Verified local facts
Trace an indicative personal loan calculator output from captured interface state through input provenance and field definitions to the matching written quotation and schedule.
Evidence and limitations
The audit package includes source address, checked date, disclaimer, screenshots or transcript, input register, output register, quotation, schedule and authenticated reconciliation.
Questions to ask before signing
A plausible estimate can be mistaken for a binding result when defaults, dates, definitions, omitted components or version changes are invisible.
Decision checklist
Capture the interface, pair only semantically matching fields, investigate each lineage break and rely on current written personal loan documents for the obligation.
Evidence reviewed
informational: resolve verify a personal loan calculator result in Malawi for a Malawi reader without inventing price, access, availability or approval.
Decision checklist
article article 1 6 Malawi decision path: Capture calculator provenance before changing a field -> Classify the output by its own authority wording -> Reconstruct the complete input provenance
Capture calculator provenance before changing a field
Open the institution-controlled interface through an independently obtained official address. Record the full page location, access date, visible product context, disclaimer, interface state and whether any session or applicant information shaped the screen. Preserve a screenshot or transcript before entering data, then another after the output appears. The checked page in the research record labels amount, tenor, interest, estimated monthly repayment and total interest separately and states that calculator information, rates and fees are indicative rather than an offer or acceptance. That narrow observation defines the authority boundary; this guide imports none of its displayed values. If provenance cannot be reconstructed, the output begins as unverified regardless of how precise it looks.
Classify the output by its own authority wording
Copy the words the interface uses for calculator, estimate, illustration, repayment, interest and any limitation. Do not upgrade an estimate to a quotation or repayment schedule. Record whether the result identifies a proposal, applicant, validity period or controlling document; absence is not filled by assumption. A public interface can help a reader explore inputs while remaining separate from a current written offer. The official consumer source supplies disclosure and complaint-process context, but it does not decide the contractual status of a particular screen. Give the capture an authority class based only on the evidence present. That class controls later language: an indicative output may be checked for reproducibility, but it cannot establish approval, availability or the borrower's obligation.
Reconstruct the complete input provenance
Create an input register in the order fields appeared. For each, record the exact label, value entered by the reader, unit, date or period, selection source and whether the interface prefilled a default. Include untouched fields, switches and choices that may influence the result. Do not substitute the quotation amount after the fact or assume that an empty field was ignored. If the interface changes a value automatically, preserve the before and after states. The register is a factual replay script, not a recommended scenario. Private values remain in the reader's worksheet and are not published here. An output cannot be reproduced when a material input, unit or default is missing from the audit trail.
Map output labels without inventing a formula
List every displayed result exactly as named, including any stated payment, interest, total or warning field. Add the unit, time basis and nearby explanatory text. Do not infer the formula, balance basis, compounding convention, rounding method or treatment of charges from the number alone. Even a result that seems arithmetically familiar can arise from an unstated definition or default. Link only definitions explicitly visible in the interface or supplied in authenticated written material. Mark all others unknown. The purpose of the output map is to identify what needs comparison, not to decode proprietary logic. A missing definition blocks semantic verification even when repeating the same inputs produces the same screen.
Pair estimate fields with written evidence
Place the current written quotation and repayment schedule beside the output map. Pair fields only when their labels, balance or amount subject, time basis, date, term, payment frequency and component scope match. A calculator payment is not automatically the same object as a scheduled payment, and a displayed total is not automatically total borrower outflow. Cite both locators for every proposed pair. If the quotation uses a different phrase, request its definition rather than translating it by intuition. Preserve unpaired fields as lineage breaks. The goal is documentary equivalence, not numerical closeness: two values can resemble one another while referring to different objects, and different values can be explained only after their objects align.
Open a hypothesis register for every difference
For each unexplained difference, list bounded hypotheses such as changed page version, different input, hidden default, date mismatch, term boundary, payment timing, component scope or field definition. A hypothesis is not a claim about the calculator. Record the evidence that would confirm or reject it and test one documented variable at a time where the interface permits. Do not assert an omitted fee, rounding rule or compounding basis merely because it makes figures reconcile. If several variables changed, the test cannot attribute the effect. Keep a residual unexplained state when no evidence closes the lineage. This register prevents a convenient story from replacing reproducible verification.
Repeat only controlled and documented tests
If another run is useful, begin from a fresh captured state and enter the exact registered inputs. Change only one documented input for a sensitivity check, retain both captures and describe the observed output change without generalising a formula. Do not publish the commercial values, present the result as a current price or use a hypothetical borrower. A repeat that matches shows interface reproducibility for that date and state; it does not prove that the quotation or schedule uses identical logic. A repeat that differs requires a new version, session or default hypothesis. Stop when the interface cannot expose enough evidence to isolate the cause. Controlled repetition supports lineage; it does not manufacture contractual authority.
Request authenticated reconciliation for broken lineage
Send a written reconciliation request that identifies the capture date, input labels, output label, quotation version, schedule version and exact mismatch. Ask what definition connects the fields and which current written document controls the applicant's obligation. Do not ask the institution to validate an invented formula. Retain the full response and any replacement document. If ordinary response channels do not resolve the matter, the checked official page describes a provider-first, internal-appeal and Registrar complaint sequence; it supplies process context rather than a promised result or legal conclusion. Until the lineage is closed, keep the affected field unverified and rely on the signed agreement and supplied schedule for binding payment information.
Publish a version-bound audit verdict
Use reproducible when the captured interface can be replayed and every field needed for the question has a documented semantic bridge to current written evidence. Use partially reproducible when the interface repeats but one or more bridges remain unknown. Use not reproducible when inputs, defaults, version or definitions cannot be recovered. State the reasons and scope. Archive the captures, registers, quotation, schedule and replies. Define expiry triggers: page change, new disclaimer, different input set, revised quotation, replacement schedule or altered definition. The verdict does not endorse a calculator, rank a provider or promise a price. It records whether one estimate's lineage could be audited at one time.