Direct Indexing vs ETF Tax Loss Harvesting: Why More Lots Beat More Funds

The choice between direct indexing and ETF investing for tax loss harvesting purposes comes down to a single arithmetic question: how many independently harvestable units does each approach give you? An ETF holding the S&P 500 is, from a tax perspective, one unit. The fund either has an unrealized loss or it doesn't. If the index is up 15% for the year, every share of every lot you own in that ETF is sitting on a gain, and there is nothing to harvest. A direct indexing portfolio that replicates the same S&P 500 by holding the underlying 500 stocks is, from a tax perspective, 500 independently harvestable units. Even when the index is up 15%, somewhere between 100 and 200 of those individual stocks will be down for the year — and every one of them is a separate harvesting opportunity. The mathematical case for direct indexing over ETFs in taxable accounts is essentially this: more lots means more harvest. The hard part is that nobody can manage 500 individual stocks manually. This is the problem AI-driven portfolio software was built to solve.

The ETF is one of the great financial innovations of the last 35 years. For investors building wealth in tax-advantaged accounts — 401(k)s, IRAs, HSAs — there is no meaningful reason to prefer direct indexing. ETFs are cheap, simple, and tax-efficient inside the wrapper. But in a taxable account, the simplicity of the ETF is also its limitation. The fund treats the entire basket of underlying holdings as a single unit. When markets fluctuate, the individual stocks inside the fund move in different directions on different days, but the ETF investor only sees the blended result. The harvesting opportunities at the individual stock level — losses that exist somewhere inside the basket on most trading days — are inaccessible. They get netted out before they ever reach the investor's tax statement.

Direct indexing fixes this by replacing the fund wrapper with direct ownership of the constituent stocks. The investor's account holds 250-500 individual positions in approximately the same weights as the target index. The portfolio's overall performance tracks the index closely, with tracking error typically under 100 basis points per year for well-constructed strategies. But because the individual stocks are held directly, every one is independently observable for tax purposes. When Coca-Cola is down 8% on a given day while the rest of the index is up, the Coca-Cola position can be harvested. When Tesla drops 15% during a sector rotation, the Tesla position can be harvested. The portfolio captures losses from idiosyncratic stock movements that the ETF would have absorbed silently into its blended NAV.

How Much More Loss Can You Actually Harvest?

The empirical answer is well-documented. Multiple studies of direct indexing strategies — by Wealthfront, Parametric, Aperio, and academic researchers — converge on a similar range: direct indexing typically produces 1% to 2% per year in additional after-tax return compared to a comparable ETF strategy in a taxable account, attributable almost entirely to harvested losses that the ETF cannot access. On a $1 million taxable portfolio, that's $10,000 to $20,000 per year in additional after-tax return. Compounded over 20 years at typical market growth rates, the difference between direct indexing and ETF investing in a taxable account runs into hundreds of thousands of dollars — purely from tax mechanics, with no change in the underlying investment exposure.

The mechanism is straightforward. In any given year, a broad-market index might be up 12%. Inside that index, the individual stocks distribute themselves widely: maybe 60% are up, 40% are down, and the down stocks span a range from 1% losses to 35% losses. An ETF investor sees +12% and harvests nothing — there's no loss to capture at the fund level. A direct indexing investor harvests every losing stock as the year progresses, replacing each with a correlated substitute (often another stock in the same sector or a small temporary tilt away from the index), banks the losses, and finishes the year with both the +12% index return and a substantial pile of realized losses available to offset gains elsewhere.

Even in years when the index itself is down — 2022, for example, when the S&P 500 dropped about 18% — direct indexing still wins. The ETF investor can harvest the entire position once and is done. The direct indexing investor harvests losses on the way down, harvests again on small bounces, harvests every individual stock that diverges from the index path, and accumulates dramatically more total realized loss than a single ETF sale could ever produce.

The Lot Multiplication Effect

The arithmetic of harvesting is governed by what TaxHarvest treats as the central insight of Capability #3 — loss identification — applied at scale. The number of harvestable units per dollar of portfolio determines the ceiling on how much loss can be captured.

Worked example. Compare two $500,000 taxable portfolios over a single calendar year, both invested to track the S&P 500.

Portfolio A: ETF strategy. The investor owns one ETF (VOO) accumulated over 36 monthly purchases — so 36 lots of a single security. Of those 36 lots, in a typical year with the index up 10%, perhaps 8-10 lots will be sitting at a temporary unrealized loss at some point during the year. Realistic harvestable losses from this portfolio: $3,500 to $6,000 over the full year, including the small losses captured on individual purchase lots that briefly went underwater.

Portfolio B: Direct indexing strategy. The investor owns 350 individual stocks in approximately S&P 500 weights, with each stock held in 6-12 lots from periodic purchases and rebalancing. Total lots in the portfolio: roughly 2,800. In the same year with the index up 10%, somewhere between 600 and 1,000 of those lots will spend at least some portion of the year in a loss position. Realistic harvestable losses from this portfolio: $22,000 to $38,000 over the full year.

For an investor in the 23.8% combined rate (top federal bracket plus NIIT), the tax savings difference: $5,236 to $9,044 per year. For someone in the 18.8% bracket (15% LTCG + NIIT): $4,136 to $7,144 per year.

These are not edge cases. They are typical results from typical years for typical portfolios. The driver is mechanical: more lots, scanned more frequently, produces more captured losses. The ETF cannot replicate this no matter how aggressively it is harvested at the position level, because the position level isn't where the opportunities live.

Why Direct Indexing Used to Be Out of Reach

For most of the history of investing, direct indexing was theoretically attractive but practically impossible for normal investors. The math worked. The execution didn't.

Three barriers blocked it. First, fractional shares weren't widely available, so replicating a 500-stock index required enough capital to buy at least one share of every constituent — which for a stock like Berkshire Hathaway alone meant tens of thousands of dollars. The minimum portfolio size for direct indexing was historically $1 million or more. Second, transaction costs at $5-10 per trade made the constant rebalancing and harvesting required by direct indexing prohibitively expensive — a single rebalance of 500 stocks could cost $5,000 in commissions. Third, and most importantly, manual management was infeasible. No human can monitor 500 stocks daily for harvesting opportunities, evaluate wash sale risk against pending purchases, replace harvested positions with appropriate substitutes, and rebalance the portfolio to maintain index tracking — all simultaneously, every market day.

All three barriers have collapsed. Fractional shares are universally available. Transaction costs at major brokerages are zero. And the third barrier — the management problem — has been solved by software. Specifically by AI-driven portfolio management systems that handle the lot-level scanning, harvesting, replacement, and rebalancing decisions automatically, making decisions across thousands of lots simultaneously in ways that no human portfolio manager can match.

This is the actual reason direct indexing has become a viable retail strategy in the last decade. Not because the math changed — the math has always favored direct indexing in taxable accounts. The math became executable. The constraint was always the management, not the concept.

What Software Does That a Human Can't

Even setting aside direct indexing specifically, the core argument for AI-driven tax loss harvesting versus manual approaches comes down to four capabilities that scale poorly under human attention.

Continuous scanning. A human checks their portfolio at most weekly, more typically monthly, and most realistically once a year in December. Software scans every lot every market day. The losses that exist for three days in March and disappear by April are invisible to the human and routinely captured by the software. Across a year, this difference alone can double the total captured loss.

Multi-variable optimization. The optimal lot to harvest is a function of cost basis, holding period, current-year tax position, wash sale risk, replacement security correlation, and bracket effects. Six variables, each interacting with the others. A human cannot reliably optimize across six variables on the fly during a trade decision. Software can — and the optimal answer is frequently not the obvious one. Manual harvesting almost always produces second-best decisions because the variables aren't all visible at the moment of choice.

Wash sale rule enforcement. The wash sale rule disallows losses if a substantially identical security is purchased within 30 days before or after the sale, in any account belonging to the investor (including IRAs and spouse's accounts). Tracking this manually across multiple accounts, multiple positions, and multiple time windows is genuinely difficult. Software does it automatically. The IRS reports that improper wash sale handling is one of the most common errors on amended returns involving capital gains.

Direct indexing-scale execution. A 350-stock direct indexing portfolio requires hundreds of trades per year — for harvesting, replacement, and rebalancing. The execution is straightforward in isolation but combinatorially complex when coordinated across the full portfolio with all the constraints above. Without software, direct indexing collapses into an unmanageable spreadsheet exercise. With software, it runs in the background.

The case for AI-driven management is not that AI is intrinsically smarter than human investors. It's that the underlying optimization problem has too many variables and too many lots and changes too often for human attention to track. Software solves a scaling problem, not a judgment problem. The judgment behind the strategy — that tax loss harvesting is worth doing, which indices to track, what the investor's tax situation requires — remains the human's. The execution is the part that benefits from automation.

When ETFs Are Still the Right Choice

Direct indexing is not always the right answer. Three scenarios favor ETFs.

Tax-advantaged accounts. Inside a 401(k) or IRA, harvesting losses produces no benefit because gains and losses don't flow through to the investor's tax return. ETFs are simpler, cheaper, and equally efficient inside the wrapper. Direct indexing inside a tax-advantaged account is pure overhead.

Small portfolios. Below about $100,000 in taxable assets, the cost of direct indexing services often exceeds the harvested-loss benefit. The math depends on the specific platform and the investor's tax bracket, but as a rough rule, the breakeven is somewhere in the $50,000 to $150,000 range for most providers. Below that threshold, a low-cost ETF held in a taxable brokerage account produces a similar net result with less complexity.

Investors in or near the 0% LTCG bracket. For retirees and other investors with taxable income low enough to qualify for the 0% long-term capital gains rate, harvested losses are worth less than they appear — there's no tax to offset. The case for direct indexing is significantly weaker in this situation, though not zero (the matched-pair strategy still applies, and losses can still offset ordinary income up to $3,000/year). For investors above the 15% bracket and certainly above the 23.8% combined rate, direct indexing's advantage compounds dramatically.

For investors in the broad middle — taxable portfolio above $100,000, ordinary income in the 22-37% range, long-term gains in the 15-23.8% range — direct indexing in a properly managed program produces meaningfully better after-tax returns than ETF investing. The advantage is not marginal. It accumulates to six and seven figures of additional wealth over multi-decade timeframes, paid for almost entirely by the federal tax code.

For background on the rate brackets that determine when direct indexing pays off, see capital gains tax rates 2026. For investors above the NIIT threshold, the marginal value of every harvested loss is meaningfully higher — see our net investment income tax NIIT 2026 piece. The mechanics of finding harvestable losses inside winning positions — which apply to direct indexing portfolios with extra force given how many positions are involved — are covered in our unrealized losses hiding in your winners deep dive. And for a worked-through example of how lot-level harvesting compounds, see Sarah's NVDA case study.

The choice between direct indexing and ETF tax loss harvesting is, in the end, a choice about how much of the underlying market's natural volatility you want to capture as realized tax losses. ETFs let you capture the volatility of one ticker symbol. Direct indexing lets you capture the volatility of every stock inside the index. The math is simply better at the second level — but it only works if the management problem is solved. Software solves the management problem. The investor solves the strategy.