Our Methodology: Чистый капитал Estimates & Tool Reviews

Обновлено July 2026

This page covers both halves of the site. For how we estimate net worth, jump to How We Calculate Чистый капитал. For how we test and score software, start here.

We put every tool through the same process before we write a word about it. The goal is simple: figure out whether it genuinely helps your business run better, save time, or bring in more revenue, and then say so plainly.

We test with real, paid accounts and live with each tool inside the kind of workflows people actually run. Demos and sales pages only show the polished version. We want to see where it gets clunky, what breaks under normal use, and how much setup it really takes before you get any value out of it. Because software ships changes constantly, we revisit our picks so the advice stays current.

What We Test

We spell out the obvious strengths and the less obvious weaknesses, then say who a tool fits well and who should look elsewhere. For outreach and marketing platforms we dig into deliverability, warmup performance, inbox placement, and whether the sending limits hold up once real volume flows through. Whatever the category, we note what changed since our last look so you always see the current picture.

How We Score Every Tool

Each tool is judged on the same six criteria. That keeps our reviews consistent and lets you compare options on the things that actually decide whether a purchase pays off.

How We Stay Independent

Some of the links on this site are affiliate links. If you buy through one, we may earn a commission at no extra cost to you. Those commissions fund the paid accounts and the hours of hands-on testing that go into every review.

We also welcome a right of reply. If a company spots a factual error in our coverage, they can flag it and we will review the evidence, correct what needs correcting, and note what changed.

The same way we used to lay out every assumption and variable when breaking down net worth numbers, we show our reasoning so you can decide whether the conclusions make sense for your own business.

How We Calculate Чистый капитал

Rates and measured data current as of 2026-08-03.

Prefer the story to the tables? How we calculate celebrity net worth, behind the scenes walks through how this system was built, including the mistakes that shaped the rules.

Every YouTuber page on this site is computed by one deterministic program from three kinds of input: measured public data, published rate tables, and documented one-off facts such as a company sale. The same inputs always produce the same number. Nothing on a page is written by hand, and any income lane we cannot yet support with public data is listed openly as one we are still sourcing.

The pipeline, year by year

1. Measure. Channel catalogs come from the YouTube Data API: views, publish dates, Shorts versus long form (60 seconds is the cut), and which videos carry paid promotions. Sponsored share is measured, never assumed, by combining YouTube’s own paid-promotion flag with description scans for disclosure language, minus the creator’s own brands. Music streams come from Kworb’s public Spotify tracker, lead-artist tracks only.
2. Date the views. Ad revenue depends on when views happened, not when videos were published. We rebuild each channel’s total-view history from Internet Archive snapshots of its Social Blade page, anchor the series to the live API total, and allocate views to calendar years along it. Channels without at least three snapshots spanning three years fall back to publish-year cohorts, and the page says so.
3. Price the views. Each calendar year’s long-form views earn that year’s rate: a genre base (median of the sourced disclosures below, each normalized to its era) times an era multiplier times an audience-geography multiplier. Shorts earn a separate rate, and nothing before YouTube turned Shorts monetization on in February 2023. Measured sponsored views earn sourced integration rates, less a 15% management fee.
4. Subtract real-world costs. Production costs by intensity tier (7.5%, 20%, or 35% of gross), then taxes by the jurisdiction the person actually lived in each year, using the residence timeline on their page.
5. Subtract spending. What remains is split between spending and saving along the savings-rate curve economists have measured for US households, rising with income.
6. Compound. Each year’s savings grow at the real calendar-year return of a 60/40 stock/bond portfolio, not a smoothed average. Money earned in 2013 rode 2013’s markets.
7. Add business stakes. Real transactions first, disclosed revenue times a comparable multiple second, modeled revenue times a low multiple last, with a 25% marketability discount on minority stakes in private companies. When a creator sells, channel statistics freeze at the deal date and the proceeds compound instead.

Evidence tiers and page grades

Every line item carries a tier: DISCLOSED (someone with the real number said it on the record), MEASURED (we counted it in public data), MODELED (measured quantity times a sourced rate), and every revenue lane carries one of these tiers. A page’s grade is the share of its value sitting in the first two tiers: A is 60% or more, B is 30%, C is anything below. Most YouTuber pages are C, which is honest: most of a career is modeled, and the grade says how much.

Ad-rate table: every RPM disclosure we use

RPM is revenue per thousand views after YouTube’s cut. Base cells take the median of their rows, US and global rows only, each row first normalized to the modern era. Rows flagged as upper bounds are shown but excluded from medians. Other geographies enter through the multiplier table, never the base.

Two luxury-real-estate creator disclosures exist beyond this table. Enes Yilmazer told the Wall Street Journal in April 2021 he was earning $50,000 to $100,000 a month from YouTube ads, and Erik Conover told the same reporter $10,000 to $30,000 a month. Neither disclosure can be turned into a rate: Yilmazer’s channel has no archived Social Blade snapshots at all, and Conover’s nearest snapshots sit 21 months apart around the disclosure window, wide enough that a derived denominator would be a guess. Both disclosures corroborate the real-estate cell’s direction without being median rows. The cell rests on one aggregate row for now, and the affected page carries a warning saying exactly that.

The base rates the engine actually uses

Era and geography multipliers

Sponsorship rates

Taxes

Effective rates by jurisdiction and year, applied to the residence timeline documented on each page. Corporate structuring that would lower these rates is ignored, which slightly overstates tax and keeps our estimates cautious.

Spending, through the savings curve

Investment returns

Business valuation multiples

Music stream payouts

Estimated lanes

Lanes without direct measurement are estimated rather than zeroed. Merchandise uses sourced print-on-demand benchmarks: 0.5% of engaged fans buying one drop a year at a $25 order, at sourced margins. Streaming royalties beyond Spotify are estimated by scaling the measured Spotify royalties to the whole market, since Spotify holds 32.2% of global music subscribers. Shorts views from before YouTube switched on revenue sharing are priced at the Fund-era floor rate. Channel memberships are estimated wherever a channel has them enabled. Instagram and TikTok brand posts are estimated from a channel’s reach at sourced per-post rates, six paid posts a year, capped at 30% of its YouTube revenue. The benchmark table below publishes every one of these rates with its source, and rows marked as assumptions say so. Real estate enters where a purchase is on record and appreciates at its metro Case-Shiller rate, and a business stake enters where a real transaction fixes its value. Art collections are the remaining lane we are building out.

How we estimate actors

Every actor page runs on the same discipline. A film role earns one of four things. A disclosed salary and documented backend, entered as reported with a citation. For undisclosed lead roles after an actor’s breakthrough, the median of disclosed salaries for that era and budget band (six bands from micro-budget under $5M through blockbusters at $200M and up), blended with the actor’s own disclosed quote history when we have it, floored at 70% of their recent personal max, and capped at 15% of the film’s budget (or at the actor’s proven personal max when that history exists, never above half the budget). Fees that are clearly non-market (union scale, far below a co-star, charity rates) stay on that person’s page as disclosed but are kept out of the median cells so they do not drag every peer down. For supporting roles after breakthrough, a documented fraction of the same lead cell: about 12%, because an established star’s supporting part is not a pre-stardom bit role. And for roles before the breakthrough, the pre-stardom ratio: across the disclosed deals in our table signed by actors who were not yet stars, from Stallone’s $35,000 for Rocky to DiCaprio’s $2.5M upfront for Titanic, the median deal paid about 2.5% of its era’s going lead rate. Films with no reported budget are treated as small productions. Every role is estimated; a role would only enter at zero if someone documented that it was played for free. Documented non-film income enters as dated rows with citations: Friends salaries season by season, Tulsa King at its reported per-episode fees, a Dior partnership at its reported minimum. Representation fees, taxes by era, the savings curve, and real benchmark returns then run exactly as they do for creators, with one addition: where court filings document actual spending, the filings replace the curve.

Three more lanes are estimated rather than zeroed. Undocumented backend enters as an expected value: across our disclosed lead deals, about 12% included points, at a median share of box office, so undisclosed post-breakthrough lead roles carry that expected value against their box office, and films whose points are documented but unquantified carry the median rate directly. A disclosed total that already includes backend is never stacked with another backend estimate. When theatrical box office is missing, residuals and backend expected value use production budget times two as a receipts proxy. Endorsements enter as the number of brand partnerships times a per-brand annual fee and term. For film actors that fee is the median disclosed ambassador deal in the endorsement table below, scaled to the actor’s own quote versus the era median so a mid-tier career is not priced like a Lancome global ambassadorship. Athletes and creators use a modeled per-brand rate, because the endorsement deals reported for them are almost all megastar outliers, and anyone whose own deal is public enters at that reported figure. Producer, director, and writer credits enter at a union-scale floor per credit, scaled to each era. Residuals are estimated from SAG residual rates applied to assumed post-theatrical receipts, split by role and spread over the decade after release. Named TV runs, Broadway engagements, and art sales enter through cited per-person rows. Real estate enters where a purchase is documented and appreciates at its metro rate, and a business stake enters where a real transaction fixes its value; art collections are the remaining lane we are building out.

Multi-decade careers forced two accumulation rules the creator pages never needed. First, the savings rate is capped at 51%, the top bracket the Dynan–Skinner–Zeldes study actually measured; the higher figure in our table is a disclosed extrapolation, and compounding an extrapolation across 40 years would let one assumption dominate the page. Second, portfolio returns are taxed at sourced capital-gains blends for each era before they compound, because untaxed compounding roughly doubled the oldest careers against every known reference point. Documented one-off liabilities also enter with citations, like the reported $90 million Willis–Moore divorce settlement. UK-filed companies are checked against the register exactly as on creator pages: Jason Statham’s and Emma Watson’s production companies file publicly, and both sit far below the model, which the pages report as documented floors rather than additions.

The Depp pilot doubles as a backtest. His former managers’ court records put his 2003–2016 earnings near $650 million. The model, built from film pay and the estimated lanes above, lands under that court-documented total for the same window, and his page reports the exact share along with which lanes hold the rest. A floor that lands under a court-documented total, for documented reasons, is the model behaving.

Two more accumulation choices matter on multi-decade pages. Entertainment households are modeled as saving a bit less of after-tax income than the plain SCF top bracket at the same cash level, unless a court filing replaces the curve. Unmarked savings then compound on a mixed path: a little over half at the actual 60/40 securities return (taxed), the rest at the long-run US home-price rate, because real careers hold more property and private assets than a pure brokerage glide path. TV series with a config credit enter through a per-episode comps table when the fee is not disclosed. We publish one central estimate on each page, not a range.

Endorsement fee disclosures

Hyphenate credit floor

Lane benchmarks

Inflation

A net worth is a figure in today’s dollars, and ours already is: every year’s income is booked in that year’s money and compounded forward at the market’s actual returns, so a 1995 paycheck arrives at its real present-day value. What we also do is run the fixed-dollar assumptions above through the price level of the year they apply to. A $25 merchandise order, a $4.99 membership, a per-post rate, and the savings-rate brackets are all today’s figures, so we deflate them with the Consumer Price Index before applying them to an earlier year. That keeps a 2010 estimate priced in 2010 money. Film salaries are already era-specific, so they need no adjustment.

Real estate, business stakes, and memberships

A documented property purchase adds only its appreciation, never the purchase price, since the money spent is already counted in the person’s savings. We grow each documented purchase at its metro’s long-run rate below, net a 7% selling cost. Business stakes enter from a real transaction where one exists, otherwise revenue times a comparable multiple, with a 25% discount for a minority holding in a private company. Channel memberships are counted only for the channels we confirmed have them switched on: of the roster, seven do and the rest do not, so we do not estimate membership income for a channel that has the feature turned off.

The backtest scorecard

A model that publishes its rates also has to publish its misses. We grade ours against every answer we can find that did not come from another estimator: court records, Companies House filings, real transactions, and on-the-record statements. Street estimates never count as ground truth, because they are somebody else’s model. Three numbers per test, per our published protocol: the median absolute error, the direction of the bias, and the worst miss. Last computed 2026-08-03, and recomputed with every model release.

Test 1: predicting disclosed salaries blind

The engine’s main modeled component prices an undisclosed lead role at the median of disclosed salaries for its era and budget band, blended with the actor’s own disclosed quote history when available, floored at 70% of recent personal max, and capped at 15% of the film’s budget. We turned that mechanism on its own table: for each of the 100 disclosed established-lead salaries above, we removed the row, rebuilt the medians without it, and predicted the figure cold. Median absolute error: 20%. The signed median is -5% and 54% of predictions come in low, so the mechanism leans under, the direction a floor should lean. 84% of predictions land within 50% of the true figure, and on ordinary market-rate deals (excluding known scale/far-below outliers as targets) the core median absolute error is 19%.

Test 2: whole-model checks against documented wealth

Test 3: career income vs court records

Johnny Depp career gross 2003-2016: the model counts $644,256,538 of film income against the $650,000,000 his former managers documented in court, an error of -1%. counted income in the window vs the court-documented total; under is the designed direction when estimated lanes hold the rest.

Tom Cruise career gross 1981-2022: the model counts $837,271,935 of film income against the $1,000,000,000 his former managers documented in court, an error of -16%. model share of documented career-earnings anchor.

Test 4: documented floors

The pattern across all four tests is the same: the model misses low far more often than high, by roughly the size of whatever the exclusions policy forced it to leave out. The KSI row is the clearest lesson this site has produced: no public income lane ever documented a Prime distribution, so the blind model saw 8% of what his own company filings later showed. That is why register data is now an input. If a documented number ever prints below one of our figures, that row goes here too, and the model gets corrected in public.

Every source this site relies on

One rule sits above the hierarchy: outside sources supply individual facts only. A salary, a filing, a settlement, a per-episode fee. No other outlet’s net worth estimate is ever an input to this model, a number we calibrate toward, or ground truth in the scorecard above. We compute the aggregate ourselves, from the facts, and publish every step.

Our sourcing hierarchy, strongest first: official records (company registries, court filings), first-party statements on the record, trade press (Variety, The Hollywood Reporter, Deadline), general press, and last, secondary compilations. One source is banned outright: CelebrityNetWorth, which prints dollar figures without citations and blocks archiving. Any figure that existed only there is treated as undisclosed, and the film models from its budget instead. Remaining secondary sources are accepted only where they attribute a figure to trade or press reporting, are always labeled as what they are, and get replaced whenever the underlying primary report surfaces. Citations point at the outlet that originally reported a figure wherever that original is locatable online; where it exists only in pre-internet print, the verified secondary stands in, names the original publication in the row, and stays on the upgrade queue. The scorecard above exists partly to measure what a table with this mix of sources actually delivers. Below is every domain cited anywhere in the model, with its class and how many citations rest on it.

Reproduce any page

Each page lists its raw inputs, retrieval dates, and rate-table version. The YouTube Data API is free, Kworb and the Internet Archive are public, and every rate above cites its source. Apply the rates to the measured numbers year by year, as shown in each page’s table, and you will land on our figure. If you cannot, write to us, because one of us has a bug.

Why We Show Our Work

Before we created this publication, there wasn't a single website that showed exactly how someone's net worth was calculated. Even major publications refused to disclose their methods, hiding behind vague claims of a "proprietary algorithm." Math is not a secret; it's simply a shield from scrutiny.

That's why our math and reasoning is always transparent. That's NetWorth Explained.