Transcribing into Excel can be automated — 3 criteria to tell

Anything that follows fixed rules and repeats is a candidate for automation. It’s not that “our work is too special” — what looks special is usually that the rules haven’t been written down yet.

① Is it automatable — 3 criteria to tell

Criterion	What it asks
Rules	Can you explain in words “if this value, then this”?
Repetition	Does it recur on a daily, weekly, or monthly basis?
Format	Is the format on the receiving end fixed — an Excel table, a report, an alert?

If all three hold, it is a candidate for automation.

Getting stuck on the first one is not a “no”. It means the criteria in the person’s head haven’t been written down yet — a rule a human cannot explain in words cannot be handed to a machine either. So the more demanding the request, the better for you. The more specific the criteria, the more there is to turn into rules.

② What do you build first — not code, but the criteria

1. Record what you do now for one week
2. Write the rules into a document — what to keep and what to drop
3. Automate only the parts that repeat
4. Add a screen where the results can be checked at a glance
5. Check whether the work done by hand has decreased

Step 4 is easy to skip. Without a place to look, you end up opening files every time to check that things are running — automation that creates a new checking job of its own.

③ How do you hand it over to the person doing the work

Package it as an executable so no runtime environment like Python needs to be installed. Double-click the icon and the results come out.

Include a button that opens the output folder directly. If the saved file has to be hunted for every time, you spend the time automation saved you right back.

④ What do you watch to keep it running long-term

The most dangerous failure is the one that gives no signal that something is wrong.

• Collection silently stops — even when it stops over an expired credential or a changed login method, the screen still looks fine. For as long as you don’t notice, you make decisions on empty data.
• Progress isn’t visible on screen — a long-running task is indistinguishable from a stuck one.

For collection that requires authentication, tie everything from login to token renewal into one unbroken flow, and show the time it last succeeded on screen. “Set up” is not “arriving”.

If you build it yourself, the first run usually works. The hard part is what comes next — keeping it from silently stopping months later. Building that into the design takes longer than building the thing itself. Ask us about the work you’re doing and we’ll start by pointing out where it could stall.

For the manual work to actually disappear

Sometimes automation is in place but the manual work stays. The cause is usually the same — only the collecting is automated, while the reviewing and the reporting are left as they were.

For the manual work to actually disappear, collect → review → report must be connected as one chain. When judging the effect, don’t look at “does the collection run” but “has the work done by hand decreased”.

If what you collect is an external site or a seller page on a mall or platform, check the terms of service before building. Some terms directly forbid exporting data from seller-facing systems — before asking whether it is legal, it becomes a contract violation. The standard for judging is laid out in Web crawling: how far does it go?.

---

You only need one task you repeat right now.

You don’t need to bring a week of records. Tell us in one line what you keep repeating, and we’ll first diagnose it with the 3 criteria above and lay out whether it can be automated and where to start.

→ Ask for an automation diagnosis · Work we’ve done

---

Cost by collection scale and difficulty — Data collection pricing