The math hedge funds use, opened up — public methodology, live calibration, retail-priced.
Framler turns an academic factor stack across 1001 tickers into one Bayesianforward-return engine — a single coherent score and 90% interval per name, every factor cited to peer-reviewed research.
Built over the past two months, served free during the calibration window; paid tiers open Q3 2026 once the 30-day horizon matures. A quantitative research platform for independent investors and small funds — public methodology, no AI guesswork in scoring.
● Product maturity — today, on the wire
Equities scored daily
1,001
US + international — full multi-factor recompute every day
Depth ≠ proof. Those snapshots have produced only ~1–2 independent non-overlapping windows so far — which is exactly why we do not yet claim a proven edge. The honest measured-vs-backtest split is below.
The honest split
Track record — measured forward returns are live
Today — synthetic backtest
IC Sharpe ≈ 1.25 on a 5-year Fama-French composite-factor backtest (n=1,826 rolling windows, sector-neutral measurement). Walk-forward CV IC Sharpe ≈ 1.0 on the test fold. Backtest only — no live trading capital deployed. Discount accordingly.
Live — measured forward returns
First measured weights persisted 2026-05-17 on 1d and 7d horizons (3 regimes, in-sample IC 0.055-0.102, n=1091). 30-day horizon now maturing; 90-day ~2026-09-15. /track-record publishes measured per-horizon IC and early hit-rate diagnostics now. Portfolio-level metrics (Sharpe, Sortino, Calmar, max drawdown) remain gated until enough completed monthly rebalance periods exist — see /backtest.
Why this split matters to a fund: live measured numbers have an entirely different evidentiary weight from backtest. We refuse to blur the line because every retail-quant tool that has done so eventually got caught.
One more caveat a fund will want, stated up front: the scored universe is survivor-only (≈1,000 currently-listed names; delisted, bankrupt and merged tickers are not in the snapshot). The forward-looking consequence: a Sell/short signal here flags mediocre-to-poor survivors, not the genuinely broken names that drive a long/short hedge in a drawdown — so a strategy built only on these signals will tend to under-hedge in a real downturn, and an allocator should pair it with its own distressed/short sleeve. We state this qualitatively on purpose: the magnitude isn't measured in-product, and we won't invent one.
Live, verifiable
Live engine state
Verdict mix today: 40% bullish · 7% bearish · 585 confluence patterns firing · 43 Phase-3 trials tracked · last refresh 03 Jul, 09:45. Verify live at /status.
● Live measurement active
Forward-return calibration is live — 1d + 7d horizons published
Forward-return weights have been measured live since 2026-05-17 (1d + 7d horizons; 30-day now maturing). Portfolio-level Sharpe / Sortino / Calmar remain gated until ≥3 completed monthly rebalance periods — see /backtest for the gating logic and /track-record for live per-horizon IC.
Why this is defensible
The opportunity & the moat
Multi-factor quant research lives behind institutional terminals — a Bloomberg or FactSet seat runs $25,000–30,000 a year, a Compustat / Refinitiv feed starts at six figures, and an in-house scoring engine takes 3–5 quant engineers a couple of years to ship. Retail investors and small funds get none of it — they fall back on analyst ratings (which underperform multi-factor scoring on every dimension we measure — the analysis), narrative, and P/E rules of thumb. Framler is the same factor literature, same engineering rigour, at roughly 1/100th of the cost — methodology published, calibration measured live.
The moat is the engine, not opacity. Most retail screeners stop at news-sentiment classifiers; most institutional stacks just add factor scores together or run them in isolation. engine v2 composes them coherently — a multi-factor signal, regime conditioning, confluence patterns and tail-dependence all flow into one Bayesian forward-return distribution per ticker per horizon (1d / 7d / 30d / 90d), with a 90% interval calibrated to actually hold up out-of-sample. Every coefficient cites peer-reviewed literature; the full derivation, parameter values and composition rules stay proprietary — methodology sketches the inputs without the formulas. That gap is the defensibility.
On market size we publish no TAM / SAM / SOM. An honest number needs measured conversion, retention and willingness-to-pay — none of which exist until the first paid tier opens (after the 30-day horizon finishes maturing). Top-down extrapolation from incumbent revenue sounds confident and doesn't survive the first cohort, so we'd rather show nothing than fiction. Until then the signal for investors is product maturity — calibration state, measured IC, free-tier retention — at track-record and backtest, not a market model.
Pricing stays honest with that: free during early access to drive citation and awareness, $29/mo Pro post-launch (alerts, historical depth, REST API), and enterprise contracts from ~$2k/mo for funds that want bulk endpoints and a custom universe. The comparison below is the head-to-head.
Competitive moat
Framler vs the incumbents
CapabilityFramlerBloombergRefinitivS&P CapIQ
Price per seat / yr$0–$348$25–30K$22K$30K+
Multi-factor composite scoreFramler 0–100No (raw data only)No (raw data only)No
Bayesian regime detectionBOCPD liveNoNoNo
Conformal prediction intervalsMondrianNoNoNo
Phase 3 trial failure predictionBayesian, every nameNoNoNo
Open methodology pagePublicProprietaryProprietaryProprietary
Factor citations to peer-reviewed papersEvery factorNoNoNo
The incumbents win on universe breadth, real-time data feeds, and qualitative content (transcripts, news, analytics packages). Framler wins on transparent factor methodology, Bayesian regime conditioning, conformal uncertainty quantification, Phase 3 prediction, and a price point that lets retail and small-fund users access the same factor literature institutional desks pay six figures for. Different product, overlapping shelf — the way Linux disrupted Solaris, not by being equivalent but by being structurally better suited to a different buyer.
Built, shipped, free
Traction so far
Engine v2 deployed
Engine v1 → v2 — an expanded academic factor stack (added Amihud 2002 microstructure, 8-K earnings press NLP, cross-asset macro). BOCPD + conformal + copula + Shapley + confluence + pharma layers live; Kalman dynamic exposures + Fama-French regression implemented and in calibration (see /status), with extensive mathematical invariants running on every request.
Public content layer
Hub pages (methodology / moat / patterns / track record / coherence / compare / about / blog) + long-form research posts + a per-pattern detail surface.
SEO surface
Sitemap covers ~2,300 URLs (per-ticker pages + research hubs). Schema.org FinancialProduct on every ticker. AI crawler allowlist explicit (ClaudeBot / GPTBot / PerplexityBot all citing).
Pipeline reliability
Automated crons keep scoring refreshed and the supporting data (prices, alerts, sector stats) on their own schedules; the public /status page shows live per-layer freshness — so this page never hard-codes a cadence that can drift.
First adaptive weights persisted 2026-05-17
Calibration window opened 2026-05-16; first measured factor weights landed the next day across 1d + 7d horizons × 3 regimes. /track-record publishes measured per-horizon IC and early hit-rate diagnostics; portfolio-level metrics (Sharpe, Sortino, Calmar, max drawdown) remain gated until enough completed monthly rebalance periods exist.
Discipline
What we will not do
Custody assets. We are not a broker. No money flow, no fiduciary duty mismatch, no regulatory perimeter creep. Research tooling, period.
Sell a black box. Every score traces to a peer-reviewed paper. The methodology page is intentionally exhaustive. The moat is engineering, not opacity.
Inflate metrics before measurement. Track record page showed zero performance numbers until 2026-05-17, the day first measured forward returns calibrated. We'd rather under-promise and deliver than do the opposite.
Take dirty money. No payment-for-flow research arrangements. No paid tickers. No promoted picks. The engine is index-agnostic — it scores what the math says, including names that hate us.
Pre-seed allocation
Use of funds
Pre-seed thesis: $500K-$2M raise extends runway 18-24 months and funds three workstreams that structurally compound (engine breadth, data depth, paid-tier conversion). The mix below assumes a $1M raise; allocations scale proportionally.
50%
Engineering — factor R&D + infrastructure
Founder runway + 1 quant engineer hire post-seed. New factors (PEAD #14, options skew, microstructure depth, alternative data integrations). Migration to dedicated infra once Vercel free-tier ceiling binds. Calibration window automation.
20%
Data sources + paid feeds
Compustat fundamentals (depth on financials), Refinitiv I/B/E/S analyst estimates, intraday OHLCV, options volume real-time. Each data feed unlocks 1-2 new factors and tightens existing factor signal-to-noise.
12%
Marketing + community + content
Long-form research (academic-style posts), founder presence on quant Twitter / podcasts, conference attendance (NeurIPS / quant finance), backlink outreach for SEO compound, paid acquisition only after track-record validates.
10%
Legal + compliance + audits
Counsel-reviewed Terms + Privacy + risk disclosures, SEC research-firm posture review, SOC2 Type I prep, optional external penetration test, data-protection-officer engagement for EU.
5%
Founder runway + operations
Cost of being a full-time founder for 18-24 months. Health insurance, accounting, tooling subscriptions, office overhead.
3%
Reserve
Buffer for unexpected — server burst (track record viral), regulatory clarification, opportunistic data deal.
Detailed budget by line item available under NDA. Burn assumptions stress-tested at 2× and 0.5× — runway 18-24 months at base case, 9-12 at 2× burn (with paid-tier launching), 36+ at 0.5× burn (founder-only mode).
Next 12 months
2026 roadmap
Q2 — DELIVERED: First out-of-sample calibration landed 2026-05-17 on 1d + 7d horizons; 30-day now maturing, 90-day ~Sep. Track record publishes live per-horizon IC now. Portfolio-level Sharpe / Sortino / Calmar remain gated on /backtest until ≥3 completed monthly rebalance periods.
Q3: Pro tier launches ($29/mo). Alerts, historical depth, REST API (1000 req/day), monthly research letter. Founder pricing for early Free users.
Q4: Universe expansion — broader international coverage: fuller UK FTSE 350, EU STOXX 600, Japan TOPIX. Same math, broader market.
Q4: Quant API for funds — model-portfolio endpoints, factor exposure decomposition, regime-conditional rebalancing recommendations.
Talk to us
Investor enquiries
We are not currently raising publicly. Investor conversations happen privately, on the basis of the moat deck (separate from this public brief), and only with funds whose thesis aligns with the public-access- first model. Email founders@framler.com if you'd like the moat deck and a 30-minute conversation.
For research collaborations, academic verification access, or press: research@framler.com.