Changelog

Everything below is in the product today. Nothing here is a roadmap item, a beta flag, or a plan. Where a feature has a real limit, the limit is stated with it. There is a short list of what is deliberately not built at the end, because knowing what a tool refuses to do is usually more useful than knowing what it claims.

13 July 2026

• Public reference library. The Salesforce import error hub went live, with the write-ups behind it — including bad value for restricted picklist field and error converting value to correct data type. Each one is written to be usable with no tool of ours: the fix comes first, the product comes last.

12 July 2026

• Map values. A new per-column transformation. It opens a modal listing the distinct values in your source column against a dropdown of the values the target field accepts, so you translate the whole set in one screen instead of discovering them three rejected loads at a time. The lookup is whole-value, trimmed, and case-insensitive.
• AI value-mapping suggestions. Optional. One click proposes a target value for each distinct source value (US to United States); you review, correct, and approve every row before it takes effect. Nothing is applied silently.
• Picklist values captured from your org. When a target field is generated from a Salesforce object, its accepted picklist values come with it. Preview now flags any cell a restricted picklist would reject, before a single record is written.
• Currency symbols accepted in Number fields. Source files say $1,200,000 and (£310,000). Those now parse into a Number field — one leading or trailing symbol, thousands separators, and parentheses for negatives. A misplaced symbol (1$2) still fails, on purpose.

11 July 2026

• Upsert by External ID. Direct upload gained an operation picker: insert, or upsert matched on an External ID field. The match field is read live from your org's describe and only fields you have actually mapped are offered. Jobs run through Bulk API 2.0. Update-by-record-Id is not offered, and never will be.
• Repeat-upload guard. The last successful payload is fingerprinted (SHA-256). An identical upload is refused, with an explicit "Upload anyway" override for the rare case where you meant it. A job whose rows were all rejected is not fingerprinted, so retrying a failed load is never blocked.
• Failed rows, grouped and explained. Per-record failures are fetched back from the Bulk job, grouped by identical message, and shown under "Error details". A second, optional click asks the AI to explain them in plain English.

10 July 2026

• Interface pass across the import workbench: clearer mapping panels and preview layout.

9 July 2026

• Multiple source files. Join up to five files — one primary plus four lookups — so the account owner in one export and the accounts in another can land in the same load. The join shows you its match rate before you rely on it.
• Searchable field pickers throughout the mapping editor, which stops being a nicety somewhere around the fortieth field on a customized Account.
• Durable file storage. Uploaded and exported files can be held in Azure Blob storage, so a multi-instance deployment does not lose a working file mid-session.
• Demo workspace. Sample data you can drive without connecting an org.

7 July 2026

• Composite mappings. Build one target field from several source columns, or from a constant, rather than one-column-to-one-field.
• Row selection with conditions. Keep all rows, only matching rows, the first row per group, one per group preferring a match, or one per group requiring a match — across 14 operators. Grouped rows can be flattened, so a mapping can read a sibling row's value.
• Template workflow: faster navigation between source file, mappings, and preview.

6 July 2026

• Direct upload to Salesforce. Bulk API 2.0 insert, behind a review screen and an explicit confirmation. It never deletes records. Workbook export stays available for anyone who does not want to connect an org at all.
• Accounts and sign-in, including Google, Microsoft, and Salesforce SSO.

5 July 2026

• Duplicate a template — mappings, transforms, row-selection rule, and its own copy of the stored source file, under a new name.
• Reverse import. Closes the round trip after a load. You export the new records from Salesforce yourself — INQUA has no read path out of your org — and hand that re-export back in as the new source. It then builds the mirrored template in one step: an output shaped like your original source file plus a record-Id column, with the mappings seeded by inverting the originals. What comes out is a workbook, not a write back to Salesforce. Transforms, constants, and row selection deliberately do not invert — an inverted transform would be a guess.

2 July 2026

• First working build. Import templates over .csv, .txt, .xlsx, and .xlsm files, read natively — no CSV conversion step, so leading zeros and long IDs survive. Field mapping, chainable per-column transformations, row selection, a typed preview with per-cell errors, and export to a typed .xlsx workbook written onto your own load template, with its header block and formatting preserved.

Not built, and not being built

Stated plainly so you do not adopt this expecting it:

What preview can catch is the class of error that lives in the file: wrong type, wrong date format, a value a restricted picklist does not hold. Upload to Salesforce is refused outright while any cell error remains.