How Robo-Advisors Work: Fees, Rebalancing, Tax-Loss Harvesting

What happens behind the dashboard of an automated investing service: risk questionnaires, fund selection, automatic rebalancing, tax-loss harvesting rules and the full fee stack.

2026-11-13 · 10 min read

Key takeaways

  • A robo-advisor is a discretionary portfolio management service, not a trading app
  • The advisory fee is only one layer of the cost stack
  • Rebalancing and tax-loss harvesting are rule-based and jurisdiction-specific
  • Cash drag and limited customisation are the main structural limits

How robo-advisors work: a managed portfolio behind a simple app

A robo-advisor is a discretionary portfolio management service delivered through software. You answer an online questionnaire, the provider maps your answers to a risk level, and from then on its portfolio engine decides what to hold, when to put new money to work and how to correct the mix when markets move.

The wrapper is regulated, not experimental. The SEC's Division of Investment Management describes automated advisers as typically registered investment advisers that use algorithms to provide discretionary asset management, and reminds firms that they are subject to the requirements of the Investment Advisers Act of 1940, including the fiduciary duty owed to clients (SEC staff guidance, 2017). The SEC's investor bulletin adds that firms providing advisory services in the US typically register with the SEC or with one or more state securities authorities and must file Form ADV, the disclosure document worth reading before opening an account (SEC Investor Bulletin: Robo-Advisers).

In the EU the same activity sits inside MiFID II. ESMA's guidelines on suitability requirements apply to investment advice and to portfolio management under Article 25(2) of MiFID II and Articles 54 and 55 of the MiFID II Delegated Regulation; they define robo-advice as advice or portfolio management delivered in whole or in part through an automated or semi-automated client-facing tool, and expect firms to document, test and monitor the algorithms that drive it (ESMA guidelines, 2023). In the UK the FCA supervises automated online discretionary investment management under the same advice and suitability standards that apply to human advisers, and its multi-firm review of the sector found recurring weaknesses: fee disclosure that was unclear or misleading, suitability assessments that did not properly evaluate knowledge, objectives and capacity for loss, and services that relied on assumptions about clients instead of adequate fact-finding (FCA, automated investment services).

The questionnaire decides more of the outcome than investors expect

Most risk questionnaires come down to four inputs: time horizon, tolerance for loss, capacity to absorb loss (how soon the money is needed) and the goal. ESMA's guidelines state that firms must collect all the necessary information about a client's knowledge and experience, financial situation and investment objectives before providing the service, and that questionnaire design matters more in robo-advice because human interaction is limited: firms should flag internally inconsistent answers, consider making human contact available during the process, and take steps against clients overestimating their own knowledge (ESMA guidelines).

Those answers are scored into a risk level, and each risk level maps to a fixed model portfolio, with a glide path that shifts the mix toward bonds as the goal approaches. That mapping is roughly where individualisation ends. Two investors with different incomes, existing holdings, tax positions and pension arrangements can end up with identical portfolios, because the scoring engine is not looking at any of those things. The upside is consistency and low cost; the limit is that the output is a template, not a financial plan.

Portfolio construction: which funds get picked, and why

Behind each risk level sit asset classes, and behind each asset class a small number of funds. Build lists differ by provider and by portfolio option: a service may run a purely index portfolio and a separate active-and-index mix, each priced differently (Vanguard advice comparison). A typical list includes a global developed-market equity fund, an emerging-market sleeve, government and corporate bond funds, sometimes inflation-linked bonds, and occasionally listed real assets. The selection screen normally covers:

  • Cost — the fund's ongoing charge, usually the most predictable drag on long-run returns.
  • Tracking quality — how closely the fund follows its index after all costs, which is not the same thing as tracking error (details here).
  • Size and liquidity, including behaviour in stressed markets.
  • Fund domicile and withholding-tax treatment for the investor's country of residence.
  • Dividend policy — accumulating versus distributing share classes change both cash flow and how income is taxed (explained here).
  • Securities lending policy and the counterparty exposure it creates (how lending works).

Two consequences follow. First, the equity/bond split does most of the work in setting both expected return and the depth of drawdowns, so the questionnaire — not the fund list — is the decision that matters most. Second, funds are deliberately chosen to be cheap and broad, which is precisely why customisation beyond the risk level is thin. There is no single published weighting that applies across providers; the allocation lives in the model portfolio shown inside your account, and the provider can change it without asking you.

Rebalancing: how the mix gets pulled back to plan

Markets drift. A balanced portfolio can become equity-heavy after a strong run, which means the investor is now carrying more risk than the plan specifies. Rebalancing sells what has grown and buys what has lagged; it is a risk-control exercise, not a return-enhancement trick.

Automated services use two main triggers:

  • Calendar rules — the portfolio is checked and corrected on a fixed schedule.
  • Threshold bands — a trade is triggered only when an asset class moves a set distance from its target weight.

Good implementations use incoming cash first: dividends and new deposits are directed at the underweight sleeve instead of forcing sales, which lowers trading costs and, in taxable accounts, realised gains. The trade-off is unavoidable: tight bands keep risk closer to plan but trade more often, and every trade carries a spread cost — plus, outside tax-sheltered accounts, potentially a tax cost (trade-offs and thresholds). When you compare providers, ask which trigger they use and what they do with cash arriving between rebalances.

Tax-loss harvesting, by jurisdiction

Harvesting means selling a position that has fallen, booking the loss for tax purposes and replacing it with a similar but not identical holding so market exposure stays intact. It has value only where losses are deductible and the money sits in a taxable account.

In the US

Losses on taxable holdings can offset capital gains and, within limits, ordinary income. The binding constraint is the wash-sale rule: buying a substantially identical security within 30 days before or after the sale disallows the loss, and the IRS states the window also captures substantially identical securities acquired for an IRA or Roth IRA (IRS, wash sales). Two costs matter. The replacement fund tracks its index differently from the original, so the portfolio slowly accumulates drift; and the benefit is a deferral rather than a gift, because the replacement position starts with a lower cost basis and the gain returns later.

In the UK: a matching rule, not a free hand

HMRC's Capital Gains Manual sets the order in which disposals of shares of the same class in the same company are matched: acquisitions on the same day first, then acquisitions in the 30 days following the disposal — the "bed and breakfast" rule — then the Section 104 holding, then later acquisitions (HMRC CG51555). HMRC notes that where the 30-day rule applies it will normally reduce or eliminate the gain or loss that would otherwise arise (HMRC CG51560). The practical effect is that selling shares at a loss and buying the same shares back inside 30 days does not deliver a usable loss: a UK investor harvesting a share position needs a genuinely different holding. Note that HMRC describes the rule in terms of shares of the same class in the same company, so how it bites on units in funds or on other instruments depends on the matching rules for those holdings — check before assuming. There is also a wrappers point: capital gains on investments held in an ISA are outside capital gains tax altogether (GOV.UK, ISAs), so harvesting is largely a taxable-account exercise in the UK.

In the EU: harmonised cost disclosure, national tax rules

EU-level rules harmonise disclosure, not tax. The PRIIPs regulation requires a standardised key information document that presents costs in a comparable format (Regulation (EU) No 1286/2014), which makes the visible fee layer easier to compare across providers. Loss treatment stays national. In Germany, disposals of securities held in a custody account are matched in the order of acquisition rather than by investor choice, losses on shares may only be offset against gains on shares, and each person has a saver's allowance of €1,000 before capital income is taxed (§ 20 EStG). The consequence is that mechanical harvesting on single shares has limited value in Germany, while losses on other capital holdings fit inside a wider capital-income bucket and can be carried forward within it. Because rules differ by country, providers marketing across the EU frequently restrict automated harvesting to selected markets or leave it out entirely.

The full fee stack

Published pricing looks simple until the layers are added. The figures below are the providers' own published rates, excluding the funds' own charges; they appear as examples of different fee structures, not as recommendations.

Provider Published advisory fee Source
Betterment Digital (US) 0.25% a year, or $5 a month below $24,000 in household balances; Premium 0.65% on the first $1m, then 0.15% and 0.10% tiers Betterment fee disclosure
Vanguard Digital Advisor (US) 0.20% gross for all-index portfolios, 0.25% for active/index mixes Vanguard advice comparison
Vanguard Personal Advisor (US) 0.35% gross all-index, 0.40% for an active/index mix Vanguard advice comparison
Schwab Intelligent Portfolios (US) No advisory fee; monetised through required cash allocations SEC order, 2022
J.P. Morgan Personal Investing / Nutmeg (UK) 0.75% a year on the first £100,000, 0.35% above; Fixed Allocation 0.45% and 0.25% Schedule of fees

On top of any advisory fee sit the funds' expense ratios. Vanguard states explicitly that its advisory fee does not include investment expense ratios charged by a fund, such as fees paid to third-party managers (Vanguard advice comparison), and the same split — advisory fee above, product costs below — is standard across the industry. Below that sit trading spreads, currency conversion if the portfolio is not held in your spending currency (FX cost mechanics), possibly platform and custody charges, and exit or transfer fees.

Because the layers are priced separately, only your own numbers produce your own total: add the advisory fee, the weighted expense ratios of the funds you actually hold (they appear in each fund's key information document or factsheet, and the difference matters — see TER versus total cost of ownership), and an allowance for trading and currency costs. Not every service charges a percentage of assets at all: the SEC's bulletin on subscription-based advisory fees describes robo-advisers charging a set recurring amount of $3, $5 or $10 per month for small accounts, a model that can be cheaper than an asset-based fee at low balances and more expensive as the balance grows (SEC Investor Bulletin: subscription-based advisory fees).

Cash drag: how a "no advisory fee" service still earns money

Schwab Intelligent Portfolios charged no advisory fee, but its model portfolios held between 6% and 29.4% in cash, and the SEC found the cash levels were pre-set so that an affiliate bank would earn a minimum revenue spread, even though the firm's own analysis showed the allocation would reduce client returns under most market conditions (SEC order). Schwab's investment adviser subsidiaries paid $187 million to settle, including a $135 million civil penalty, without admitting or denying the findings (SEC press release). The lesson generalises: when a service looks free, the revenue model is somewhere else — cash, order routing or fund fees (how cash sweep economics work).

Robo-advisor, DIY brokerage or human adviser?

  • A robo-advisor fits simple, long-horizon goals, smaller balances and automatic monthly contributions, where the alternative is doing nothing at all. Its cost is usually a fraction of a human adviser's and it rarely needs attention.
  • A DIY brokerage account is the cheapest route if you will actually follow rules, choose accumulating or distributing share classes deliberately and rebalance yourself; if you won't, the predicted saving often fails to materialise.
  • A human adviser costs more and earns that fee mainly in complexity: cross-border tax positions, employer share schemes, pension consolidation, drawdown and estate planning. Automation does not cover these.

Risks and costs

  • Market risk. No automation removes drawdown risk; a risk-appropriate allocation only sets its size.
  • Limited customisation. Standard portfolios ignore existing holdings, employer stock concentration and multi-country tax exposure.
  • Cash drag. Check the cash weight in the model and why it is there — it is a drag on return with no reduction in investment risk.
  • Provider concentration and operational risk. Custody, platform outages and provider changes sit outside your control; compensation schemes typically cover firm failure, not investment losses.
  • Harvesting side effects. Tracking drift, deferred rather than erased tax, and disallowance where a 30-day rule is breached.
  • Cost opacity. The advisory fee is rarely the total; fund charges, spreads and FX sit underneath, and regulators have repeatedly found these disclosures unclear.

Bottom line

Robo-advisors do the boring parts well: they pick low-cost funds, hold a risk-appropriate mix and correct drift on a rule. What they do not do is look at your whole financial life, and the advertised fee is only the top of the cost stack. Compare providers on the total cost of ownership, the cash allocation, the rebalancing trigger and the jurisdiction-specific tax handling before deciding whether automation, a DIY account or a human adviser matches the job.

This article is for information and education only. It is not investment advice and does not take into account your personal situation. Read the full disclaimer.

Follow the shortlist

Building or reviewing an ETF portfolio? Our global ETF and stock research watchlist is tracked daily. Premium members get the full list, the reasoning behind every entry and updates when entries change.

See premium access

Information and education only, not investment advice.

Related