Crypto marketplaces attract a funny species of review.
You know the one. Five stars, contractor is “an absolute legend”, project supposedly smashed every target, no date, no order number, no scope, no payment trail, no deliverable, and somewhere near the bottom there is a referral link wearing a fake moustache.
Then there is its evil twin: one star, “SCAMMER, STOLE EVERYTHING”, posted ten minutes after a deadline slipped, while an escrow dispute is still open and half the story lives in private messages nobody else can inspect.
I’ve seen enough fake portfolios, vanished deposits, suspicious hiring offers, copied deliverables, hostage-style milestone disputes and revenge feedback to distrust both performances.
A useful service review is neither applause nor prosecution. It is an auditable account of a real engagement.
Street rule: a strong review lets a stranger distinguish what happened, what you can prove happened, and what you personally think about what happened. Those are three different things.
That distinction matters beyond forum etiquette. In the UK, current Competition and Markets Authority guidance treats fake reviews, concealed incentivised reviews, and misleading review information as consumer protection problems, and says businesses publishing reviews need reasonable and proportionate processes to detect and address them. See the UK guidance on publishing consumer reviews.
The same basic principle shows up elsewhere. US Federal Trade Commission guidance says material connections behind endorsements should be disclosed and that platforms featuring reviews should have processes aimed at reflecting feedback from genuine customers. See the FTC guidance on endorsements and reviews.
This is not about turning forum members into lawyers. It is about making reviews hard to fake, hard to weaponise, and genuinely useful to the next buyer.
Start with an engagement that actually happened
Before I care whether you loved or hated a crypto service, I want to establish that there was something real to review.
That means an engagement reached an auditable end state: delivered, partly delivered, refunded, cancelled, abandoned after payment, or formally disputed. “I spoke to them on Telegram and got bad vibes” may be a useful warning discussion, but it is not the same thing as a service review.
The basic record should answer these questions:
| Field | What a useful review should say |
|---|---|
| Date | When the order or engagement began, and when it ended or entered dispute |
| Purchased scope | What was actually bought, preferably in the language of the proposal, listing, statement of work, or accepted message |
| Commercial relationship | What you paid, how you paid, or what other benefit changed hands |
| Expected result | The deliverable, deadline, performance target, service level, or other agreed outcome |
| Actual result | What was delivered, when, and in what usable condition |
| Support experience | Response times, attempts to resolve problems, refunds, revisions, escalation, or silence |
| Evidence | Records that independently support the important factual claims |
| Unresolved issues | What remains contested, unpaid, incomplete, or under investigation |
That commercial relationship field is not optional just because crypto people dislike talking about incentives.
If the reviewer received the service free, was refunded in exchange for reviewing it, holds an affiliate arrangement, received tokens, got a referral payment, received a discount, has an investment in the provider, works for the provider, or was otherwise rewarded, say so plainly.
The UK advertising rules provide a useful sanity check here. ASA guidance says marketers using testimonials in advertising should hold documentary evidence that those testimonials are genuine, and it specifically warns against omitting material commercial interests. Its examples also show that ordering records, payment information and communications can help establish authenticity. See the ASA guidance on testimonial evidence.
Spoiler: the disclosure that saves everybody a lot of detective work
A simple statement works:
“I purchased this service myself for approximately £X. I received no refund, referral payment, free service, tokens, commission, affiliate revenue, discount, or other compensation for posting this review.”
Or, where there was a benefit:
“The provider refunded 25% of my fee after the engagement. The refund was agreed before this review was posted and was not conditional on the rating.”
That is much more useful than “not sponsored”, which can hide half a dozen other commercial relationships.
A positive review with a disclosed incentive is evidence that can be weighed.
A glowing review with a hidden incentive has an entirely different credibility problem.
Build the review around checkable evidence
Crypto gives people a dangerous confidence in the phrase “it’s on-chain”.
Fine. Show me what is on-chain.
But understand what it proves.
A transaction record can help establish that an amount moved between particular addresses at a particular time. By itself, it generally does not establish why the payment was made, what service was promised, who controlled every relevant address, whether the work met the agreed specification, or what happened in off-chain communications.
That is why a serious review uses an evidence chain, not one magic screenshot.
For a straightforward freelance engagement, I would usually look for some combination of:
Scope evidence: the marketplace order, accepted proposal, written statement of work, or messages in which both sides clearly agreed what would be delivered.
Payment evidence: marketplace payment status, escrow funding and release records, invoice, receipt, or transaction identifier where appropriate.
Timeline evidence: dated order events, delivery messages, revision requests, dispute filings, refund records, and support tickets.
Work evidence: the actual deliverable, a redacted extract, repository history, file metadata where useful, or a cryptographic hash if publishing the work itself would breach confidentiality.
Communication evidence: the messages directly relevant to disputed promises, deadlines, revisions, refunds or acknowledgements.
This is where bad reviews often stop adding up.
The buyer says “they missed the deadline by three weeks”, but the evidence shows the buyer changed the specification after the original deadline.
The seller says “work completed successfully”, but there is no delivery event and the escrow was ultimately refunded.
The reviewer says “I paid 2 ETH”, but the transaction shown was sent to an exchange deposit address with no evidence connecting it to the provider.
The contractor says “client disappeared”, while the support log shows the client chasing an update every second day.
Any one of those inconsistencies is worth investigating.
None of them, standing alone, proves fraud.
Suspicion gets you to the evidence. Evidence gets you to the conclusion. Do not swap those around.
There is also no prize for doxxing yourself to make a review look convincing. Redact home addresses, telephone numbers, personal email addresses, identity documents, authentication tokens, private account information and unrelated conversations. Never publish wallet seed phrases, private keys or credentials. Where a full record contains sensitive material, publish the relevant portion and retain the original in case a moderator needs to verify it privately.
And do not crop evidence so aggressively that nobody can understand the chronology. A screenshot saying “fine, refund it” means very little without enough surrounding context to establish who was speaking, what was being refunded, and when.
Separate the facts from the verdict
This is probably the most important habit in the whole format.
Write the review in two layers.
First, establish factual propositions.
Then give your judgement.
For example:
Fact: “The order confirmation specified delivery by 14 March. The first complete files were sent on 19 March.”
Opinion: “For my launch schedule, I considered that delay unacceptable.”
That is much cleaner than:
Bad version: “These people are completely unreliable.”
Another example:
Fact: “I reported the defect on 3 June. I sent follow-ups on 5 June and 9 June. I received the first substantive response on 11 June.”
Opinion: “I found that support response time poor for a service marketed to professional users.”
Or:
Fact: “The portfolio example offered during the sales conversation substantially matched work published earlier under another developer’s account.”
Conclusion requiring caution: “I could not verify that the contractor created the portfolio example.”
That is a much more defensible statement than jumping directly to “this contractor steals work”.
Words such as scammer, thief, fraudster, criminal, stole, hacked, forged, fake company and similar accusations carry substantially more factual baggage than “I was dissatisfied”, “I could not verify this claim”, or “the evidence I received did not establish authorship”.
If you have evidence for a serious allegation, present the evidence.
If you do not, describe precisely what remains unverified.
For readers in England and Wales, there is a legal reason not to blur these categories. The Defamation Act 2013 includes provisions dealing separately with truth, honest opinion, publication on a matter of public interest, and the requirement of serious harm. The exact legal position depends on jurisdiction and circumstances, so a forum template is not a substitute for legal advice.
That does not mean sticking “in my opinion” before an accusation magically makes it safe.
Street translation: an adjective is not evidence.
“I think he is a scammer” is still a lousy substitute for:
“Payment was released on 8 January. The agreed deliverable was never provided. I requested either delivery or refund on 12, 18 and 25 January. The provider acknowledged the requests but neither delivered nor refunded as of 10 February.”
Now the reader can make up their own mind.
And that is exactly what a good review should allow.
Watch for promotion dressed as experience
A review can be technically true and still function primarily as an advertisement.
There are tells.
The reviewer talks more about features than about the actual engagement. Every criticism sounds like a sales objection that the provider conveniently overcame. The prose includes discount codes. The review repeats marketing claims that the reviewer could not personally test. The account has no history except praise for one service. The reviewer will not say what was purchased or paid. The screenshots resemble campaign assets. Everybody supposedly got “amazing ROI”, but nobody can define the baseline.
None of that is automatic proof of promotion.
It is enough to start asking questions.
My first verification would be boring on purpose:
When was the service purchased? What exactly was ordered? What did it cost? Was any consideration received for reviewing it? Can the reviewer show a redacted order record? Does the account pre-date the review? Was the referral arrangement active when the review was written?
That last category matters because commercial relationships can change how readers interpret supposedly independent praise. UK CMA guidance specifically identifies concealed incentivised reviews as problematic and advises publishers to have a clear policy covering incentives. The FTC likewise treats disclosure of material connections as an important part of endorsement transparency.
A referral link should never be allowed to quietly answer the question the disclosure section avoided.
For a forum, I would either prohibit referral links inside reviews entirely or place them in a separately labelled commercial area.
Keep the evidence review and the sales funnel in different rooms.
Give the other side a real right of reply
Now we get to the part everybody suddenly hates when they are convinced they are right.
A review system that lets buyers publish accusations without correction is easy to weaponise.
A review system that lets sellers erase criticism by complaining loudly is just as rotten.
So the answer is not “believe the buyer” or “believe the seller”.
It is a structured reply process.
When a material factual claim is disputed, the reviewer or moderator should identify the specific claim under challenge.
Not:
“Provider denies review.”
Use:
“Provider disputes the statement that no deliverable was sent before 19 March and says a working build was delivered on 13 March.”
Good. Now we know what evidence matters.
The provider should be invited to supply the minimum relevant evidence: delivery timestamp, marketplace event, repository commit, support ticket, transaction record, signed acceptance, refund event, or whatever would actually test the statement.
The provider should not need to publish confidential client material to the entire internet. A moderator can verify sensitive evidence privately and report a narrow conclusion such as:
“Moderator verified that a file was transmitted on 13 March. We did not independently assess whether it satisfied the contracted specification.”
That wording matters.
Transmission is not acceptance.
Payment is not proof of satisfaction.
A refund is not necessarily an admission of wrongdoing.
An unresolved chargeback is not a final ruling.
A marketplace suspension is not automatically proof that every allegation against the account is true.
And a five-year-old account is not incapable of lying.
Convenient stories come from both sides.
Spoiler: a correction is not the same thing as deleting embarrassment
Suppose a reviewer originally writes:
“No work was delivered.”
The provider later supplies evidence showing that partial files were delivered before the review.
The right correction is not to quietly delete the post.
Update it:
“Correction, 22 April: I originally stated that no work was delivered. Evidence subsequently provided shows that partial files were sent on 17 March. I dispute whether those files satisfied the agreed scope. The review has been amended accordingly.”
The historical mistake is acknowledged, the corrected fact is visible, and the remaining disagreement is accurately narrowed.
That is what an honest correction looks like.
A fair right of reply gives the criticised party a chance to contest evidence. It does not give them editorial control over the review.
I would also distinguish four statuses: verified, partly verified, disputed, and corrected.
“Disputed” should not mean “false”.
It should mean exactly what it says: a material factual issue has competing accounts and the available evidence does not yet settle it.
That little label can save a forum from pretending to know more than it knows.
Use a review format that forces the story to add up
Here is the template I would actually use.
Service reviewed:
Provider name and marketplace profile.Engagement period:
Start date and end date, cancellation date, refund date, or dispute date.What I purchased:
Specific scope, quantity, milestones, deliverables and relevant service level.Commercial disclosure:
Amount paid and payment method where safe to disclose. State any refund, discount, free service, referral payment, affiliate arrangement, token compensation, investment, employment relationship or other benefit.What was agreed:
Deadline, expected deliverable and any measurable result actually promised.What happened:
Chronological factual account. Keep conclusions out of this part where possible.Actual result:
What was delivered, partly delivered, refunded, abandoned or disputed.Support and resolution:
Dates of important support contacts, revision attempts, escrow actions, refund requests and responses.Evidence available:
Order record, invoice, escrow record, transaction identifier, correspondence, delivery record, repository history, screenshots, hashes or other corroboration. Redact sensitive information.What I could not verify:
Claims or explanations for which sufficient evidence is unavailable.Unresolved issues:
Matters still in dispute or awaiting payment, refund, marketplace action or additional evidence.My assessment:
Clearly identified personal judgement based on the facts above.Right of reply:
Note whether the provider was invited to respond, what factual points they dispute, and whether supporting evidence was supplied.Corrections:
Date and describe any material amendment made after publication.
That format has one useful side effect: bluffers hate filling it in.
A fake reviewer has to invent a scope, date, payment path, chronology and deliverable that remain internally consistent.
A dishonest provider replying to it has to explain specific events rather than yelling “competitor attack”.
A genuine customer usually has some combination of records because real commercial engagements generate records.
Not perfect records.
Not always public records.
But something.
And the absence of a particular document should not automatically be treated as proof against either side. Cash jobs happen. Telegram deals happen. Marketplace records disappear. People lose accounts.
You weigh the entire evidence chain.
The research question worth testing
Here is the question I would actually like marketplaces and forum researchers to investigate:
Primary research question: among crypto service reviews labelled as coming from completed engagements, what proportion can an independent moderator verify by matching at least four elements, engagement date, purchased scope, payment or escrow evidence, and delivery or dispute outcome, and how does the rating distribution of those verified reviews differ from reviews that cannot meet that threshold?
That question gets past the usual “are fake reviews bad?” discussion.
Of course fake reviews are bad.
The interesting question is whether structured verification changes the picture of provider reputation.
My working hypothesis is that it would.
A marketplace might discover that some highly rated providers are being boosted by commercially connected or poorly evidenced praise. It might also discover that certain one-star reputations are being dragged down by reviews posted before disputes were resolved, reviews concerning services the poster never actually purchased, or claims that collapse when the contract scope is examined.
Either result would matter.
A useful study could record, without exposing private customer information, whether each review had evidence for date, scope, payment, delivery, support history and final outcome. Researchers could then compare verification rates against rating extremity, account age, reviewer activity, incentive disclosure, referral relationships, dispute status and later corrections.
The CMA’s current approach is relevant because its guidance explicitly treats review authenticity as something publishers should manage through risk assessment, detection, investigation and proportionate response rather than simply assuming whatever users post is genuine.
That is the bit I would steal for a crypto marketplace.
Not bureaucracy.
Just healthy suspicion with receipts.
A service review should leave the reader able to say:
I know what was bought. I know roughly when. I know what was promised. I can see what evidence supports the important events. I know which parts are the reviewer’s judgement. I know what the provider disputes. I know whether money, discounts or other benefits affected the relationship. I know what changed after publication.
Once you have that, five stars actually means something.
So does one star.
Without it, you may just be watching two anonymous wallets argue in a trench coat.
Granular Verification > Street rule: Separate facts from opinions in every label.