When a European retail investor's portfolio drifts from its target allocation, the rebalancing logic has to answer two distinct questions. First: which positions to trade. Second: in what order to execute the sells. The first question gets most of the attention. The second is where the money lives.
Selling the wrong lot — the one carrying embedded gains rather than the one sitting at a loss — can generate 40 basis points of unnecessary tax drag before the market has moved a tick. On a 10,000 EUR portfolio, that is 40 EUR handed directly to the tax authority for no reason other than the order in which the trades were sent. Scale that across 50,000 accounts rebalancing quarterly, and the aggregate number is no longer small.
This piece walks through why sequence matters, how different European tax wrappers impose different rules, and what a rebalancing engine needs to know to get it right.
The lot selection problem
When you buy the same ISIN at different prices over time, you accumulate multiple tax lots — distinct purchase tranches with distinct cost bases and holding periods. When you need to sell, which lot you sell determines how much capital gain (or loss) you realise.
Consider a simple example. An investor holds three lots of a Xetra-listed ETF (ISIN DE000A0H0744):
- Lot A: 20 shares acquired at 48.00 EUR each (cost basis 960 EUR)
- Lot B: 15 shares acquired at 52.00 EUR each (cost basis 780 EUR)
- Lot C: 10 shares acquired at 44.00 EUR each (cost basis 440 EUR)
Current price: 56.00 EUR. The portfolio requires selling 12 shares to rebalance back to target. Which lot do you sell?
If you sell Lot A (FIFO — first in, first out): gain of 96 EUR (12 × 8.00 EUR gain per share).
If you sell Lot B: gain of 48 EUR (12 × 4.00 EUR gain per share).
If you sell Lot C: gain of 144 EUR (12 × 12.00 EUR gain per share).
In a taxable account subject to German Abgeltungsteuer (flat 25% capital gains tax), the difference between selling Lot B versus Lot C is 24 EUR of tax on a transaction that accomplishes the same rebalancing outcome. On larger positions, this gap widens considerably.
FIFO — the default in most retail brokerage systems — is not optimal. It is simply the path of least resistance.
How tax wrappers change the calculation
The wrinkle is that European retail investors often hold the same underlying securities across multiple account types with completely different tax treatment. The optimal sell decision in one wrapper is not the optimal sell decision in another.
ISA (Individual Savings Account, UK)
Inside a UK ISA, capital gains are fully sheltered. It makes no difference which lot you sell — the gain is exempt from CGT regardless. The correct lot selection strategy inside an ISA is therefore to ignore cost basis entirely and optimise for execution quality: sell the most liquid lot, the one with the smallest bid-ask spread, or the one that avoids a fractional share residue that is hard to rebalance later.
The common mistake is applying the same tax-optimal lot selection logic inside an ISA that you apply in a GIA. This wastes computation and, more importantly, leaves tax-loss harvesting candidates unused in the taxable account because the system consumed them in the wrong wrapper.
GIA (General Investment Account, UK)
In a GIA, capital gains above the annual CGT exemption (£3,000 for the 2024/25 tax year) are taxable at 18% or 24% depending on the investor's income band. The optimal strategy is to harvest losses first — sell the lot with the highest unrealised loss to offset gains elsewhere — and defer embedded gains where possible. The cost basis of each lot, its acquisition date, and the current price all matter.
There is an additional constraint specific to the UK: the 30-day wash sale rule equivalent (the "bed and breakfasting" rule under TCGA 1992). If you sell a security and buy it back within 30 days, the loss is disallowed for CGT purposes. A rebalancing engine that handles GIA tax-loss harvesting must track this window per ISIN and wrapper to avoid triggering disallowance.
Depot (Germany and Austria)
German and Austrian retail accounts (Depot) are subject to Abgeltungsteuer — a flat 25% withholding tax on dividends and realised capital gains, plus a 5.5% solidarity surcharge, making the effective rate approximately 26.375%. Brokers in Germany withhold at source.
Unlike the UK, Germany does not have an annual CGT exemption (the Sparer-Pauschbetrag covers only 1,000 EUR per person in 2024, covering interest and dividends rather than capital gains specifically). The marginal lot selection benefit in a Depot is therefore to defer gain realisation, sell loss-making lots where available, and avoid realising long-term positions unnecessarily.
German brokers typically default to FIFO for cost basis tracking. An active lot selection strategy that picks highest-gain lots last can save 10–35bps versus FIFO on a diversified portfolio rebalancing quarterly.
PEA (Plan d'Épargne en Actions, France)
The PEA is a French tax-advantaged account that exempts capital gains from income tax after a 5-year holding period (subject to social charges at 17.2%). The wrapper has strict eligibility rules: only equities from EEA-resident companies can be held, which excludes most US equities and many global ETFs.
Inside a PEA, lot selection follows a different logic: the account has a global cost basis rather than lot-by-lot tracking for the purpose of withdrawal taxation. Partial withdrawals before the 5-year mark trigger full account closure and taxation of the entire gain. For rebalancing purposes, the PEA's lot selection is therefore constrained by the wrapper's redemption rules rather than per-security optimisation.
The relevant question for PEA rebalancing is not which lot to sell but whether to rebalance at all inside the PEA or to handle the drift via contributions or by rebalancing other account types. This requires the engine to model cross-wrapper positions holistically.
SIPP and CTO: additional wrapper contexts
For completeness, two other wrappers appear with enough frequency in European retail portfolios to warrant mention. The UK Self-Invested Personal Pension (SIPP) is another tax-sheltered vehicle, like the ISA in that gains are not subject to CGT within the wrapper — but unlike the ISA, withdrawals in retirement are taxed as income. Lot selection inside a SIPP, therefore, follows similar logic to the ISA: ignore cost basis, optimise for execution quality.
The French Compte-Titres Ordinaire (CTO) is the French equivalent of a GIA — a standard taxable brokerage account. Gains are subject to the prélèvement forfaitaire unique (PFU) at 30%, which combines income tax (12.8%) and social charges (17.2%). Like the German Depot, lot selection in a CTO should prioritise deferring gain realisation and harvesting available losses. Unlike the Depot, the CTO does not have source withholding — taxes are settled annually via the investor's income tax return.
Cross-wrapper sequencing
When an investor holds the same underlying security across multiple wrappers, the rebalancing engine must decide not just how much to sell but where to sell it. This is the sequencing problem that most platforms underestimate.
Consider an investor who holds a European equity ETF in both an ISA (100 shares, 30% of ISA) and a GIA (80 shares, 25% of GIA). The overall portfolio has drifted and requires selling 30 shares of this ETF to restore target allocation. Where should those sells come from?
Naively, you might split the sell proportionally across accounts. But the correct answer depends on the embedded gain in each account's lots, the investor's remaining CGT allowance for the year, and whether there are harvesting opportunities elsewhere in the GIA that should take priority.
In most cases, the optimal sequence is:
- Satisfy the sell requirement first from the ISA, where gains are sheltered
- Use any remaining GIA sells as an opportunity to harvest losses, not gains
- Only realise GIA gains when the ISA capacity has been exhausted and the gain cannot be deferred
This requires the engine to hold a full view of all wrapper positions, not just per-account portfolio state.
A worked scenario: ISA/GIA/Depot cross-wrapper rebalance
To make the cross-wrapper logic concrete, consider a UK investor who also holds a German brokerage Depot acquired when they previously lived in Germany. Their portfolio holds MSCI World exposure across three wrappers: 12,000 GBP in an ISA (100 shares, embedded gain of 28%), 7,000 GBP in a GIA (58 shares, embedded gain of 9%, plus 10 shares in a different ETF sitting at a 400 GBP unrealised loss), and 6,500 EUR in a Depot (55 shares, embedded gain of 18%).
The portfolio has drifted: the global equity sleeve is 6 percentage points overweight relative to target. The engine must sell approximately 25 shares of MSCI World exposure. Where to sell?
Tax-optimal logic: sell 18 shares from the ISA first (gain is sheltered, no CGT cost), then harvest the 10-share loss position in the GIA to offset the remaining 7-share sell (realising the 400 GBP loss as a CGT offset), and leave the Depot position untouched. The Depot sell would trigger Abgeltungsteuer withholding at 26.375% on embedded gains — avoidable here because the ISA and GIA can satisfy the rebalancing requirement without touching the German account.
Without cross-wrapper visibility, a naive FIFO approach might sell proportionally across all three accounts, triggering Depot withholding unnecessarily and missing the GIA loss harvesting opportunity. The after-tax cost of the naive approach on this rebalance is approximately 180 EUR higher than the tax-optimal sequence.
The 40bps claim, unpacked
The 40bps figure cited in the context of wrong-lot selection comes from analysing portfolios with typical European retail profiles: a mix of equity ETFs, bond ETFs, and a small allocation to individual equities, held across ISA and GIA with embedded gains averaging 8–15% of position value.
In this scenario, the difference between FIFO lot selection and tax-optimal lot selection (sell losses first, defer gains, prioritise ISA capacity for gain-realisation) runs to approximately 30–50bps of portfolio value per rebalancing event for the taxable portion of the account. The 40bps midpoint is illustrative, not a guarantee — actual savings depend on portfolio composition, holding periods, current price levels, and the investor's marginal tax rate.
We are not saying that FIFO lot selection is negligent or that it produces uniformly bad outcomes. For portfolios with minimal embedded gains — recently constructed portfolios, or portfolios in persistent drawdown — FIFO and tax-optimal selection converge. The gap emerges specifically when there is meaningful dispersion in embedded gains across lots. That condition is common in portfolios that have been running for two or more years in a generally positive market environment.
What is not illustrative: the direction of the impact. FIFO always underperforms tax-optimal selection when there is dispersion in cost basis across lots. The magnitude varies; the sign does not.
What a rebalancing API needs to encode
To implement tax-optimal sequencing correctly, a rebalancing engine requires:
- Per-lot cost basis: acquisition price, acquisition date, and original quantity for each tax lot per ISIN per wrapper
- Wrapper classification: ISA, GIA, PEA, Depot, SIPP, CTO — each with its jurisdiction-specific tax treatment rules
- Cross-wrapper position view: the ability to consider sells across wrappers together rather than in isolation
- Tax-year context: remaining CGT allowance (UK), whether the PEA 5-year period has been reached, current year's realised gains for wash-sale tracking
- Harvesting pass: a pre-sell scan that identifies loss-making lots available for harvesting before the rebalancing sell is executed
Without per-lot cost basis data, none of this is possible. Platforms that aggregate positions to a single average cost basis lose the ability to do tax-optimal selection entirely. The infrastructure decision to track cost basis at lot level or average cost level, made early in platform development, has compounding consequences for every rebalancing event that follows.
Data sourcing and custody layer constraints
A persistent obstacle for platforms building lot-level tax optimisation is that custody and brokerage layers frequently provide only averaged cost basis data, not lot-level records. Some custodians provide lot data in CAMT.053 ISO 20022 message format as part of their end-of-day statement feeds; others aggregate to a single average cost in their position APIs.
If your custody integration provides only averaged cost basis today, the path to lot-level optimisation runs through either switching custodians or maintaining a shadow lot ledger — tracking purchases and sells at lot level within the rebalancing engine independently of what the custody layer reports. A shadow lot ledger is feasible but requires the engine to reconcile against custody-reported positions at each rebalancing cycle to detect discrepancies from corporate actions, dividend reinvestments, or manual trades made outside the rebalancing system.
The shadow ledger approach also introduces a class of edge case: what happens when a client transfers a position in from another broker? The incoming position has no lot history in the shadow ledger. Common practice is to treat the transfer date as the acquisition date and the current market price as the cost basis, accepting some loss of tax optimisation accuracy on transferred positions. This is a known limitation, not a failure of the approach.
Practical implications for platform engineers
If you are building or integrating rebalancing logic, the sequencing problem has a few concrete engineering implications worth internalising.
Lot-level data is a prerequisite, not a nice-to-have. You cannot retroactively reconstruct tax lots from averaged position data. If your custody layer gives you only averaged cost basis, you lose the ability to do tax-optimal rebalancing. This is worth flagging early in conversations with your custody provider or prime brokerage.
Wrapper-agnostic engines will underperform. A rebalancing engine that treats all positions identically, regardless of their tax wrapper, will default to FIFO and leave the tax drag in place. The wrapper classification has to feed directly into the lot selection algorithm, not be an afterthought applied at reporting time.
Simulation is essential before go-live. The interaction between lot selection, wash-sale rules, and cross-wrapper sequencing produces edge cases that are difficult to reason about theoretically. Running rebalancing scenarios in simulation mode — with real or realistic portfolio data — before the logic runs live is the only reliable way to catch sequencing errors before they become tax events.
Annual reviews of tax-year parameters are not optional. CGT exemption thresholds change. The UK annual exemption fell from £12,300 to £6,000 and then to £3,000 across recent tax years. The logic that correctly computes remaining allowance in one tax year can quietly produce wrong answers the following year if the threshold is hardcoded rather than drawn from a configurable parameter. Tax-year parameters need explicit maintenance in any long-running rebalancing system.
The 40bps problem is not a marginal optimisation. For a platform managing portfolios at scale, it is the difference between offering a service that genuinely serves clients' after-tax returns and one that simply executes trades in the order they arrive.
Sequence matters. The infrastructure has to know that.