Your closed claims files, washed on your own computer.

Everything happens in this browser, before anything is sent. The engine is that of Proplace's partner page, byte for byte (version wash-1.0.0).

• Washing recognises neither a diagnosis, nor a treatment, nor a policy or file number: what is not a name you validate stays in the file. Remove those columns before uploading.
• The engine, Proplace's, knows the vocabulary of mergers and acquisitions, not yet that of claims: in General format, a “Claimed”, “Settled”, “Reimbursed”, “Medical expenses”, “Indemnity” or “Remaining cost” column is not recognised as an amount. This page therefore treats unclassified numbers as amounts by default. A whole number from 1900 to 2100 without a recognised label is read as a year and stays as it is: the page has you confirm them before washing, and a currency or thousands format given to their column in Excel makes them recognised as amounts.
• For a claims file, claimed and settled amounts are transformed, but caps and deductibles stay as they are: the remaining-cost calculation, which mixes amounts and caps, can no longer be redone on the washed version, and a remaining cost equal to the deductible, the usual case below the cap, gives k to within about 3% through their ratio. Treating caps and deductibles as amounts too does not restore that calculation (each amount is perturbed on its own), and a public cap, that of a known contract, then gives k to within about 3% to whoever recognises it.
• The washed version keeps each amount to within a few percent, no more: the ratio of two amounts differs from the original one by 6.3% at most, a result computed as a difference (claimed minus settled) by much more; whoever knows one original amount estimates the others to within 6.3% at most, and can then guess those that were very round. What the perturbation prevents: recovering k, and with it the exact amounts, from the round numbers of the washed version alone, as a plain scaling allowed. The ratios you declare are read on the original, before any transformation.
• One washed version per original file: two washes of the same file draw two factors k and two random values, and whoever held both versions could cross them to narrow their estimate of the amounts. The first version received is the one that counts.
• Name detection is a suggestion: every name spotted is ticked by default and replaced, except those you untick. Numbers written inside text and constants typed inside formulas are not transformed: the page lists the latter, review the preview for the former.
• A settlement case only enters Maximum's data (the real-cases page of /comparer) with its real amounts, validated by Maximum Insurance or by the insurer.
• Workbooks with macros (.xlsm), binary workbooks (.xlsb), password-protected files and images placed in cells are refused.