The reporting workflow we automated in a week
A weekly report that used to eat half a day now builds itself. What made it a good candidate for automation, and what we left manual on purpose.
Automate the assembly, not the judgement
The weekly report that eats a morning is rarely eating it on analysis. It goes on exporting from three systems, pasting into a sheet, fixing the date formats, renaming the columns that changed, refreshing the charts and saving a PDF. All of that is assembly, all of it is identical every week, and none of it needs a person.
The judgement — what the numbers mean, what to flag, what to leave out — is the part worth a person's morning. Separating the two is most of the work in any automation job, and it is also the honest test of whether something should be automated at all.
The first week is spent on the data, not the script
Writing the script is fast. Discovering that one source exports dates as text, another pads its IDs with leading zeros, and a third occasionally emits a blank row that shifts every column one to the left — that is the week. A pipeline built before those are known is a pipeline that produces confident, wrong numbers.
So the cleaning and transformation step gets built first and gets built to complain. A row that does not match the expected shape stops the run and says which row and why, rather than quietly averaging itself into the total. A report that refuses to build is recoverable; a report that builds wrong is not, because nobody checks a number that looks plausible.
Hand over something your team can change
An automation nobody but its author can modify is a liability dressed as a saving. The first time the source system adds a column, it either gets fixed by the person who wrote it or it gets abandoned and the morning comes back.
We document every script, hand over the source, and walk your team through running and changing it. Whether it ends up as Python, an Excel and VBA build or a scheduled job depends on what your team can actually maintain — which is a question worth asking before the tool is chosen, not after.
Full-time teams in Chile and India, about nine hours apart. Chile covers your working day; India picks the file up as you sign off, so weekday work does not wait overnight.
Tier 3 is the top of our creative scale — and the only tier we run. The thinking behind this article is the same thinking behind every brief we take on.