA history

Employment and wage data

• Bureau of Labor Statistics, Occupational Employment and Wage Statistics (OEWS)
• Bureau of Labor Statistics, Employment Cost Index
• Bureau of Labor Statistics, Quarterly Census of Employment and Wages

Tax data

• Internal Revenue Service, Revenue Procedures (annual tax brackets, contribution limits)
• Internal Revenue Service, Publication 590-B (RMD divisor tables)
• Tax Foundation, State Tax Data
• State Departments of Revenue (each of 50 + DC)

Markets and returns

• Robert J. Shiller, Online Data (monthly S&P 500, dividends, earnings, CPI from 1871)
• Aswath Damodaran, NYU Stern School of Business (annual returns and risk premia)
• U.S. Department of the Treasury, Yield Curve Data
• Freddie Mac, Primary Mortgage Market Survey

Mortality and demographics

• Centers for Disease Control, National Center for Health Statistics, Mortality Tables
• Society of Actuaries, Experience Studies
• U.S. Census Bureau, demographic data

Energy and other macro

• U.S. Energy Information Administration
• National Center for Education Statistics (tuition data)
• The College Board, Trends in College Pricing

Methodology references

• Bureau of Labor Statistics, "Updated Response to the Recommendations of the Advisory Commission to Study the Consumer Price Index" (the Boskin Commission response, 1998 and updates)
• John Williams, ShadowStats methodology notes (alternative inflation measures)
• Consumer Expenditure Survey, Bureau of Labor Statistics
• Health and Retirement Study, University of Michigan