Model v0.2 · November 2026
How the BRI model works
Every number on BRI is our own estimate, computed the same way for every method. The inputs are cited; the assumptions are labelled; the formula is below.
1. The scenario
Each estimate answers one question: what does one person in the United States earn a month, working on this about 10 hours a week, after 12 months? Fixing the scenario is what makes methods comparable.
2. Unit economics
Each method has a unit-economics model: hourly pay × hours booked, pay per unit × units, pay per 1,000 views × views, price × conversion × audience, or client fees × clients, minus costs such as platform fees, gas or ads. Sourced inputs link to their source on every page; inputs we cannot source yet are marked “BRI assumption” with the reasoning.
3. The reality correction
Earnings figures on the web come mostly from people who succeeded. To correct for that, every estimate is anchored to the US side-hustle distribution: Bankrate's 2025 survey found a median of $200 and a mean of $885 a month. We model earnings as lognormal; that median and mean imply a spread of σ = 1.72, widened by 0.2 for high-competition methods and narrowed by 0.2 for low-competition ones.
The median is a weighted geometric mean of the unit-economics figure and the US median. The weight depends on data confidence: A 70%, B 60%, C 45%. Methods without unit data (grade D) start from the US median, adjusted for trend (×1.2 rising, ×0.7 cooling), competition (×1.15 low, ×0.85 high) and platform risk (×1.1 low, ×0.8 high).
From the median and σ we report the lower quarter (P25), upper quarter (P75) and the top-10% threshold (P90). When the press reports a success case, we show where it falls in this distribution.
4. Data confidence
- A: the main pay figure is sourced, at most one input is a BRI assumption, and the evidence includes platform data, a survey or job listings plus three or more independent sources.
- B: the main pay figure is sourced, with platform/survey data or three independent sources.
- C: the main pay figure is sourced from a single secondary source.
- D: no sourced pay figure yet; the estimate is the adjusted US prior.
4b. How the data stays current
The index updates itself. Every day an automated research job revisits the methods checked longest ago, looking for newer pay rates, fees, platform changes and demand signals. Each new figure must come with the exact text from its source; the job opens that page and checks the figure is really there. Verified figures are applied, unverified ones wait for an editor, and a change larger than three times the previous value always waits for review. Then the whole index is recalculated and a daily snapshot is stored, which is what the history charts show.
4c. AI impact
Each method carries one of three labels: AI-boosted (AI makes the work faster, cheaper or newly possible), AI-disrupted (AI is taking the work, the traffic or the buyers), or neutral. The label is editorial, based on the cited evidence, and is shown next to the reason on each method page.
5. BRI Score
A weighted sum of five parts, each scored 0–100:
| 25% | Momentum: rising 100, stable 50, cooling 10 |
| 30% | Earnings: model median on a log scale, $50 = 0 to $2,000 a month = 100 |
| 15% | Room to compete: low competition 100, medium 60, high 30 |
| 20% | Platform durability: low risk 100, medium 60, high 20 |
| 10% | Data confidence: A 100, B 75, C 50, D 25 |
6. BRI verdict
The verdict is our call on whether a newcomer should start now. It uses the same rules for every method.
| Stage | Signals | Verdict |
|---|---|---|
| Early growth | Rising demand, low or medium competition | Worth joining now (score 60+, platform risk not high), otherwise join with a sharp niche |
| Crowded growth | Rising demand, high competition | Join with a sharp niche |
| Mature | Stable demand, low or medium competition | Steady, not a gold rush (only with a real edge if platform risk is high) |
| Saturated | Stable demand, high competition | Only with a real edge |
| Declining | Falling demand or platforms cutting payouts | Avoid for now |
Each verdict lists the reasons it was reached, who the method suits, and the signal that would change our view. The one-line BRI Take is written by our editors from the cited evidence; it never adds facts the evidence does not support.
7. Sources and what we don't do
Inputs come from platform reports, public filings, surveys, job listings, press and figures people publish about their own business, each labelled by type and date. We do not copy other sites' estimates and do not scrape sites that forbid it. Estimates are third-party figures, not data reported by YouTube or any other platform, and they are not financial advice.
The model is versioned. When inputs or weights change, the version number changes and the previous version stays documented.