Research

Lendwise is built on original research into how lending yield is priced, reported, and dispersed across DeFi protocols. This page hosts our work.

Working paper

Cross-Protocol Yield Dispersion in DeFi Lending Markets (working title)

Abstract. Placeholder — replace with the paper abstract. We study the dispersion of realized lending yields for identical assets across the major on-chain money markets (Aave V3, Morpho, Compound V3) over multiple chains. Using a normalized, direction-aware APY methodology that converts all reward APRs to compounded APY and nets protocol fees, we quantify the persistent same-asset spread available to lenders, characterize its drivers (incentive programs, utilization dynamics, and fragmentation across chains), and estimate the yield left on the table by capital that does not rebalance.

• Authors: TBD
• Status: Working paper — draft
• Download: PDF link TBD — drop the file in docs/public/ and link it here.

Publishing checklist

1. Add the PDF to docs/public/research-paper.pdf.
2. Replace the abstract, authors, and status above.
3. Link the PDF: [Download the PDF](/research-paper.pdf).
4. Add a citation block (BibTeX) below once there's a DOI or arXiv id.

What the paper formalizes

The research underpins three things you can see in the product:

1. The same-asset spread is real and persistent — not a fleeting arbitrage, but a structural feature of fragmented lending markets. This is the basis of the optimizer.
2. Naïve rate comparison is systematically wrong — mixing APR with APY, or base with net, mis-ranks markets. Our normalization (methodology) removes that bias.
3. Dispersion is measurable and trackable — quantifying it over time is what the GraphQL API exposes.

Citing this work

Once published, cite as:

@techreport{lendwise_yield_dispersion,
  title  = {Cross-Protocol Yield Dispersion in DeFi Lending Markets},
  author = {TBD},
  year   = {2026},
  note   = {Working paper, Lendwise},
  url    = {https://docs.lendwise.fi/research/}
}

Have feedback on the methodology or want the dataset? Reach out on X.