Methodology

How we rank

One published rubric, applied identically to every company, with the full results shown.

Published September 13, 2026. Last verified September 13, 2026.

Every ranking on BuyGold comes from a single scoring method, published in full below and applied identically to every company. We weight the things that matter most to real buyers: total cost of ownership, minimums and accessibility, pricing transparency, liquidity, storage and security, IRA terms, sales experience, concierge support, and product breadth. Each company is scored on the same criteria.

Because the weights favor cost and transparency, different companies win different categories. Value platforms win on cost and accessibility. Concierge dealers win on high-touch support, where they earn their higher per-ounce pricing. Bullion retailers win on selection. No company wins everything, and any site that tells you one company is best at all things is selling you something.

Methodology v1.0 Changelog

BuyGold ranks on a published, weighted rubric. We rank on structure and cost, NOT on review volume (which favors older, higher-spend brands). Review data is shown separately as raw, unweighted reader trust signals.

Dimension Weight Tends to win
Total cost of ownership (10yr) 25% platforms (0% premium)
Minimums & accessibility 15% platforms
Pricing transparency 15% platforms
Liquidity / speed to sell 10% platforms
Storage & security 10% tie
IRA terms 10% platforms competitive
Sales experience (low pressure) 5% platforms
Concierge / white-glove support 5% dealers
Product breadth / selection 5% retailers
Total 100%

How each dimension is scored

A dimension with no sourced input for a company is left unscored, never guessed. The company's total is the weighted average over the dimensions that ARE scored, with weights renormalized across them, so every company's total sits on the same 0 to 10 scale regardless of how many dimensions it was scored on. The parity gate protects the connected brand from a thumb on the scale: it fails the build if OWNx is scored on any dimension no other verified company is scored on, or leaves unscored any dimension a majority of verified companies carry.

Total cost of ownership (10yr) 25%

Input: scoring.totalCost.value = the all-in effective annual cost rate as a percent of holdings, computed from PUBLISHED fees only (annual storage percent plus any account or platform fee amortized over a $10,000 allocation). Premiums are excluded because they vary by product and are not a published flat rate.

Scores about 2
About 1.2% per year all-in (high storage plus account fees).
Scores about 5
About 0.75% per year.
Scores about 9
About 0.15% per year or an at-cost model with no percentage storage drag.

Minimums & accessibility 15%

Input: scoring.minimums.value = the smallest purchase a buyer can make in US dollars (purchase minimum; use the account minimum if no separate purchase minimum is published).

Scores about 2
A five-figure minimum that excludes most first-time buyers.
Scores about 5
A few thousand dollars to start.
Scores about 9
$25 or less; fractional buying.

Pricing transparency 15%

Input: scoring.transparency.value = a sourced 0 to 10 assessment of how fully pricing is published, against the anchors. The source must link the company's own pricing page (or its absence).

Scores about 2
No pricing published; a phone call is required to learn the cost.
Scores about 5
Fees published but premiums or spreads are not.
Scores about 9
Full pricing published on-site: premium, storage, account, and sell terms, no contact required.

Liquidity / speed to sell 10%

Input: scoring.liquidity.value = the fee to sell back, as a percent of the sale (sell fee or buy-sell spread the company charges the customer to liquidate). Zero is best.

Scores about 2
A wide dealer buy spread and slow settlement on physical metal.
Scores about 5
A modest spread or fee, settlement in about a week.
Scores about 9
Sell at spot with no sell fee, same-day or next-day settlement.

Storage & security 10%

Input: scoring.storage.value = a sourced 0 to 10 assessment of storage and security, against the anchors. The source must document the depository, allocation, and insurance.

Scores about 2
Unallocated or commingled, or storage terms not documented.
Scores about 5
Allocated storage at a named depository, insurance unclear.
Scores about 9
Allocated, segregated where offered, at a named insured depository, with published audit terms.

IRA terms 10%

Input: scoring.iraTerms.value = the published all-in annual IRA cost in US dollars (account plus custodian plus depository fees where the company publishes them; use the company's own published IRA annual fee if that is all that is disclosed).

Scores about 2
High four-figure first-year IRA cost.
Scores about 5
About $300 per year.
Scores about 9
Under $100 per year, or an at-cost IRA.

Sales experience (low pressure) 5%

Input: scoring.salesExperience.value = a sourced 0 to 10 assessment of sales pressure, against the anchors (low pressure scores high). Source: the company's stated sales model, documented complaint patterns, or a first-hand account.

Scores about 2
Documented high-pressure calling or upsell patterns.
Scores about 5
Standard sales contact, no notable pressure signal.
Scores about 9
Self-serve with no sales contact required.

Concierge / white-glove support 5%

Input: scoring.concierge.value = a sourced 0 to 10 assessment of white-glove support, against the anchors. Source: the company's published support model.

Scores about 2
App or email support only, no human guidance.
Scores about 5
Phone support available on request.
Scores about 9
A dedicated representative who handles rollover paperwork end to end.

Product breadth / selection 5%

Input: scoring.productBreadth.value = a sourced 0 to 10 assessment of catalog breadth, against the anchors. Source: the company's published product catalog.

Scores about 2
One or two metals, no coin selection.
Scores about 5
Common bullion coins and bars.
Scores about 9
Wide catalog including numismatics, fractional coins, and large bars.

Weights are frozen at the published version and change only through a dated entry on the changelog. A ranking that does not match this table is a mistake on our part. Tell us and we will correct it or explain it.