Daily, weekly, or monthly: when does the cadence pay for itself?

Daily-cadence rebalancing is expensive in operational terms. So is weekly. Monthly is cheap. The question that determines which cadence belongs on which account is whether the additional harvest stream from a tighter rebalance schedule pays for the additional operational and trading cost. The honest answer is size-dependent, and the break-even NAV is lower than most practitioners assume — but not at the very small end of the market.

We ran the same vintage of tax-aware DI on a US large-cap (100) universe at eight account sizes, with three rebalance cadences. Same constraint set, same risk model, same marginal rates. Trading cost modelled per-name with a square-root impact term plus a flat 1 bp commission. Operational cost modelled as a fixed per-rebalance overhead — solver compute, order generation, reconciliation, surveillance — that's the same in dollars regardless of account size.

The cadence-premium curves are concave in NAV: at very small accounts the operational overhead dominates and all three cadences struggle; past about $250k each cadence settles into its asymptotic alpha. Daily wins above ~$400k; weekly is competitive in the $80k–$400k band; monthly is the right choice below $80k.

The case for daily isn't more harvest per dollar. It's more opportunities to harvest what already exists, before the replacement security has time to drift on you.

• More harvest opportunities don't expire. A name that drops on Tuesday and recovers on Friday is a harvest the daily cadence catches; the weekly cadence measures Friday vs Friday and never sees the dip. In high-vol regimes (2020, 2022) this is the largest mechanism.
• Tighter wash-sale-aware substitution. The 30-day lockout is the same regardless of cadence, but daily rebalancing has finer control over when the lock starts. The optimizer can sequence harvests to keep the substitution window short.
• Lower drift between rebalances. Daily realised TE is by construction smaller than weekly or monthly. The hard TE_max constraint binds less often, so the tax penalty has more weight in the objective.

Two cases tilt toward weekly:

• Sub-$400k NAV. The per-rebalance op cost is fixed in dollars; at $250k that $80/rebalance is 3.2 bp/yr if monthly, 32 bp/yr if daily. Daily eats too much fixed overhead for the harvest to recover.
• High-spread universes. On thinly-traded constituents (small-cap indices, emerging markets) the per-trade cost rises faster than the harvest. Weekly rebalancing reduces trade count and is competitive with daily even at larger NAVs.

Cadence is per-account on the platform — the DailyRunner treats every account as eligible to rebalance every day, but the strategy module's `should_rebalance(as_of)` hook returns False for non-daily accounts on non-rebalance days. A weekly account solves only on Monday's morning run; a monthly account only on the first business day of the month. The same orchestration path serves all three; the only thing that changes is the boolean gate.

• Manager attention cost. A daily-rebalanced account requires daily compliance review, daily trade reconciliation, daily exception handling. The op cost number ($80/rebalance) folds compute and storage but not the people-time at scale. Real fleets need to add a per-account staffing layer.
• Custodian batching. Some custodians batch trades across accounts and pass the slippage advantage back. Daily cadence captures the batch benefit; weekly may miss it.
• Tax-loss carryforwards. Households with substantial existing carryforward already have headroom for harvested losses; the marginal value of an additional harvest is full. For households with no carryforward and no near-term external gains, the marginal loss is worth less, and the cadence math shifts.

For the orchestration layer that runs the cadence decision in production, see The DailyRunner.

Educational illustration · synthetic sweep · numbers size-dependent. The right cadence is account-by-account.