The hard part is not the arithmetic, it is **alignment**: the sheets name columns differently, the same word means different things in two departments, and last year's definition is not this year's. Done by hand, the gaps are filled from memory — and memory does not transfer.
Which channel a client belongs to, which month a rebate is spread across — these judgements are rarely written down. When the person who held them leaves, a discontinuity appears, and it usually surfaces a quarter later.
Exports arrive with stray spaces in headers, a word changed, the order shuffled. Reading by position does not error on any of that; it produces a plausible wrong number.
Many finance systems only offer import and export templates. That segment stays manual, and the whole line breaks in two at that point.
Client, channel, project and company follow your existing attribution rules, held as configuration rather than hardcoded in a script.
When a match genuinely fails it stops and asks rather than skipping the column and carrying on. This rule exists because the alternative was learned the hard way: skip a column silently and the report still looks tidy, while the wrong number surfaces months later.
How rebates spread, which accounts roll into which metric, how periods are cut: all of it lives in a definitions profile in the cloud, not in one person's workbook.
Where the ERP offers no API, files are generated against its own import template and a person approves them before they go in. A human still gatekeeps — they just stop filling rows by hand.
Every run stores its inputs, the version of the definitions used, and the output. Any period can be recomputed later and compared.
Yes. Where there is no interface, its own import and export templates are used: exports for reading, generated import files for writing, approved by a person first. That is not a workaround — it removes the most mechanical part of the manual work.
No. Definitions are configuration, maintained per organization in the cloud, and apply from the next run. Changes are versioned so you can see which period they started from.
Every run stores inputs, the definition version and the output, so any period can be re-run and compared.