Make the next decision smaller than the purchase
A useful first decision is not whether a robot will make money. It is whether there is enough inspectable information to justify a defined test. Start with one question: can this particular result be reproduced with the supplied version and settings? Another useful question concerns a specific failure case, such as recovering state after a restart. Neither question requires a promise about future performance.
This is a proposed review workflow, not a recommendation for a named product. Make a separate record for each candidate and put its intended use at the top. A buyer evaluating an existing executable needs a different scope from a developer commissioning new source code. Do not let a polished sales description silently become the acceptance specification.
Give every claim an evidence address
Use three columns: claim, evidence location and unresolved question. An evidence location might be a dated report, a settings file or a particular log range. Avoid entries such as "looks convincing". Those record an impression rather than something another reviewer can inspect. Missing evidence should remain missing, not receive a guessed score.
Keep the product version, exact symbol, server label, timeframe, date interval and assumptions together. Preserve the settings associated with each run before changing anything. Add a one-line reason for every rerun. This small discipline lets the reviewer distinguish a reproduced result from a fresh experiment that happened to look better.
Understand what a Market demo actually permits
MQL5 Market rules distinguish the protected demo of an EA from a demo trading account. A Market EA demo cannot run on an online chart; its trading behavior is tested in the built-in Strategy Tester [1]. A broker demo account instead uses virtual funds [2]. A seller providing a Market demo has not thereby supplied a forward-demo observation.
Write an access boundary in the evidence pack: what can be run lawfully, what files are available, and what cannot be examined. Do not try to remove product protection. When source is unavailable, describe the planned work as behavioral testing rather than a complete source-code review. A limitation is useful information, not a blank space to fill with reassurance.
Use a reproducibility question before a leaderboard
MetaTrader documents a fixed-parameter historical test separately from optimization across parameter sets. Its forward option reserves a later part of the selected historical interval for checking chosen parameters [3]. Ask which process produced the advertised result and which settings were chosen before the comparison period was examined.
Consider an invented procurement case. Candidate A has a report and matching settings for one stated interval. Candidate B has three cropped screenshots without a version identifier. This does not prove A is profitable or B is dishonest. It means A offers a more concrete starting point for a reproducibility test. That is the limited conclusion the evidence supports.
Request a failure-case conversation
Choose a few operational questions rather than a long feature wish list. What should happen when data is unavailable? How is a duplicate action identified? Which controls restrict new exposure? What evidence records the outcome after an interruption? These are suggested questions for scoping an audit, not claims that a product has already passed them.
For every answer, ask for the input condition, expected behavior and recorded observation. A setting named "protection" is not the observation. A useful reply might identify the exact test that remains to be run. Do not share live trading passwords, recovery phrases or unrestricted desktop access just to obtain an initial review scope.
Keep the commercial conclusion honest
Mark the pack as ready for a specified test, missing evidence, or containing a documented inconsistency. Do not convert that label into a profitability certificate. The Market rules explicitly do not make product testing a guarantee of future profits [1]. A narrow review can answer a narrow technical question while leaving substantial trading risk unresolved.
BamaUp has a before-you-buy evidence questionnaire and a separate service page explaining an existing-EA audit. The free EA checklist helps organize the initial questions without purchasing a review. Beginners can start with the free trading lessons instead. For paid work, clarify the scope, available files, deliverables and price for the specific order before proceeding.
