Where every number comes from
The board below is drawn from the same daily file that makes the score. If a source is late, you see it here. If a part of the score had to be dropped for the day, you see that too.
Per-source board
Snapshot as of 2026-09-20. 6 of 7 parts used in today's reading.
| Part of the score | Source | How often it updates | Newest value | Days behind | Allowed delay | Freshness | Raw values published |
|---|---|---|---|---|---|---|---|
| Cash after inflation | Fed H.15, inflation-adjusted 10-year yield | daily | 2026-09-17 | 3d | 7d | fresh | yes |
| Entry price | World Bank Pink Sheet monthly gold price (CC BY 4.0) | monthly | 2026-08-31 | 20d | 45d | fresh | yes |
| Central-bank buying | IMF IRFCL for central banks' gold holdings, plus SPDR Gold Shares tonnes for ETFs (internal) | varies by source | 2026-09-17 | 3d | 45d | fresh | yes |
| US dollar | Fed H.10, broad trade-weighted dollar index | once a week | 2026-09-11 | 9d | 14d | getting old | yes |
| Trader bets | US regulator's weekly trader-bets report (CFTC), professional bets as a share of all open bets | weekly | 2026-09-15 | 5d | 10d | fresh | yes |
| Price jumpiness | Cboe Gold Volatility Index; used internally, raw values not republished per licence | daily | 2026-09-18 | 2d | 7d | fresh | no (see notes) |
| Retail premiums | The Gold Barometer Retail Premium Index (our own daily reading, the middle value across the dealer basket) | daily | 2026-09-20 | n/a | n/a | not counted today | n/a |
Licences, one source at a time
| Source | Licence | Credit shown on the site |
|---|---|---|
| Inflation-protected and plain 10-year Treasury yields, broad dollar index (series H.15, H.10) | US Government public domain | Source: Board of Governors of the Federal Reserve System (US) |
| US consumer price index (CPI-U, series CUSR0000SA0) | US Government public domain | Source: US Bureau of Labor Statistics |
| Trader-bets report (Commitments of Traders, weekly detailed version) | US Government public domain | Source: US Commodity Futures Trading Commission |
| International Reserves and Foreign Currency Liquidity (IRFCL) | IMF, republication allowed including commercial use and derived works | Source: International Monetary Fund, IRFCL |
| Long gold price history for the rebuilt record | Creative Commons Attribution 4.0 | Source: World Bank, Pink Sheet |
| SPDR Gold Shares (GLD) daily holdings | Restrictive site terms. Used only inside our calculations. We publish only the money-in-or-out signal we compute from it. | Derived ETF flow signal only |
| Cboe Gold Volatility Index | Cboe proprietary; raw values are never republished | Cboe credited in the methodology text; the score is a ranked version of the value, not the value itself |
| World Gold Council numbers quoted as evidence | Short-quote use for review and commentary | Source: World Gold Council |
| Displayed spot price | Paid commercial-display licence | Displayed spot from Metals.dev (their unbranded gold price). Branded benchmark rates, such as LBMA and IBA, are never shown. |
| Dealer retail prices | Openly-served public pages; per-dealer values are never republished | The Gold Barometer Retail Premium Index (a median across dealers). Dealer names appear in the methodology text only. |
If a source goes silent
Every source has a limit for how late it may be. If it goes silent inside that limit, the last value is carried forward and marked "getting old" on the board above. If it goes silent past that limit, the missing part is dropped from today's reading. The remaining parts share out the missing weight, so the weights still add up to 100%. The number on the home page shows how many parts were used today ("N of 7 parts used"). Nothing is quietly hidden.
- Daily inputs (TIPS, Cboe Gold Volatility Index, GLD, spot): limit 7 calendar days. Weekly releases (the dollar index): 14 days.
- Weekly input (trader bets from the US futures regulator): limit 10 days.
- Monthly inputs (CPI, World Bank price, IMF reserves): limit 45 days.
How the basket of sellers is chosen, and what that means
The retail-premiums part of the score is a daily median premium over the spot price across a basket of dealers. A dealer is in the basket only when all three of these are true at the same time:
- The dealer's terms of use do not forbid automated reads of their public product pages.
- There is no active anti-bot check (Cloudflare interstitial, CAPTCHA, JavaScript riddle) in front of the product pages. We never try to get around one.
- The dealer accepts the fetcher below. Its name says who we are and links back to this page. Any dealer that does not want to be included can simply block it.
The basket today is 7 dealers and 18 products. The dealers are: SD Bullion, Texas Precious Metals, Camino Coin, Pinehurst Coin Exchange, Miles Franklin, Colorado Gold, and Liberty Coin. Almost every other US dealer falls into one of these groups:
| Reason they are out | Rough count | Dealers we looked at |
|---|---|---|
| Terms of use forbid automated access, or forbid derived works that would cover price data | 16 dealers | JM Bullion, BGASC, Provident Metals, BullionMax (part of the A-Mark group, one shared set of site terms), Gainesville Coins, US Gold Bureau, Kitco store, ModernCoinMart, Universal Coin & Bullion, Asset Strategies International, Sprott Money, Golden State Mint, Hero Bullion, PIMBEX, Pacific Precious Metals, First National Bullion. |
| Active anti-bot check on product pages (we do not try to get through one) | 11 US dealers | Money Metals, Bullion Exchanges, Monument Metals, APMEX, Golden Eagle Coin, Silver Gold Bull, SchiffGold, GoldSilver.com, Bullion Trading LLC, Bold Precious Metals, American Gold Exchange. |
| Their server refuses our named collector outright. We do not work around a block. | 3 dealers | CMI Gold & Silver, DGSE, Fidelitrade. |
| Outside the market this index covers (US retail in US dollars) | 3 dealers | CoinInvestDirect (UK), Atkinsons (UK), Sprott Money (Canadian entity). |
What this means for the number you read. Several of the biggest US dealers by sales volume (JM Bullion, APMEX, Money Metals) are not in the basket. So the number measures premiums where prices are shown openly and the terms allow their use. It is not a size-weighted market average. Watch how the number moves from day to day. That is the useful signal, not the exact level.
The fetcher (says who it is, links back here). Every request comes from an agent string that names the project and links to this page:
Mozilla/5.0 (compatible; TheGoldBarometer/1.0; +https://thegoldbarometer.com/provenance/)How often, and what we do not do. One request per product per day. No login, no account, no session cookie, no clicking through a checkout, no crawling of category pages beyond the specific product pages we already know. The prices are blended into one combined number that starts from a base of 100. Individual dealer prices are never republished.
How to opt out. Block the agent string above at your edge, or email hello@thegoldbarometer.com. Your dealership is removed the same day and listed in the excluded set above. No reason needed.
A note on the sourcing chain
The Federal Reserve Board series (H.15, H.10) are US public domain, no matter which channel we fetch them from. We take them from the producer first: the Board's own Data Download Program. The Fed is retiring its old download service, so we are moving to FRED, the public data site it points to instead. If one path fails, we can use the other.
The Cboe Gold Volatility Index is used inside the pipeline under Cboe's rules. The score for that part is a ranked version of the value against its own past. It is designed so you cannot work back from it to the raw value without our full internal history. The raw values are never published.
The disclaimer
The Gold Barometer measures gold buying conditions. It does not give investment advice. It does not recommend any transaction. It is not a broker, a dealer, or a registered investment advisor. The numbers are here for research and press use. See the about page for how the site stays neutral, and the methodology page for how every published number is worked out.