How Barguvo analyses: what it knows, what it estimates and when it does not conclude
Barguvo does not treat all information as if it carried the same weight. It separates what is declared, what is observable, what the person provides, what is externally confirmed and what remains unknown. Then it explains what supports a recommendation, how much it depends on assumptions and which check could change it.
A photo does not prove authenticity on its own. A seller’s text is not the same as independent verification. An active listing is not a completed sale. And a missing piece of data is not filled in to make the answer look more complete.
Contents
What do you want to understand?
Jump to the part of the method that affects your question.
Four labels so evidence is not mixed up
These labels describe what backing a claim has. Where a piece of data comes from and how strong the evidence is are different things: a piece of data may come from the user and still need an independent check.
Current primary documentation from a platform, manufacturer or competent body, within its scope.
It serves for processes, limits, specifications or requirements; it does not prove the condition of a specific unit.Public experiences that reveal language, friction points or patterns worth checking.
It is not a statistic and does not prove causation.Editorial interpretation or recommendation derived from visible data, context and limits.
It must be possible to explain it and take it apart step by step.A lead with insufficient, contradictory or overly context-dependent backing.
A lead, not a conclusion. It never decides a material recommendation on its own.PROVENANCE
First we separate where each piece of data comes from
Aim: so that “it is written”, “it can be seen” and “it is verified” do not become synonyms.
Title, description, condition, accessories, history or any claim by the seller. It is kept as a declaration, not as an independent fact.
Features visible in photos, screenshots or documents provided. What is observable can support a hypothesis, but an image may also be incomplete, old or manipulated.
Conversation, known use, defects, measurements, costs or context that was not in the listing. Its provenance is recorded so it is not attributed to the listing.
A piece of data matches a primary source or a check appropriate to the case. Confirmation always has a scope: confirming a specification does not automatically authenticate the unit.
If a battery, repair, defect, accessory or condition cannot be supported, it remains pending. The absence of evidence does not become “all correct”.
An output must be able to answer two questions: “where does this data come from?” and “how much weight does it carry for this decision?”. If it cannot, the explanation is incomplete.
When the decision depends on visual evidence, the photos guide applies this separation to the cover photo, angles, defects, labels and AI editing without misrepresenting the product.
COMPARABILITY
A price is only useful if the reference product is truly comparable
Aim: to avoid a seemingly precise figure built from incompatible references.
What we normalise before comparing
- Model, variant, capacity, year or version.
- Condition, wear, defects and known repairs.
- Accessories, warranty, documentation and delivery conditions.
- Country, currency, date and marketplace when they change the context.
- Whether the data is an asking price, a reduced price or a verifiable sale.
What can degrade a sample
- Duplicates or relisted listings treated as if they were separate transactions.
- Outliers without explanation.
- Old references for a fast-moving market.
- Mixing new, refurbished and used units without adjusting the comparison.
- Treating active listings as completed sales.
When coverage is not enough to support a reproducible valuation, Barguvo can offer context and questions, but it should not present a supposed “market price” as fact.
When a specification, limit or procedure depends on a platform, manufacturer or body, the primary source describes that point. It does not replace the evidence about the used unit you are assessing.
EFFECTIVE COST
The listing price is not always the cost of the decision
Aim: to compare alternatives with the visible costs that actually change how much you pay or how much you get back.
The agreed or listed figure, clearly distinguished from an estimate.
Only when the applicable amount is visible or can be checked in the current flow.
Known or sufficiently supported costs. If they are only possible, they are shown as uncertainty.
They are not invented to complete the sum. It is flagged which data is needed before closing the comparison.
A RANGE, NOT FALSE PRECISION
A range narrows when the evidence improves, not because we want a neat figure
Aim: to reflect real variation and make visible the assumptions that move the estimate.
Wider range
Few comparables, mixed conditions, scattered dates, uncertain accessories, a pending cost or an unchecked material condition.
More defensible range
Well-normalised identity, recent and comparable references, important costs known and contradictions resolved.
Barguvo does not need to pretend there is a statistical formula that does not exist. The explanation must show what widens or narrows the range, which references support the context and which assumption remains open.
To turn comparables and condition into a listing decision, selling on Wallapop applies this logic to the starting price and to later changes to the listing.
CONFIDENCE
“High”, “medium” or “low” are not mathematical probabilities
Aim: to express the practical quality of a recommendation without manufacturing a percentage of certainty.
| Dimension | What improves confidence | What reduces it |
|---|---|---|
| Completeness | The data that changes the decision is present. | A variant, defect, cost or material condition is missing. |
| Consistency | Text, photos, conversation and references do not contradict each other. | There are incompatible versions or unexplained changes. |
| Comparability | The references are truly equivalent. | The sample mixes different conditions, versions or contexts. |
| Freshness | Dynamic rules and references are dated and current. | The decision depends on old fees, limits or processes. |
| Source | The claim uses the right source for what it sets out to prove. | An opinion or lead is presented as if it were primary data. |
| Independence | There is corroboration that does not depend on a single interested party. | The whole conclusion rests on a single statement. |
| Sensitivity | The remaining unknowns would barely move the decision. | A single check could change the price, risk or recommendation. |
Confidence describes how defensible the recommendation is with the available evidence. It does not mean “80 % probability that the listing is good”, nor is it a guarantee of the outcome.
In negotiation, a recommendation on a buyer’s maximum or a seller’s minimum must keep these limits. The guide to making an offer uses effective cost, comparables and conditions without turning them into a universal haggling percentage.
ABSTENTION
UNKNOWN and INCONCLUSIVE solve different problems
Aim: not to turn uncertainty into artificial certainty.
UNKNOWN
A specific piece of data needed to answer is missing. The right output is to identify the gap and, when useful, ask for the check that would resolve it.
Example:The exact storage capacity is not stated, and that variant materially changes the comparables.
INCONCLUSIVE
There is evidence, but it points in directions compatible with more than one explanation. A box is not necessarily missing: what is missing is the ability to resolve the hypothesis responsibly.
Example:A photograph looks odd, but it does not allow you to distinguish between editing, lighting or a different product.
Directional output
States such as Compatible or Suspicious guide the next check. They do not authenticate, condemn or certify a transaction.
Example:Several inconsistencies justify stopping and verifying before moving on.
An interesting lead can be kept as a hypothesis, but it must never become a title, metadata, an accusation or an irreversible recommendation just because it “sounds likely”.
LAYERED RISK
A transaction can be strong in one layer and weak in another
Aim: to avoid the shortcut of declaring “all safe” or “all fraud” on the basis of a single signal.
Product / authenticity
Identity, variant, consistency, visible provenance and signals that require verification.
Condition
Wear, defects, functioning, repairs, accessories and correspondence with what was declared.
Counterparty / account
Consistency of the conversation, claimed identity, session control and signs of impersonation.
Transaction
Payment, shipping, delivery, traceability, issue resolution and any move outside the expected flow.
The guide to scams and risk patterns goes deeper into these layers. To review a specific listing, the methodology refers to how to spot a fake listing.
Users repeatedly describing a friction point may justify a specific check, but it does not authorise labelling a behaviour as fraud. Before escalating a signal, Barguvo must consider plausible innocent explanations and look for a second piece of evidence when it is material.
If the question changes from “what pattern am I seeing?” to “what do I do before continuing?”, the methodology refers to how to avoid scams instead of duplicating preventive checks.
SOURCES AND CONTRADICTIONS
The evidence label does not replace date, scope or context
Aim: to use each source for the claim it can actually support.
| Type | Especially useful for | Must not become |
|---|---|---|
| Official | The issuer’s rules, processes, specifications or requirements within its remit. | Automatic proof of the condition, authenticity or behaviour of a specific unit. |
| Community signal | Discovering friction points, language, doubts and patterns worth investigating. | Frequency, causation or statistics without valid sampling. |
| Barguvo’s judgement | Translating evidence and limits into an understandable action. | An external rule disguised as fact, or an unexplainable recommendation. |
| Unverified | Keeping a hypothesis or an open question. | A conclusion, accusation, final price or structured data presented as truth. |
When two sources contradict each other
First we check whether they refer to the same country, date, version, account type, marketplace or flow. Incompatible rules are not averaged. If the disagreement persists and is material, it is kept as uncertainty or the strength of the output is lowered.
When information changes over time
A dynamic rule —a fee, limit, payment method or procedure— needs a higher freshness requirement than a stable principle, such as photographing real defects. The more irreversible a decision may be, the greater the need for date, context and source.
There is no word count that closes a guide. It is closed when coverage, evidence, clarity, accessibility and usefulness no longer show a material improvement worth adding. Making it longer to reach a figure is not part of the method.
AI AND AUTOMATION
AI can assist; it is not allowed to fill gaps as facts
Aim: to explain the role of automation without promising capabilities or reviews that do not exist.
It can help to
- Structure scattered information.
- Detect contradictions or open questions.
- Summarise comparables and differences.
- Suggest next checks and messages.
- Improve presentation when the real content is preserved.
It must not
- Invent a specification, defect, cost or sale.
- Turn a visual inference into authentication.
- Present confidence as probability without a model validated for that purpose.
- Hide contradictions to give a cleaner answer.
- Simulate an individual human review if it has not happened.
Photography rule
AI can improve presentation, but never alter the apparent condition, identity or real characteristics of the product. Defects, wear, labels, materials, geometry, real colour and existing components or accessories must be preserved. “Improve the photo without misrepresenting the product.”
WHO · HOW · WHY
The methodology must explain who answers for the judgement and how it is reviewed
Aim: for transparency to describe a real process, not to be decorative trust signalling.
Who
Barguvo is editorially responsible for this methodology and for the criteria it presents as its own. An external source supports its facts within its scope; it does not automatically take over responsibility for the recommendation.
How
An output is built by separating provenance, normalising data, contrasting comparables and sources, preserving contradictions, assessing sensitivity and choosing an output level that may include abstention.
Why
This page exists so that a person can assess a recommendation, challenge it, provide new evidence and understand which part could change. The purpose is not to appear certain, but to make the recommendation auditable.
It indicates when this methodology was reviewed, but it does not automatically make all dynamic data recent.
Fees, limits, promotions and flows must keep country, version and date when they affect the recommendation.
If a claim becomes outdated or wrong, it must be corrected. In production, the visible contact/correction mechanism will only be published when a real route exists.
Changing a date without substantively reviewing a claim does not count as new research.
AUDITABLE CASE
From an incomplete listing to an explainable recommendation
Illustrative example: the figures serve to teach the process; they are not a real valuation.
The seller declares “very good condition” and 256 GB. The photos show light scratches, but battery health is not shown.
420 € and 256 GB are declared. The scratches are observable. The battery stays `UNKNOWN`. An official specification can confirm which variants exist, not the specific unit.
References with another storage capacity, new units and duplicate listings are excluded. If only a few comparable references remain, the range stays wide.
The recommendation might be “price compatible with the available context, but medium/low confidence until the battery and effective cost are checked”.
What could change the output?
If a verifiable screenshot shows a heavily degraded battery, if a major repair comes to light, if the total cost changes because of shipping/protection or if it turns out the variant was not 256 GB, the recommendation must be recalculated. The methodology does not protect an old conclusion from new evidence.
Do you have a specific listing?
The methodology helps organise the evidence; the analysis must work with that listing’s real data.
LIMITS
What Barguvo does not guarantee or do for you
Aim: so that a support tool does not look like a certification, a professional valuation or an intermediary in the transaction.
A good output is not one that always decides. It is one that distinguishes when to decide, when to ask for proof, when to lower confidence and when to leave a hypothesis open.
FROM METHOD TO ACTION
Where this methodology applies
Aim: to use this page as a cross-cutting reference and let each guide go deeper into its specific purpose.
Do you want to apply the method to a specific task?
The guides carry these principles over to selling, negotiation, risk, photos and cost.
FAQ
Does a piece of data appearing in the listing mean Barguvo considers it verified?
No. It means it is declared in that context. It may be relevant, but it is distinguished from what is observable, what the person provides and what is confirmed through an appropriate source or check.
Does Barguvo use active listings as if they were completed sales?
It should not. An active listing shows a price expectation; it does not prove that anyone has paid that amount. The type of reference must be made explicit.
Why does it give ranges instead of an exact price?
Because condition, variant, date, demand, costs and the quality of the comparables introduce real variation. An explainable range can be more faithful than a figure with false precision.
What does high confidence mean?
That the recommendation has few material gaps, good consistency, comparable references and appropriate provenance. It does not mean a mathematical probability or a guarantee.
What is the difference between UNKNOWN and INCONCLUSIVE?
`UNKNOWN` indicates that a specific necessary piece of data is missing. `INCONCLUSIVE` indicates that evidence exists, but it does not allow the hypothesis to be resolved responsibly.
Does one suspicious signal turn the whole listing into fraud?
No. The analysis separates product, condition, counterparty and transaction. An inconsistency may justify a check or stopping a step without proving fraudulent intent on its own.
Does AI decide on its own what is true?
It should not. It can help structure, compare and detect gaps, but it has no methodological permission to fill in unknown data as facts or to hide contradictions.
Are opinions from Reddit or other communities treated as facts?
No. They can work as a community signal to discover problems or real language, but they do not become frequency, causation or a universal rule without adequate evidence.
What happens if a Wallapop rule or a fee changes?
Dynamic claims must keep their date, scope and source. If they change, the applicable conclusion is updated instead of protecting an old version.
Does Barguvo require a guide to have a minimum length?
Not as a public quality rule. A guide is closed on the basis of coverage, evidence, usefulness, clarity and the absence of pending material improvements; not by reaching a word count.