Distribution Safety Score™

Definition

The Dividend Vision Distribution Safety Score™ is Dividend Vision's own 0–100 estimate of how safe a fund's distribution looks going forward. Higher is safer. It appears on every covered fund's ticker page as a gauge with a color band, and the same number shows up as a sortable column in the screener, watchlist, holdings table, compare pages, and the portfolio risk view.

It is deliberately not an AI model or a black box. It is a transparent, rules-based penalty engine: every fund starts at 100, points are deducted for specific, named red flags — a falling or recently-cut payout, price/NAV erosion, deep drawdowns, leverage, crypto exposure, a short track record, and so on — and the result is clamped to 0–100. Because every deduction is a labeled factor, the ticker-page card can always show a plain-language "What's pulling it down" breakdown next to the number, and the rows on that card add up to the points the score lost. No hidden weights shift around behind the scenes, and the same fund scores identically on every surface because every surface runs the exact same engine on the exact same daily data.

One thing the score is emphatically not: a yield penalty. A high distribution rate on its own is not treated as a red flag. Many income funds yield high by design, and the score judges them on their payout trend, price behavior, and track record — not their headline rate. Only genuinely extreme yields, far above what a fund's structure normally supports, draw a deduction.

Distribution Safety Score™ is a trademark and proprietary metric of Dividend Vision LLC — the score, its factors, and its weightings are developed and maintained in-house.

Why It Matters

Income investors face an asymmetric problem: the headline yield is the easiest number to see and the least informative one. A 12% yield tells you nothing about whether the payout is growing or being quietly cut, whether the share price is eroding underneath it, or whether the fund has existed long enough to prove anything at all. Chasing the biggest number is exactly how investors walk into a yield trap.

The Distribution Safety Score compresses the checks a careful analyst would actually run — payout trend, past cuts, price erosion, drawdown, beta, leverage, coverage, track record, cost — into one comparable number with a visible breakdown. That does three useful things:

• It surfaces problems the yield hides. A fund can pay a flawless-looking 11% while its distribution has been trimmed twice and its NAV has ground down 20% — the score flags all three even though the headline yield still looks generous. See why high yield isn't high income.
• It sets expectations for young funds. New funds simply don't have the history to prove a durable payout, so instead of projecting confidence the score caps them and labels them Unproven — a "no verdict yet," not a "risky."
• It makes portfolios comparable. Because every fund is scored by the same engine, you can sort a screener by safety, spot the lowest-scored holding in a portfolio, or compare two candidates side by side without re-reading ten fund pages.

What it is not: the score is an educational estimate, not investment advice, a prediction, or a credit rating. A high score isn't a promise the distribution holds, and a low one isn't a prediction of a cut.

How It's Calculated

The mechanics are simple to state: start at 100, deduct points for each red flag present, clamp the result to 0–100, and round. Every deduction has a fixed maximum, so no single signal can dominate beyond its cap. The factors, what triggers them, and their maximum deductions:

A few design choices in that table deserve emphasis:

High yield per se is not penalized. The "extreme yield" caps are structure-specific because context matters: 9% is unusual for a plain equity ETF but routine for a BDC or CEF, and a covered-call fund can sustainably distribute far more than either. Only yields beyond ~40% for option-income funds, ~12% for BDCs/CEFs/MLPs/REITs, or ~8% for everything else draw the deduction — the territory where chronic NAV decay becomes the norm. Likewise, being an option-income fund is not treated as a defect: the strategy itself carries only a token deduction, because its real risks (an eroding NAV, a cut payout) are already measured directly by the erosion and payout-trend factors.

Price-behavior factors share a combined cap. Erosion, the covered price decline, the long-term decline, drawdown and beta all describe how the fund's price moved, so their combined deduction is capped (at 35 points, equal to the payout-trend maximum) to stop one bad market move from being charged five ways at once. When the five together exceed 35, each is scaled down in proportion, so the breakdown keeps its ordering. The lowest scores are reserved for funds failing on multiple independent dimensions. Crypto exposure deliberately sits outside that cap: it's a structural forward risk that holds even after a calm year, and backtesting showed crypto-linked funds cut far more often than same-scored plain funds.

The trailing-year payout factors share a cap of the same size. Falling payout, Missed distribution and Skipped distributions all read the past twelve months of the distribution stream, and they overlap: a payment that did not arrive is also a smaller trailing-12-month total, so a quarterly payer with one missed payment used to be charged for the gap and for the 25% drop in its yearly payout that the gap caused. Their combined deduction is capped at 35 points, scaled in proportion like the price factors, so the trailing year's stream and the price around it can each cost at most 35. "Missed distribution" and "Skipped distributions" are also mutually exclusive on their own: a fund is charged one or the other. "Past dividend cut" sits outside the cap because its window is one to five years back, which a gap today cannot touch, so a fund that cut deeply then and has stopped paying now is still charged for both. "Irregular distributions" and "Infrequent payout" sit outside too, because they fire only where the missed/skipped factors cannot and read the absence of a commitment rather than the stream's behavior.

Three price-decline factors, two price series. The 12-month Price / NAV erosion factor is measured on the distribution-adjusted price — the total-return view — so a fund whose share price slips 1–2% over a year while paying out 7% takes no erosion deduction: the payouts more than covered the dip, and the holder came out ahead. But covered is not harmless. Backtesting across 187,000 fund-month observations showed that funds whose raw price fell even as payouts covered it went on to cut at steadily higher rates as the dip deepened — a Safe-band fund with a covered dip of 3–10% cut within a year at the Generally-safe base rate, and one beyond 10% cut 63% of the time, worse than the High-risk band (the observation count behind that deepest rung was not recorded with the run, so read it as a direction, not a precise rate) — so the Covered price decline factor charges the raw slide directly, one point per 1% of decline. A fund with a slightly red 1-year price return therefore reads a few points below a rising twin, not identically. The long-term price decline factor extends the same raw-price reading over 3–5 years. An adjusted series adds every payout back in, which is the right way to measure what an investor earned — but it is exactly the wrong way to spot a fund handing back principal, because the payout and the price drop cancel each other out. A fund yielding 48% whose share price slipped over the year can read +61% on an adjusted basis. Measured on the raw price, the glide shows up as the decline it is. That is why this factor watches the multi-year window and only charges once the raw price is compounding down faster than ~3% a year: a mild drift over years is tolerable (a flat-NAV bond fund legitimately gives up a percent or two), but a price compounding steadily downward while a large distribution is paid is close to arithmetic proof that some of the "income" is your own capital coming back.

The track-record gate counts payouts, not birthdays. What proves a distribution durable is the number of payouts observed: roughly eight distributions across at least six months of history. A monthly payer clears that in about eight months; a quarterly payer needs about two years; a weekly payer even less. A fund that hasn't cleared the gate takes a "Too new to score" deduction and has its score capped at 50, with the band relabeled Unproven — shown in a neutral slate color rather than red, because it means "no verdict yet," not "dangerous." A fund that has cleared the gate but is still young takes only the mild "Short track record" nudge from the table (3 points under two years of history, fading to nothing by year five). Proven-but-young option-income funds face an extra confidence ceiling: they can't read Safe (80+) until about two years of record, and can't score above 90 until about five — not a strategy penalty, but a requirement that a volatility-premium payout prove itself across more than one market regime. Single-underlying option funds are held to a stricter version of that ceiling — topping out in Caution (49) under two years, 59 under three, and 74 under five, with the ceiling lifting fully at five years. Backtesting showed why: young single-stock option funds whose trailing signals looked clean went on to cut within a year at rates worse than the Caution band, and which underlying craters is not predictable from payout history — funds with identical structures and records split between 0% and 50%+ cut rates purely on what their one stock did next. A payout tied to a single stock's volatility premium simply needs a longer record before any confidence badge; a proven single-stock fund with five years of history can still read Safe.

What history does the annual payout comparison need? For regular quarterly through twice-weekly payers, it normally pairs two full payout windows at the current frequency: 4 against 4 quarterly payments, 12 against 12 monthly payments, or 52 against 52 weekly payments, roughly two years in total. Normal payment-date shifts do not count as income changes. A cadence change, missing payments or a long interruption can make the windows incomparable; the annual reading then stays unavailable until comparable history exists. Annual and semiannual payers retain the existing comparison of at least three payments per side, which spans multiple years. An unavailable annual reading is not a zero change. The existing guarded shorter-history payout trend may still be usable by DSS; this does not activate a separate provisional scoring method.

A payout that stops arriving is scored as hard as a payout that shrinks. The payout-trend factor compares payout sizes only when comparable windows exist; missing payments can leave that comparison unavailable. Three independent cadence factors evaluate those gaps. "Missed distribution" fires when a quarterly-or-better payer has no unbroken on-cadence run at all right now — it is overdue, or a skip broke the run and only one payment has landed since — and it carries one of the heaviest charges in the engine (up to 25 points), because a payment that never came is the least ambiguous distress signal there is. "Skipped distributions" is the milder version: a gap inside the past year that the fund has since recovered from. Both are charged only to frequent payers with a stated cadence, since for an annual payer a "missed" payment is indistinguishable from date wobble — the infrequent-payout factor already owns that risk. Both also measure the current run against how long the fund has been on its present schedule, not against its whole payout history, so a fund that changes cadence is not charged for the change: when a quarterly payer moves to monthly, its old quarterly spacing is no longer a hole in a monthly calendar, and the score waits for the new schedule to build its own record instead. "Irregular distributions" covers the opposite case: a payer with no stated schedule whose history averages under four payouts a year is paying discretionary specials, not running an income stream, and the charge escalates sharply when the posted forward rate leans on those one-offs rather than the realized stream. A fund is never charged both "Infrequent payout" and "Irregular distributions" — the first requires a stated cadence, the second its absence.

Missing data shrinks the score toward its structure's typical score, and caps it. A missing input is not itself a deduction, but a record scored on fewer inputs has fewer chances to lose points, so treating the gap as clean reads optimistic. Instead, for each unavailable core signal (payout trend, 12-month price change, track record, payout schedule), a quarter of the score is replaced by the median score of fully-evaluated funds of the same structure: about 87 for an ordinary fund, 90 for a dividend stock, 80 for a BDC, CEF, MLP or REIT, 73 for an option-income fund and 4 for a crypto-linked fund. One gap is a 75/25 blend, two is 50/50, and a record with nothing measurable simply reads as its structure's typical fund. The blend only ever lowers a score: the deductions already taken are measured evidence about this fund, and filling an unseen dimension with an average must not erase them. The Limited data chip spells out the blend. On top of that, a record missing two or more of the four core signals is flagged Low confidence and capped at 79, so a sparse record can never read as Safe. The payout schedule counts as known when the fund states a cadence or its tracked history shows a frequent (roughly quarterly-or-better) rhythm — with no cadence at all, the score cannot even tell whether a payment is late, so a whole risk dimension goes unevaluated.

For CEFs and option-income funds, return-of-capital availability is also disclosed as known zero, known nonzero, estimated, or unknown. ROC is retained as tax-character information, not a scored DSS input. Missing or preliminary ROC does not lower DSS confidence or add a warning to the score card; unknown ROC remains distinct from verified zero.

A rock-bottom share price is a stock-only distress flag. For an individual company, a very low nominal share price is an independent fragility signal: sub-$1 US listings face the exchange $1-minimum-bid delisting rule, and low-priced shares broadly carry wider spreads, thinner liquidity, and more frequent payout cuts than company size alone captures. The charge is deliberately gated on a small market cap (under $2B), so it fires only where cheap and small overlap — a large company quoted at a low nominal price is following a share-count or foreign-listing convention, not signaling distress, so Ambev, Banco Bradesco and Wipro (all low-priced but tens of billions in cap) are never flagged. For a fund, share price means nothing — it is set arbitrarily at inception, so a $5 ETF is no riskier than a $50 one — and the factor never applies there.

The final number maps to bands: 80+ Safe, 60–79 Generally safe, 40–59 Caution, 20–39 Elevated risk, below 20 High risk — plus the special Unproven band for funds that haven't cleared the track-record gate.

What isn't scored at all: money-market funds (a pinned $1 NAV has no safety question), funds with no recurring distribution (a lone special dividend isn't an income stream), and inverse/bear daily trading products, whose distributions are incidental collateral interest. Short-strategy income funds that sell option premium as the product are still scored.

When it updates: the score is recomputed with the daily data build, and one engine produces the number for every surface — the ticker card, screener, compare pages, watchlist, holdings, portfolio risk view, and Ask DV chat all read the same daily inputs, so a fund never shows two different scores. It is not a live intraday metric.

Example

Here is how the ticker-page breakdown reads for an illustrative fund — call it a seven-month-old, single-stock option-income ETF paying monthly, yielding 45%, with seven distributions banked so far and a payout trend already drifting down about 20% a year. Every figure is illustrative, but the arithmetic is the real engine's:

Distribution Safety Score: 50 / 100 — Unproven

What's pulling it down:
  Falling payout            -14.0   distribution trending down ~20% annualized
  Too new to score          -10.0   only 7 distributions so far
  Leverage & concentration   -4.0   tied to a single underlying asset
  Very high yield            -2.0   45% yield vs. the ~40% option-income cap
  Manufactured income        -2.0   income from selling options

  100 - 32 penalty = 68 (one core input missing; the option-income typical score
  of 73 would raise it, so no blend) → capped at the Unproven ceiling of 50
  → band: Unproven

The raw arithmetic lands at 68, but seven distributions is short of the roughly-eight gate, so the score is capped at 50 and the band reads Unproven — the cap, not the deductions, is the binding constraint. Notice what did not hurt it: the 45% yield cost only 2 points (it barely exceeds the option-income cap — the yield level is mostly tolerated), and being a covered-call fund cost only the 2-point token charge. Notice also what couldn't be measured yet: at seven months old the fund has no trailing-12-month price window, so the price-erosion and drawdown factors show nothing — the gap would pull the score toward the typical option-income fund's 73, but because that would raise it the blend is skipped, and the card flags it with a Limited data chip. The real damage came from the payout already falling and the fund being too young to have proven anything — the things that distinguish a real distribution problem from a scary-looking headline rate. One more payout on the same trajectory clears the gate, lifts the cap, and lets the deductions (plus the young option-income ceiling) speak for themselves.

At the other end of the spectrum, a large, decade-old dividend fund with a growing payout, a stable price, modest yield, and a low expense ratio — the SCHD profile — accumulates almost no deductions and reads in the high 90s. A proven option-income fund like SPYI or QQQI is judged the same way — payout trend, NAV behavior, track record — with the young-fund ceiling lifting as its history lengthens.

Does It Work?

The table below is the July 11, 2026 historical calibration, not a fresh validation of today's method or a probability forecast. The backtest reconstructs each fund's inputs as they stood on monthly evaluation dates from mid-2022 through mid-2025, scores them with the same rules-based engine as it stood at the 2026 calibration run, and then looks at what actually happened over the following 12 months — across 774 funds and more than 21,000 fund-month observations. A "cut" is measured one way for each structure, variable payers included: the fund's trailing-12-month payout total on the evaluation date is compared with the same total 12 months later, so a monthly payer's normal month-to-month wobble does not register, while a fund whose yearly income fell by 15% (or 30%) does. The share of fund-month observations that went on to cut by that definition, by band:

The universe for that run was the catalog as it stood in July 2026, so funds that had already closed or been liquidated before then are under-represented, which flatters each band a little; the ongoing validation described below retains closed funds for exactly that reason. The thresholds were also tightened against the same 2022–2025 data that produced the table, so it is an in-sample calibration, not a test of the method on data it had not seen. Confidence intervals, a held-out period and a comparison against simple screens (yield rank, a recent-cut flag, a NAV-trend rule) are part of the ongoing validation process and are not yet published.

Cut rates rise strictly band by band: a High-risk fund was roughly 7x more likely to cut by 15%+ within a year than a Safe one, and roughly 14x more likely to cut deeply. Two of the engine's thresholds — the crypto factor stacking outside the price-behavior cap, and the young option-income ceilings — were subsequently tightened directly against this evidence, so the engine that ships today is slightly stricter than the one that produced the table above. Later documented calibration runs on August 26 and 27, 2026 added covered price decline and stricter ceilings specifically for young single-stock option-income funds (method 1.3). Current published scores use method 1.6. The version on each score identifies the rule set that produced it; the historical table above is not a backtest of method 1.6. Do not compare different method versions as if only the fund changed.

Ongoing validation

Daily score updates and historical validation are separate processes. The validation workflow is scheduled monthly, with a quarterly review of calibration evidence. It refreshes historical data, includes retained closed funds and current non-payers, requires distribution-gap reconstruction, and reports missing inputs and incomplete outcomes. Closed funds without observable liquidation outcomes are reported separately; they are not counted as successful non-cuts.

Validation separates earlier development dates from later evaluation dates, with a gap to prevent overlapping outcome windows. Confidence intervals resample whole funds because monthly observations from the same fund are related. Candidate changes require fresh, untouched validation evidence and review; scheduled runs keep the scoring weights unchanged. Some historical inputs remain approximations, so even a completed run does not establish an individual fund's cut probability. The first report from this expanded validation process is pending; no new performance claim is made here.

That's the honest case for the score. The honest case for reading it carefully:

• These are cohort base rates, not per-fund predictions. "Safe" means the band historically cut at a 6.7% rate — it does not mean this fund has a 6.7% chance, and it certainly doesn't mean zero. Some Safe funds cut; some High-risk funds never do.
• The inputs are trailing, so the score lags regime changes. There are no forward-looking inputs (no options-overwrite math, credit quality, or rate sensitivity). After a long calm market, scores drift up — and in the backtest, Safe-band funds evaluated in 2022 went on to suffer 12-month price crashes at several times the rate of those evaluated in 2023–2025, because a bear market arrived through perfectly calm trailing signals.
• The weights are hand-tuned heuristics. The band-level results validate the overall calibration, and two thresholds were empirically recalibrated, but most individual multipliers and cutoffs are sensible rules, not fitted parameters. It is a rules-based estimate — never a "model" output or a probability.
• Missing data is filled with a structure average, not with evidence. A fund with a sparse record reads partly as its structure's typical fund, which is the best available guess and still not a measurement of this fund. The confidence chip and the Low-confidence cap exist precisely to flag that case.
• Coverage is checked for BDCs only. Distribution coverage is the best forward-looking cut signal for BDCs, but an earnings-based ratio is misleading for CEFs and plain ETFs (whose payouts routinely include return of capital and capital gains), so the score may catch a quietly under-earning ETF only later, through NAV erosion or the cut itself.

Common Mistakes

• Treating the score as a guarantee. An 85 is a favorable trailing profile, not a promise — 6.7% of Safe-band observations in the backtest still cut within a year. Use it to prioritize attention, not to replace it.
• Comparing scores across structures without context. A 75 on a two-year-old option-income fund (sitting under its confidence ceiling) and a 75 on a decade-old plain dividend ETF got there differently. Read the factor breakdown, not just the number.
• Reading "Unproven" as "risky." The Unproven cap is a statement about evidence, not danger — the fund hasn't banked enough payouts to judge. Some Unproven funds will mature into 90s; the score just refuses to guess which.
• Using it as a standalone sell signal. A score that drops is a prompt to look at why — one factor (a mid-year drawdown that later recovers, say) can move the number without changing the fund's payout economics. Check the breakdown, the payout history, and the NAV trend before acting.
• Assuming a high yield lowered the score. It almost never does — yield only draws a deduction above the structure-specific extreme-yield caps. If a high-yielder scores low, the breakdown will show the real reasons: a falling payout, erosion, or a thin record.

FAQ

What is a safe distribution?

A safe distribution is one the portfolio can keep paying from what it earns, not from shrinking its own asset base. Dividend Vision's Distribution Safety Score™ turns that idea into a 0–100 reading from payout history, price behavior, structure, and track record. Higher scores mean fewer visible red flags — not a guarantee. A high headline yield is not the same thing as a safe distribution.

Is the Distribution Safety Score a dividend safety score?

It answers the same question for a wider set of holdings. Dividend safety scores built for individual stocks lean on payout ratios, earnings, free cash flow, and balance-sheet debt. An ETF, CEF, or option-income fund has no earnings payout ratio, and much of its cash can come from option premium, interest, or capital gains rather than dividends. The Distribution Safety Score works instead from evidence that funds and stocks both have: the distribution history itself, price and NAV behavior, drawdowns, leverage, structure, and track record. It also scores dividend stocks, with stock-specific checks such as the very-low-share-price deduction, so a dividend ETF and a dividend stock read on the same 0–100 scale and bands.

What is a good Distribution Safety Score?

80 and above reads Safe, and 60–79 Generally safe — in backtesting those bands went on to cut by 15%+ within a year at roughly 7% and 15% rates respectively, versus ~49% for the High-risk band. But "good" is contextual: a proven option-income fund under 3 years old tops out at 84 by design (a single-stock option fund at 59), so a 79 — or for a young single-stock fund a 55 — can be a strong result for its class. Always read the number next to its band and factor breakdown.

Does a high yield lower the score?

On its own, no — that is a deliberate design choice. Yield only draws a deduction above a structure-specific extreme threshold: roughly 8% for plain funds, 12% for BDCs, CEFs, MLPs and REITs, and 40% for option-income funds. Below those caps, a fund is judged on its payout trend, price behavior, track record, and cost — not its headline rate.

Why is my new fund capped at 50?

It hasn't banked enough payouts to prove its distribution durable — the gate is roughly eight distributions over at least six months, so a monthly payer clears it in under a year while a quarterly payer needs about two. Until then the score is capped at 50 and labeled Unproven, which means "not enough evidence yet," not "unsafe." The cap lifts automatically as real payouts accumulate.

How often does the score update?

It is recomputed with the daily data build, and every surface — ticker page, screener, compare, watchlist, holdings, portfolio analysis, Ask DV — reads the same engine and the same inputs, so a fund shows one consistent score everywhere. It is not a live intraday number, and it does not react to today's price move until the next build.

Can a fund with a Safe score still cut its distribution?

Yes. In the 21,000-observation backtest, 6.7% of Safe-band fund-months were followed by a 15%+ cut within a year. The bands are historical base rates, not per-fund guarantees — and because the inputs are trailing, a sharp market regime change can hit before the score reflects it. Treat Safe as "few red flags visible," never "nothing can go wrong."

Where do I see the Distribution Safety Score?

The flagship view is the gauge card on each fund's ticker page (for example QQQI), which includes the full "What's pulling it down" factor breakdown. It's also a sortable column and filter in the screener, a row on compare pages, a column in your watchlist and holdings tables, and a band-by-band breakdown in the portfolio risk view. Money-market funds, non-payers, and inverse trading products show no score by design.