Flat editorial illustration for a finance article about the break-even threshold for upgrading to paid direct-indexing portfolio software. A clean electric-blue gateway arch stands at center with an abstracted gold portfolio token crossing through it. An abstract directional path extends from a soft open circle on the left, through the arch, toward open horizon on the right. The mood is calm, confident, and deliberate — representing the measured decision to move from sandbox simulation to a live brokerage account.
The Sandbox-to-Guided upgrade is a threshold decision: one where portfolio size, marginal tax rate, and annual harvest potential may together determine whether the subscription is likely to pay for itself in after-tax savings.

The Sandbox tier is free for a reason. It exists to answer one question: does the engine think the way I'd want my money managed? Once you've watched a few harvest cycles, the next question is whether to bring real money.

Below the right account size, the math generally does not favor it. A broad-market ETF is cheaper, simpler, and approximately 95% of the way there. Above it, the math flips. This article is about the threshold.

What Guided buys you

What does upgrading from the Sandbox to Guided tier add to HarvestEngine?

Guided connects the engine to a real broker account with live cost basis and wash-sale state, adds the full AI Designer against real capital, and enables the Tax Bank — while keeping manual approval required for every proposed trade.

  • One linked broker. The engine sees your real positions, real cost basis, real wash-sale state. Not a sandbox approximation.
  • The full AI Designer. Same model the Sandbox uses, but designing against the actual capital you have at your actual broker.
  • Manual approve every trade. The engine proposes; you press Approve with PIN. Same review surface as Sandbox; just real orders flow through.
  • Tax Bank, year-round. Realized losses bank against future gains. The dashboard tracks it as a running total per tax year.

What Guided does not buy: autonomous execution (Autopilot), short overlay (Alpha), multi-account households (Alpha). One broker, manual approve, long-only.

LIVE BROKER Real positions + cost basis connected brokerage account live wash-sale state AI DESIGNER Design vs. real capital full portfolio design toolset against actual holdings TAX BANK Year-round loss tracking realized losses accumulate offset future gains MANUAL APPROVE Investor reviews every trade engine proposes, you confirm PIN-confirmed execution
The four core capabilities included in the Guided tier: a live broker connection brings real positions and wash-sale state into the engine; the AI Designer operates against actual holdings rather than sandbox approximations; the Tax Bank tracks realized losses year-round; and every proposed trade requires manual review and confirmation before execution.

The threshold math

What portfolio size and tax bracket make Guided's subscription cost-effective?

The break-even point depends on marginal tax rate and annual harvest yield — at approximately 1% annual harvest yield, Guided typically pays for itself at portfolios of around $100K–$220K depending on bracket, with higher-bracket investors reaching break-even at a lower portfolio size.

Guided is $49/mo monthly, $490/yr annual (two months free).

For tax-loss harvesting to be worth $490/yr, the engine has to bank at least $490/yr of after-tax savings. Translating that to required loss harvested:

Your marginal rate Losses needed to break even on $490/yr
22% (mid-bracket)$2,228 / yr
32% (high-bracket)$1,531 / yr
37% (top-bracket federal + ~10% state)$1,043 / yr

Translating to portfolio size: a typical TLH program harvests 1-3% of portfolio value per year as realized losses. So at the bottom end (1%), Guided breaks even at:

  • 22% bracket: ~$220K portfolio
  • 32% bracket: ~$150K portfolio
  • 37% + state: ~$100K portfolio

Below those numbers, a Vanguard S&P 500 ETF (VOO, 0.03% expense ratio) is cheaper and ~95% of the result. There is no shame in that — direct indexing is a leverage tool that needs leverage to work.

22% BRACKET ~$220K typically, at ~1% annual yield ~$2,228 / yr losses needed to break even on $490/yr sub (estimated; varies by year) 32% BRACKET ~$150K typically, at ~1% annual yield ~$1,531 / yr losses needed to break even on $490/yr sub (estimated; varies by year) 37%+ BRACKET ~$100K potentially, at ~1% annual yield ~$1,043 / yr losses needed to break even on $490/yr sub (estimated; varies by year)
Estimated break-even portfolio size across three federal tax brackets, assuming approximately 1% annual harvest yield. Higher marginal rates reduce the required portfolio size because each dollar of realized loss may produce more after-tax savings. All figures are estimates; actual break-even depends on the specific portfolio, market conditions, and annual harvest yield achieved.

Why direct indexing fails below the threshold

Why does direct indexing not produce meaningful tax savings below a certain portfolio size?

Direct indexing can only harvest losses from individual positions, and below a certain portfolio size each position is too small for a typical price drop to generate a dollar loss worth banking after friction — the math requires individual position sizes large enough that a typical move produces a meaningful realized loss.

The mechanic that makes direct indexing valuable is also why it doesn't work small. To harvest a loss on a position, the position has to drop enough that the loss is worth banking. With a single ETF, you can't harvest part of the holding — it's one position. With direct indexing, you can harvest the names that dropped while keeping the names that rose.

That asymmetry only matters when individual position sizes are large enough that a typical 5% drop produces a meaningful dollar loss. The math:

harvestable position = position size × typical drop %
    = $5,000 × 5% = $250 per harvest
    = $50,000 × 5% = $2,500 per harvest

At a $5K position, the harvest is too small to bother with after fees and friction. At $50K, it's a real number. So the direct sleeve needs positions averaging ~$5K minimum, which means a 25-stock sleeve needs ~$125K. With the typical 60/40 direct/beta split, that's ~$200K total — right around the Guided threshold.

VS SMALL POSITION ~$5,000 × 5% drop ~$250 harvest may not justify harvesting too small after friction (estimated; actual varies) LARGE POSITION ~$50,000 × 5% drop ~$2,500 harvest meaningful after-tax figure worth banking and tracking (estimated; actual varies)
The harvest arithmetic scales with individual position size. A smaller position may generate a harvest too small to be meaningful after friction; a larger position at the same percentage drop may produce a figure worth banking and tracking. This is one reason direct indexing can require a minimum account size — the direct sleeve needs enough capital per position that typical market moves may generate worthwhile individual harvests.

The sandbox-to-real bridge

How does sandbox performance translate to a real Guided account?

The engine's harvest logic is identical in Sandbox and Guided — the difference is that Guided uses real positions, real cost basis, and real wash-sale state — so sandbox loss estimates at a given portfolio size are a reasonable proxy for what Guided can be expected to harvest on a similar real account.

If you've been running the Sandbox at $250K-$1M, you've probably noticed:

  • Loss Watch finds 5-15 candidates per month on a $500K simulated account.
  • Realized losses accumulate at $4K-$15K/year on ordinary market noise.
  • At a 32% rate that's $1,300-$5,000/yr of after-tax savings.

If those numbers track in your sandbox, they'll track on your real account. That's when Guided pays itself back.

Flat editorial illustration representing the conceptual link between a portfolio sandbox simulation and a live brokerage account. A clean electric-blue abstract bridge arch spans the center of the composition, connecting a left platform with a subtle grid pattern suggesting simulation and a right platform with a solid grounded form representing real capital. An abstract gold value token travels across the bridge, suggesting that harvest-yield estimates from a sandbox simulation may translate to a live account when the same engine logic is applied to real positions. The mood is calm and confident — translation, continuity, readiness.
The engine's harvest logic is the same in Sandbox and Guided modes — the difference is live positions and real wash-sale state. Sandbox performance at a given portfolio size may be a reasonable proxy for what a real Guided account can be expected to harvest on a similar portfolio, though actual results depend on the specific holdings, entry prices, and market conditions at the time.

What's the path

What is the recommended sequence for moving from Sandbox to a live Guided account?

The typical path is to run Sandbox until the portfolio reaches the break-even threshold and the engine's judgment looks right, then upgrade to Guided in Settings → Billing, connect a broker, and approve the first several proposals manually before evaluating whether to stay on Guided or graduate to Autopilot.

  1. Stay in Sandbox until you trust the engine and your portfolio is at or above the threshold.
  2. Upgrade to Guided in Settings → Billing.
  3. Connect your broker (E*TRADE, Schwab, or one of ~10 SnapTrade-supported brokers) in Settings → Brokers.
  4. Apply your existing manifest design to the real account, or run the Designer against your real capital.
  5. Watch the first proposal. Approve it. Watch the second. Approve it. After about a month, you'll have your own data on whether the engine + the subscription is a good trade for your specific portfolio.

If at any point the answer is no, downgrade. Settings → Billing handles it; your data stays.

For further reading: what changes when moving from Guided to Autopilot, how flat subscriptions compare to percentage-of-assets fees at scale, and what direct indexing requires in terms of account size and complexity.