The two dominant asset-model patterns in tokenised private issuances are the basket model (a single token represents a pro-rata claim on a pool of assets) and the share-class model (multiple token classes represent different waterfall positions within the same vehicle). Each has appropriate use cases; each has characteristic failure modes. This paper compares them systematically, drawing on 22 production issuances, and identifies the common pattern of choosing the basket model when the share-class model would have been more appropriate.
- 01The basket model is appropriate for open-ended evergreen pools with homogeneous assets; the share-class model is appropriate for closed-end private vehicles with heterogeneous investor positions.
- 02Of the 22 issuances surveyed, 9 used the basket model in a situation where the share-class model would have been more appropriate; none used the share-class model inappropriately.
- 03The bias toward the basket model is technological: ERC-4626 is simpler to implement than ERC-3643 with custom share-class logic, and implementation teams default to the simpler primitive even when it fits poorly.
- 04The cost of the mismatch is primarily borne at the point of redemption and waterfall distribution, where basket tokens cannot express the per-class preferences that private investors have typically negotiated.
1. Two patterns, two use cases
The basket model (one token, pro-rata claim on the pool) is the natural structure for evergreen open-ended funds where all investors hold economically identical positions. ERC-4626 encodes it cleanly. The share-class model (multiple token classes, differentiated waterfall positions) is the natural structure for private closed-end vehicles where institutional investors, family offices, and sponsors hold economically distinct positions.
Neither pattern is universally superior. The relevant question is whether the underlying economic structure is homogeneous or heterogeneous across investors.
2. The implementation bias
ERC-4626 is simpler. It has better tooling, more production-tested audits, and a larger pool of engineers who have shipped it. ERC-3643 with custom share-class logic is more capable but less accessible.
When a team new to tokenisation confronts the choice, the simpler path frequently wins, even in situations where the underlying economics are heterogeneous and the basket model will force the issuer into contortions at redemption.
3. Where the mismatch bites
The mismatch shows up at redemption and waterfall distribution, not at issuance. At issuance, a basket token serving heterogeneous investors looks like it works: different investors subscribe at different NAVs, and the token accounting reflects pro-rata claims at subscription.
At redemption, the mismatch surfaces. Preferred returns, hurdles, and catch-up provisions are features of the share-class model; they are not expressible in the basket model without either off-chain reconciliation (defeating much of the purpose of tokenisation) or a retrofitted share-class layer grafted onto the ERC-4626 vault (which works but is awkward and error-prone).
4. Recommendations
Pick the asset model before the token standard. If investor positions are genuinely homogeneous, the basket model is clean and should be preferred. If they are heterogeneous in any material way, accept the implementation cost of the share-class model rather than deferring it to an operational problem that will eventually cost more to retrofit.
This paper will be available as a signed PDF. The Association publishes PDFs of all research on release; they carry a cryptographic signature anchored to a Swiss qualified electronic-signature provider to ensure provenance.
Boli Association. (2026). Basket tokens versus share-class tokens: a design comparison. Boli Association Working Paper No. BA-2026-08. Zurich.