Copy Trading in Switzerland: 12 Checks Before You Mirror a Portfolio

The decision in one minute

Do not decide whether to copy a trader from a profit percentage, a polished equity curve, a badge, or a long list of closed positions. First ask whether the record describes the same account, rules, costs, risk and execution conditions that you would actually receive. If a provider cannot answer those questions with dated, verifiable evidence, the record is not enough to support a decision. That is a statement about evidence quality, not a prediction that the strategy will lose money.

The most useful first screen is a twelve-check evidence card. It separates identity, account type, population, time window, return calculation, drawdown, exposure, execution, incentives, independent evidence, selection effects and legal/service boundaries. A missing answer is itself a result: mark it unknown and decide whether that uncertainty is acceptable before linking an account or granting permissions.

Copy trading is not one uniform product. Some services transmit another person’s orders automatically; others provide alerts or let the customer approve each trade. The FCA says that, in its UK regulatory perimeter, a copy arrangement can amount to portfolio or investment management when the customer has no clear manual input. That is a useful classification warning, not a Swiss legal conclusion. The FCA’s copy-trading guidance describes the UK perimeter and the importance of how the service actually works. Swiss customers should check the specific service, contracting entity, permissions and advice channel applicable to them rather than importing a UK answer.

This page supplies a reusable buyer-side method, not a trader ranking. It contains a synthetic comparison, an evidence card and a small local CSV completeness checker. No real account, provider, platform, performance record, follower result or Swiss legal status was tested. The examples are fictional and deliberately do not imply that a particular return or risk level is attractive.

If you want the group’s separate overview of how prop firms and CFD platforms present their models and client-protection controls, see this Swiss market risk checklist. That page is group-owned editorial content. Its subject is the provider and business-model side; the method here is for an individual reviewing a copy-trading record. Neither page is independent performance evidence.

Start with the object being copied

Before reviewing a statistic, write down exactly what you might connect to. Is it one person, one named strategy, a model portfolio, a signal feed, a managed mandate, or a broker’s catalogue entry? “Copy this trader” can hide different arrangements. The displayed leader may publish an idea, while a platform chooses when and how to route it. A follower may receive different prices, partial fills, rejected orders, currency conversion, financing charges, or risk limits. The record can be genuine and still fail to describe the follower’s experience.

Separate the actor who originates a trade from the entity that executes it, the entity that holds assets, and the entity that displays the performance page. Those roles may belong to different companies. Ask for the exact legal name in the customer contract, the jurisdiction, the regulator or registry claimed, and the document that supports that claim. A logo or a social profile is not a legal-entity check. The FCA page is helpful precisely because it ties the UK classification to the actual degree of customer control; it does not say every feature called “copy” is identical.

Record your own control path too. Can you approve each order? Can you set a maximum amount, stop copying, close positions, or revoke access? Does a pause stop new orders only, or also cancel queued orders? Can the leader alter risk settings or add instruments without your approval? Answers should be taken from current terms and product documentation, with the version date saved. If the only answer is an interface screenshot, ask for the controlling contract or help page.

This distinction changes what evidence you need. A signal feed that requires approval can be judged partly as an information service, while automatic order transmission raises execution and account-control questions. A model portfolio with periodic rebalancing has different timing and cost questions from a stream of intraday orders. Do not use a single score to collapse those differences. Write the arrangement in plain language before you compare records.

The twelve-check evidence card

Use the following card for every profile. “Pass” means the evidence exists and is clear enough for your stated use; it does not mean the strategy is good. “Fail” means a contradiction, a material omission, or a fact that conflicts with the promised process. “Unknown” means you could not verify it. Never silently convert unknown into pass.

For each row, write the evidence location, date obtained, exact wording, unresolved issue and consequence. A compact record is more valuable than a high-level rating because another person can reproduce the decision. The card also prevents one strong feature, such as a long displayed history, from compensating invisibly for a missing live-account statement or unknown execution method.

1. Identity: match the record to the account

Ask which account generated every displayed number. If one profile combines several accounts, request the mapping and the aggregation method. A strategy may migrate between brokers, use a master account, or show a model account while followers trade elsewhere. Those can be legitimate arrangements, but they need to be named. Capture the account or strategy identifier, currency, account type, broker/custodian, owner, and record period. Masked identifiers may be acceptable if a qualified independent reviewer can check the match.

Then compare the identifier on the performance source with the identifier in the service terms. A common weak point is a public profile name that is not bound to the underlying account. Renaming, copying a profile, or changing its owner can break the link between past evidence and future orders. Ask what happens to the record after a material strategy change: does the old history remain attached, is the record restarted, or are periods combined? The answer should be visible in documentation or an auditable export, not inferred from a continuous chart.

Do not ask for someone else’s passwords or personal account data. The aim is provenance, not access. A redacted statement, a platform export with a verification route, or an auditor’s scoped confirmation can establish more than screenshots while limiting sensitive disclosure. Record what you did not see. An unverifiable identity is a reason to withhold reliance, regardless of how persuasive the display looks.

2. Live versus simulated: label the operating conditions

Simulation can be useful for testing whether a rule behaves as designed. It can also omit queue priority, spread changes, market impact, liquidity constraints, rejected orders, funding, and the emotional or operational effects of real capital. Therefore, “backtested,” “paper,” “demo,” “verified,” and “live” are not interchangeable labels. Ask when the record was simulated, when it became live, whether the rules changed at that point, and whether the chart combines the periods.

Request a date-stamped transition note. If performance is displayed across a simulation-to-live boundary, split the timeline and assess each portion with its own assumptions. Do not treat a small live segment as evidence that a long simulated history will transfer. Conversely, do not treat simulation as useless; use it for what it can support, such as testing operational logic or understanding a stated rule, and keep the inference narrow.

The evidence card’s second check should be “pass” only if the type is explicit and each transition is dated. If the provider cannot tell you whether the selected record is simulated, mark “fail” for disclosure and “unknown” for live performance. That is more accurate than calling it a bad strategy or a good one.

3. Population: ask what is missing from the catalogue

If a platform shows its current leaders, you are looking at a selected set. Ask for the full population used to compute any “success rate,” including profiles that closed, stopped publishing, were renamed, became private, or failed eligibility rules. Ask how each profile enters and exits the denominator. A rate without a denominator and inclusion rule is not reproducible.

This is a selection-bias problem, not evidence of misconduct by itself. A directory may reasonably remove inactive accounts. The practical question is whether that removal is disclosed and whether historical claims still refer to the original population. Request snapshots of the catalogue over time or a methodology note that explains deletions and eligibility. If the platform only presents current survivors, do not infer how the original cohort performed.

Academic research on social networks and investor behavior offers context for why social exposure and market activity can be related, but it does not validate an individual profile or establish that copying causes a particular result. The NBER working paper News Diffusion in Social Networks and Stock Market Reactions studies information diffusion and market reactions, while Social Interactions and Lottery Stock Mania examines social interactions around lottery-like stocks. These are specific research settings, not tests of every copy-trading service. Use them to motivate careful questions about selection and social influence, not to claim that a particular leader or follower will behave in a known way.

4. Time window: look for rule changes and quiet gaps

Request exact start and end dates, not “three years of experience.” Identify days or months with no trading, pauses, strategy revisions, account migrations, or missing data. A quiet period may be benign, but the reason matters. If a strategy changed its instruments, sizing, stop rules, or trade frequency, split the record around the change. A longer chart is not automatically a more comparable chart when the underlying process is different.

Check whether the record includes open positions at both ends. If a displayed metric is based only on closed trades, compare it with an account equity series that includes unrealized gains and losses. Ask what time zone is used and how timestamps are rounded. These details may seem small; they determine whether a follower can reconcile an alert with the actual order history.

Your evaluation horizon should also be written down before looking at the leaderboard. If your intended holding period is short, a multi-year aggregate may not answer the operational question you care about. If you are examining a long-term service, a short promotional period may be inadequate. This is not a universal minimum-duration rule. It is a requirement to align the question and observation period.

5. Return method: define the number before comparing it

“Return” can mean cumulative account growth, time-weighted return, money-weighted return, trade-level percentage, or a platform-specific score. Each answers a different question. Ask for the formula and whether deposits and withdrawals are handled. Request the currency, start value, end value, treatment of cash, and exact dates. If there are external cash flows, a percentage without a method can be misleading even when every input is accurate.

Ask whether costs are included: subscription, management or performance fees, commissions, bid-ask spread, financing, conversion, custody, and any copying-specific charge. Some expenses may be account-specific and some may be charged separately. Request a worked example that starts with a follower’s actual statement fields. Do not compare a provider’s gross record with your expected net result as if they were the same metric.

The Swiss provider VZ’s page about its online broker describes features and fee arrangements for its own platform. It is useful as a concrete example of how a provider may present platform features and costs, but it is promotional material from the provider, not an independent performance study. See VZ’s online broker overview and verify current terms directly with the contracting provider. A fee page can support a fee question; it cannot corroborate a trader’s returns.

6. Drawdown: include open risk and the recovery path

Ask how maximum drawdown is calculated: equity or balance, intraday or end-of-day observations, peak-to-trough rules, and whether open positions are included. Two vendors can use the same label and report different values because their data frequency or account basis differs. Request a dated equity curve or statement series and the method description. Ask for both the largest loss from a peak and how long it took to return to that peak, if recovery occurred.

Closed-trade summaries can understate the risk visible in open positions. A strategy that holds losing positions for a long time may appear stable when only closed trades are shown. That does not prove abuse or future loss; it means the observation method may omit information material to your question. Compare the series with broker statements or an independent audit whose scope explicitly includes open positions and account equity.

Do not treat “low drawdown” as a complete safety description. It says nothing by itself about leverage, tail exposure, liquidity, hidden correlation, or the possibility that losses have not yet been realized. It also does not specify how a copy service would behave during a gap, suspended market, or rejected order. Record the figure, formula, frequency and exclusions together.

7. Exposure: count economic bets, not ticker symbols

Request the instrument list and an exposure summary over the same period as the performance record. Look for leverage, notional size, largest position, currency, overnight exposure, sector and correlated positions. Ten symbols can share one market driver; a diversified-looking list may still be concentrated. If the provider cannot give position-level detail, ask for a safe aggregate such as asset class, direction, gross exposure and concentration band.

Exposure should be viewed as a time series, not just a snapshot. The largest position may occur outside the displayed sample or at a time when the equity curve looks calm. Ask how sizing changes after losses or deposits. Check whether the follower’s account uses proportional sizing, fixed lots, risk-based scaling, or a cap. Each method behaves differently when account values or instrument contract sizes differ.

Do not assume that the account’s stated risk limit is a hard guarantee. Ask which layer enforces it: the strategy, the platform, the broker, or the customer’s own account settings. Ask whether a disconnected follower, delayed feed, or partial fill can exceed the intended limit. Save the answer as a workflow description; marketing terms such as “risk-managed” do not specify the control path.

8. Execution: a copied instruction is not your fill

Record the leader’s signal time, platform transmission time, follower order time, and execution price where available. Ask whether orders are copied at market, by limit, or through another rule; how partial fills are handled; what happens when minimum size cannot be met; and whether the system skips a trade if a price threshold is exceeded. A leader’s record may show a trade that a follower did not receive, or a fill at a different price.

Compare the execution mechanism with your account: broker, instrument availability, contract size, currency, trading hours, leverage settings, and any regional restriction. A provider may show an average execution statistic across accounts; ask for the distribution or at least the methodology and dates. One average may conceal follower accounts that repeatedly receive a different outcome.

Ask how the system handles a connection failure or rejected order. Is there a retry? Could it duplicate an order? Who receives the alert? Can the customer cancel a pending instruction? Test these questions in a demo if one exists, but mark the test as a workflow test, not performance evidence. A successful button press does not prove the strategy or the account record.

9. Incentives: identify who gets paid and when

List every way the leader, platform, broker, affiliate, or publisher may be compensated: subscription, referral, trading volume, revenue share, performance fee, ranking payment, or a combination. The presence of compensation does not prove a recommendation is wrong. It changes what you should disclose and what alternative explanations to consider. Ask whether the compensation depends on followers opening accounts, trading more, retaining the service, or producing a displayed performance metric.

Save the answer and compare it with the product’s default settings. A ranking that rewards engagement or revenue may not rank by follower experience. If a profile is labeled “popular,” ask what that means and whether the metric is paid, user-selected, performance-based, or a mixture. If compensation cannot be disclosed, mark it unknown and decide whether that conflicts with your own standard for transparency.

Do not assume that an affiliate disclosure resolves all incentive questions. It may identify a commercial relationship but not explain how rankings, leader selection, or order frequency are determined. Conversely, a referral relationship alone does not invalidate account evidence. Keep the issue specific: what incentive exists, where it applies, and how it might affect the record or service you are reviewing.

10. Independent evidence: test provenance, not badges

Ask who verified the record, what data were reviewed, which account identifiers were matched, whether open positions and costs were included, and what period the verification covers. A third-party badge is not enough unless the scope and date are available. A verified profile may mean only that identity was checked, not that every return statistic was audited.

Prefer evidence with a clear chain: a dated broker statement or platform export, a disclosed calculation method, and a reviewer whose role and scope are identifiable. Ask if the reviewer can confirm the record without receiving your credentials. A screenshot can be a useful lead but is weak as the final evidence because the source, date, completeness and editing history may be unknown.

When comparing two records, use the same test for both. Do not require one provider to provide an impossible level of disclosure while accepting a badge from another. If a provider cannot reveal account-level information for privacy reasons, ask whether a redacted statement, independent attestation or reproducible aggregate can answer the question. Record the residual uncertainty.

11. Presentation and selection: inspect how the screen is built

Take a dated capture of the search, leaderboard, filters and default sort you actually saw. Ask whether profiles are promoted, sponsored, sorted by activity, ranked by performance, or personalized. Ask whether paused or removed profiles remain in historic statistics. A platform interface is part of the evidence environment: defaults shape which profiles a user encounters, even when the underlying data are genuine.

Look for survivorship bias, look-ahead bias, cherry-picked periods, and ranking instability. If a strategy appears only after a period of strong performance, the screen may answer “what is visible now” rather than “how did all eligible strategies fare before selection?” Ask for the selection algorithm and historical catalogue snapshots. If these are not available, narrow your claim: you have reviewed a current profile, not a representative sample.

Compare the displayed profile against a direct export and the service’s methodology. If the chart starts at a favorable point, ask why that date was chosen. If badges change with activity, ask whether they indicate quality, recency, or engagement. The objective is not to accuse the platform; it is to understand which question the screen can and cannot answer.

12. Legal and service boundaries: read the contract for the actual service

Write down who provides each function: publishing signals, selecting leaders, routing orders, holding assets, customer support, fees and dispute handling. Read the contract for the customer’s jurisdiction and the entity name. Check the relevant regulator or official register for the exact legal entity and permission claimed. A search result, copied logo, app-store listing, or social post does not establish authorization.

The FCA guidance is specific to the UK perimeter and explains why service design and customer discretion matter. It should not be transplanted into Swiss law. For a Swiss customer, verify the contracting entity and applicable Swiss requirements through current official channels or a qualified Swiss professional. If the service includes discretionary management, investment advice, custody, execution, or only technical transmission, the legal analysis can differ. This article does not decide which category applies to any provider.

Keep legal status separate from performance evidence. A service can have a clear legal entity and still provide weak performance data; a performance record can be well documented and still leave contractual or regulatory questions. Do not let one check substitute for the other. If the provider’s answer changes between the website, contract and customer support, preserve the versions and ask a professional before relying on the service.

A worked fictional comparison

Consider two fictional profile cards. Profile North shows a long, smooth cumulative chart, a “verified” label, and a high recent ranking. It does not state whether the record is simulated or live, excludes closed profiles from its success-rate claim, omits open positions, and does not explain follower execution. Profile South shows a shorter record, labels the live start date, provides a methodology for returns and drawdown, lists costs and open-position treatment, and offers redacted statements with an independent review scope. It still has unknowns about ranking incentives and how a delayed order is handled.

The correct conclusion is not that South will perform better. The evidence card supports a narrower judgment: South answers more of the questions required to interpret its displayed record, while North’s presentation leaves material gaps. If your goal is to compare record quality, you can say that. If your goal is to predict future results, neither profile has established that. Do not turn a documentation comparison into a performance forecast.

Here is a synthetic scoring example. It is a teaching device, not research data, not a pass threshold, and not an investment recommendation.

The synthetic example deliberately avoids numbers for returns, drawdown and win rates. Numbers can make fictional cases feel like observed evidence. The point is to practice separating completeness from attractiveness. In your own worksheet, you can use pass/fail/unknown labels and a separate confidence note. Do not sum them into a single grade unless you explain why each item has that weight and how the user should interpret the result.

Reusable evidence card template

Copy this block into a dated note for each profile. Keep original provider wording in quotation marks and distinguish it from your own interpretation. Link each answer to a document, statement page, export, regulator register, or support response with its date. If you are comparing products, duplicate the same block rather than changing the questions between providers.

Profile / strategy ID: Provider legal entity and jurisdiction: Contract version and date: Account owner / account type / currency: Live period: Simulated period: Strategy changes and inactive periods: Population included; deleted or closed records: Return formula and external cash-flow method: Costs included / excluded: Drawdown definition and observation frequency: Open-position treatment: Instruments / leverage / concentration / overnight exposure: Signal-to-follower execution path: Partial fill / rejection / delay / retry handling: Leader, platform, broker and affiliate compensation: Independent reviewer, scope, date and matched identifier: Leaderboard defaults, filters and paid placement: Customer control / revoke / pause / queued orders: Evidence links and date captured: Unknowns that could change my decision: Next verification date or event: Decision boundary: what this evidence does not prove: 

The template is intentionally descriptive. It does not calculate expected return, risk tolerance, suitability, or an optimal allocation. Those require personal circumstances and professional judgment outside a public profile review. The template’s value is that it makes uncertainty visible before you authorize access or make a payment.

Runnable local CSV completeness checker

The following Python 3 program checks whether a CSV evidence card contains all twelve named fields and whether each row is marked pass, fail, or unknown. It does not verify the source, assess a trading strategy, calculate performance, or determine Swiss legal status. It checks completeness only. It reads a file you choose locally and prints missing fields; it does not upload data or contact a provider.

Save the code as audit_card.py. Create a CSV with two columns named check,status and one row per check. Use the exact keys shown below. The output is a checklist diagnostic, not a grade.

#!/usr/bin/env python3 """Check a copy-trading evidence card for completeness; no performance scoring.""" import csv import sys from pathlib import Path REQUIRED = { "identity_account": "Identity and account", "live_simulated": "Live or simulated", "population": "Complete population", "time_window": "Comparable time window", "return_costs": "Return method and costs", "drawdown": "Drawdown and recovery", "exposure": "Exposure and concentration", "execution": "Follower execution", "incentives": "Incentives", "independent_evidence": "Independent corroboration", "selection": "Selection and presentation", "legal_boundary": "Legal and service boundary", } ALLOWED = {"pass", "fail", "unknown"} def main(): if len(sys.argv) != 2: raise SystemExit("usage: python audit_card.py evidence.csv") path = Path(sys.argv[1]) with path.open("r", encoding="utf-8-sig", newline="") as handle: reader = csv.DictReader(handle) if not reader.fieldnames or not {"check", "status"}.issubset(reader.fieldnames): raise SystemExit("CSV must contain columns: check,status") rows = list(reader) found = {} errors = [] for number, row in enumerate(rows, start=2): key = (row.get("check") or "").strip() status = (row.get("status") or "").strip().lower() if key not in REQUIRED: errors.append(f"line {number}: unknown check key {key!r}") continue if key in found: errors.append(f"line {number}: duplicate check key {key!r}") continue if status not in ALLOWED: errors.append(f"line {number}: status must be pass, fail, or unknown") continue found[key] = status missing = [key for key in REQUIRED if key not in found] print(f"rows_valid={len(found)} required={len(REQUIRED)}") for key, label in REQUIRED.items(): print(f"{key},{found.get(key, 'missing')},{label}") for error in errors: print(f"ERROR: {error}") if errors or missing: raise SystemExit(2) print("complete: evidence fields are present; no strategy quality was inferred") if __name__ == "__main__": main() 

To run it, prepare a CSV and execute python audit_card.py evidence.csv. A complete file produces twelve normalized rows and an exit code of zero; a missing key, duplicate key, invalid status or unknown field produces a diagnostic and non-zero exit. The program intentionally treats unknown as a valid recorded state. It does not convert it to pass or fail. That distinction prevents a blank field from looking like evidence.

For a safe smoke test, create a temporary CSV with all twelve keys and set every status to unknown. The checker should report completeness, not a favorable profile. Remove one row and rerun it; the process should exit with code 2. This is a reproducible software test of the checklist parser only. It is not a test of provider data, market behavior or customer outcomes. Do not put account numbers, credentials, personal identifiers, or full statements in the sample CSV.

The program accepts only these fields by design. It will reject a typo rather than silently create a thirteenth category. If you adapt it, keep a version number and preserve the original input; otherwise an updated checklist can make two reviews appear comparable when they asked different questions. Store evidence links separately and make sure local access controls are appropriate for any private documents.

Method, date, sample and limits

This article was prepared on 12 November 2026. The reusable 12-check card and checker are editorial assets created for this article. The “sample size” for original empirical findings is zero: no trader accounts, platform records, follower statements, trade executions or customer outcomes were sampled or measured. The synthetic North/South profiles are fictional and are not observations. The CSV smoke test is a software-behavior illustration, not a financial test.

The source review used the live FCA copy-trading guidance for a UK regulatory framing; two NBER working-paper pages for research context about social networks and investor activity; a Swiss provider’s own platform page for an example of provider-described features and fees; and the group’s earlier B2B prop-firm/CFD overview for editorial continuity. Each source has a limited role. FCA material does not state Swiss law. NBER working papers do not establish outcomes for an individual copier. VZ is a provider describing its own service, not an independent auditor. The group’s article is not an independent source.

GSC was checked for the primary publication site before drafting. The query-page view for the preceding 28-day and 90-day windows returned no eligible query rows in positions 8–30 with at least two impressions for this topic. Page-only observations existed for unrelated pages, but they do not validate demand for copy trading. This is a report observation at the time of the check, not evidence that no one searches for the subject. It also provides no basis to predict ranking or traffic.

Search-result inspection and the source set informed the boundary of the article: general explanations of social trading, platform features and legal labels do not replace a user-side evidence reconciliation method. The original contribution is the operational card and completeness checker, not a new market statistic. No prevalence estimate is claimed for forum questions, reader concerns or provider behavior. Without a representative corpus and documented sampling method, a single discussion thread cannot establish how common a concern is.

The 12 checks are a structured review aid, not a validated psychometric instrument, a compliance standard, a FINMA checklist, or a minimum legal disclosure list. A “pass” records that the requested evidence was obtained and interpreted for the stated purpose; it does not certify truth or quality. A dishonest record can pass a completeness check. An honest provider can have a legitimate privacy reason to withhold information. In both cases, record the limits and decide whether the remaining uncertainty is acceptable for your use.

Failure modes that make a neat record misleading

Treating a badge as an audit. A badge may confirm identity, activity, or data connection rather than the calculation and completeness of performance. Ask for scope and date.

Comparing incompatible returns. A leader’s gross, time-weighted series and a follower’s net account result answer different questions. Align currency, cash flow, dates and costs before comparison.

Ignoring open risk. A closed-trade table may omit unsettled losses. Ask whether account equity includes open positions and how often it is observed.

Assuming proportional copying is exact. Minimum order sizes, account value, leverage limits, rejected orders and execution delay can change the follower’s position. Ask for the actual routing rules and an account-specific statement.

Reading the visible leaderboard as the full population. Current profiles may exclude discontinued or removed entries. Ask for inclusion rules and historical catalogue snapshots.

Confusing an operational test with an outcome test. A successful demo order proves only that a workflow worked in that test. It does not validate returns, suitability or legal classification.

Using a risk label without its mechanism. “Low risk,” “conservative” or “protected” may be marketing descriptions. Identify who enforces a limit, what event triggers it, and what happens during a gap or system failure.

Over-reading an academic result. Evidence about social interaction in one market or dataset does not predict a named trader’s future behavior. Keep the context and population attached to every cited finding.

Treating missing data as proof of misconduct. A missing statement lowers what you can verify. It does not establish why it is missing. Use “unknown” until you have evidence for a stronger conclusion.

Letting a score hide the boundary. A combined rating can compensate for a critical unknown with several minor passes. Keep the twelve rows visible and decide which gaps are disqualifying for your own use.

A practical review sequence

1. Save the product name, contracting entity, terms version, date, and the exact profile or strategy identifier.
2. Describe the control path in one sentence: who chooses, who approves, who routes, who executes, and who holds assets.
3. Request the twelve evidence items before comparing a headline return. Mark each pass, fail, or unknown and link the supporting document.
4. Reconcile account identity, record dates, cash-flow treatment, costs and open positions. Split live and simulated periods.
5. Review leverage, concentration, order delay, partial fills and the ability to pause or revoke access.
6. Inspect compensation, leaderboard defaults, deleted profiles and any provider-selected sample.
7. Verify the legal entity and the service’s claimed permissions through current official channels. Escalate Swiss legal questions to a qualified local professional.
8. Run the local CSV checker if you want to confirm that all card fields are recorded. Keep its output separate from your judgment about evidence quality.
9. Write the decision boundary: what remains unknown, what new fact would change your view, and when you will review the file again.

The sequence is designed to prevent an attractive chart from setting the questions. Start with the record and the service; then decide which claims it can support. If the service refuses a question, capture the refusal and move on. Repeatedly asking support the same thing is unlikely to improve your evidence. A missing answer may be acceptable for a low-stakes research exercise but unacceptable before granting live account permissions. The threshold depends on your use, the contract, and your ability to reverse the action.

Decision map

flowchart TD A[Identify service and contracting entity] --> B{Live or simulated record clearly labeled?} B -->|No| U[Mark unknown; request dated evidence] B -->|Yes| C{Account, period, costs and open positions reconcilable?} C -->|No| U C -->|Yes| D{Execution, risk limits and incentives described?} D -->|No| U D -->|Yes| E{Legal/service boundary verified for your jurisdiction?} E -->|No| L[Pause; verify with official register or qualified adviser] E -->|Yes| F[Record remaining uncertainties and decide whether to proceed] U --> G{Does the missing evidence matter to your intended use?} G -->|Yes| P[Do not grant access yet; request evidence or stop] G -->|No| F

This diagram is not an automated approval rule. “Yes” means the evidence exists, not that the service is safe or suitable. “No” identifies a question that remains open. If a service involves a legal or financial judgment you cannot resolve, stop and obtain advice from a qualified professional in the relevant jurisdiction.

Checklist to keep with the decision

Source：https://dev.to/isabelle_dubuis_d858453d7/copy-trading-in-switzerland-12-checks-before-you-mirror-a-portfolio-4d3b
My favorites