The G-Street Value Methodology

The one-liner

Every G-Street Value on this site is computed by the same fixed formula — no vibes, no per-stock exceptions. That's what makes the "Est. GAINZ" numbers comparable across the board:

Analyst leg = weighted analyst target × 0.92 (sell-side optimism discount)
Factor leg = our factor value, capped at the analyst leg (discipline, never hype)
G-Street Value = 70% × analyst leg + 30% × factor leg

The two governors are the point: whenever analyst data exists, our value can never exceed 92% of what the weighted street thinks, and our own fundamentals model is only allowed to pull the number down. We would rather understate the upside than sell you a fantasy.

Our AI's job is to gather the inputs (fresh analyst price targets, live fundamentals). The number itself comes out of deterministic code. Same formula, same bands, every stock, every night.

The Weighted Analyst Target

We collect recent price targets from major research shops — but we don't weight them equally, because analysts aren't equally good.

• Firm credibility — rated from day one, refined by data. Every research shop carries a GainStreet rating (tier A ×1.3, B ×1.0, C ×0.6) — our editorial judgment on the quality of their work. On top of that, we backtested every public rating action these firms made on our coverage universe since 2022 (12,800+ graded calls across ~116 firms), scoring each bullish or bearish call against the stock's next six months versus the S&P 500 — a firm at the universe-average hit rate weighs ×1.0, better firms weigh more, worse firms weigh less. Going forward, every price target we capture is graded against the realized price six months later, so the scorecard keeps updating monthly and, with enough data, a firm's record overrides its rating entirely. Hover any weight badge on a company card to see the firm's hit rate and graded-call count.
• Freshness. Targets under 3 months old count at full weight, 3–6 months at half weight, and anything older is dropped entirely.
• Outlier control. Targets are winsorized to ±40% of the median target, so one moonshot call can't drag the blend.

Each analyst card on a company page shows the exact weight that firm carried (the ×N badge) — hover it for the firm's graded track record.

The GainStreet Factor Value

Sell-side targets have a documented optimism bias, so 30% of the blend is our own fundamentals check: a fair multiple applied to forward earnings.

• Industry-anchored multiples. The fair multiple starts from the company's industry base (semiconductors ~22×, software ~28×, utilities ~17×, and so on) — memory stocks are judged against memory multiples, SaaS against SaaS.
• Growth and margin adjustments. Revenue growth moves the multiple up or down around the anchor within a hard clamp (0.6×–1.8× of base); margin quality nudges it ±10%.
• Rule of 40 — software only. The growth + FCF-margin screen applies exclusively to recurring-revenue software, where it was invented. It never touches cyclicals or hardware.
• The cyclical guard. Memory and storage names (MU, SKHY, SNDK, and peers) carry a hard fair-P/E cap of ~14× no matter how good peak-cycle growth and margins look. We never capitalize peak margins.
• Pre-profit companies are valued on a growth-banded price-to-sales multiple instead of earnings.

Conservatism governors

• The sell-side optimism discount. Analysts are structurally paid to be optimists, so the analyst leg is discounted 8% before blending. G-Street Value can never exceed 92% of the weighted street view.
• Factors discipline the street — they never out-bull it. When our factor value comes out above the analyst leg, it's capped down to it. The factor model exists to pull inflated consensus back toward fundamentals, not to push values higher; when it's capped the card says so.
• Proven-bad firms are excluded. A firm whose earned credibility weight falls below ×0.5 (a genuinely poor graded record — small samples are protected by shrinkage) is dropped from the calculation entirely. Its target still renders on the card, struck through and marked excluded, so you can see exactly whose opinion we're ignoring and why.
• The street sanity clamp. We sample 3–5 targets a night; the full covering street is often 40–60 analysts. If our sample strays more than ±20% from the street's mean target, the analyst leg is clamped back into that range and the card says "street-clamped." One misread or unrepresentative target can't hijack the number.
• Earnings reset freshness. A target published before the company's latest earnings report is halved in weight regardless of calendar age — its information set is obsolete the moment new numbers drop.
• Unpriceable profiles don't vote. When our factor model's output falls below its own sanity floor (typically pre-profit, hyper-growth shapes), that's a safety rail — not an opinion. The value goes analyst-only and the card says "factors n/a for this profile" instead of blending in an artifact.

Honesty rules

• Coverage-adaptive weights. With 3+ fresh firm targets we blend 70/30. With thin coverage (1–2 targets, or street consensus only) we drop to 50/50. With no analyst data at all, the value is 100% factor-based and labeled as such.
• Disagreement is disclosed, not hidden. When the analyst view and our factor value diverge by more than 50%, the company card shows a ⚠ flag and our stated confidence drops a notch.
• No grounded inputs → no number. If we can't compute either side of the blend from real data, we show nothing rather than invent a target.

The G-Street Score

Beside every value sits a second, independent number: the G-Street Score (0–100). The Value answers "what is this worth?"; the Score answers "how strong is the setup right now?" A stock can be cheap with a terrible setup, or expensive with a great one — you want to know both.

The Score is a fixed formula: six categories and sixteen checks, computed by code from filing-grade data — 10-K and 10-Q numbers wherever they exist, never a random webpage. The same formula applies to every stock and is refreshed nightly. The six categories are:

• Revenue Growth — top-line growth, its acceleration, and whether guidance points to a faster next quarter.
• Profit Margins — the level of margins and whether they are expanding or compressing.
• Revenue Visibility — booked backlog, recurring revenue, and how concentrated the customer base is.
• Cash Flow & Balance Sheet — real cash generation, real profits, and the strength of the balance sheet.
• Governance & Dilution — capital allocation, management alignment, and whether the share count keeps rising.
• Valuation & Market Signals — analyst estimate revisions, what you pay for the earnings, and price momentum.

The exact weights, bands, and thresholds are not published. What we will promise: it is deterministic code with no judgment-based inputs, every stock gets the identical formula, data it cannot find scores neutral rather than invented, and it never overrides the lifecycle — it informs judgment, it is not a trading signal.

Bands: 90+ Max Conviction · 75–89 Core · 60–74 Probation · under 60 Weak. The bar is strict — most stocks score Weak, and that's the point.

The obvious disclaimer

GainStreet is an AI-run research experiment, not a licensed financial advisor. G-Street Values are model outputs, not price predictions, and nothing on this site is financial advice. May the Gainz be with you — but do your own homework.