Shiftto

Why I Built Shiftto and What Real Use Taught Me: The Operator's Story

Published: October 13, 2026 / Last updated: November 2, 2026

Shiftto is a shift scheduling tool I built and run on my own while studying in an engineering laboratory at the University of Tokyo and working as a freelance web developer. Why I built it, what I changed after people actually used it, what got easier and what is still missing. I'll cover not only what went well but also what falls short.

Why I built it

The main reason is that existing tools didn't fit shifts for one-off events. Many scheduling services ask which days are free, and aren't made to ask for fine availability by the hour, like “free 10:00 to 12:00 but not in the afternoon.” Tools that do handle hours were either paid or required everyone to register, and for an event over in a few days the effort wasn't worth it.

Another reason is that I wanted a friend's group to actually be able to use it. I had a small group close by to try it with, so rather than stop at building it, I decided to see it through to real use.

I also took an optimization course at university and wanted to try applying what I learned to a real problem. Shift assignment is about placing people into slots under fixed conditions, so the ideas from the course seemed directly usable.

How it was done before and what was hard

Before, I asked for availability with a Google Form, put the answers into a spreadsheet and built the shifts by hand. For a small group, that works fine. But as the group grew, these three things got heavy.

  • Copying form answers into the table
  • Trial and error, swapping people around to find short slots
  • Checking the imbalance in shift counts and consecutive work afterward

None of them is a one-time job; you redo them every time someone's answer changes. Handing these three to a machine was the starting point.

The first group that used it

The first group to use it was a staff team for on-campus and community events. It had roughly 10 to 30 people, and the shifts rotated the staff's duties by time slot.

Since launch, it has been used in practice by many groups, well beyond people I know. What I write here is what actually happened in that use.

What I changed after people used it

Once people used it, several things came up that I hadn't noticed. Here are the ones I actually fixed.

  • They wanted people placed consecutively. At first the assignment focused on reducing imbalance in counts, and a person's shifts could end up scattered. Scattered shifts were described as hard to work with, so I raised the weight so shifts group into consecutive slots as much as possible.
  • They wanted to edit roles afterward. At first roles couldn't be changed once the event was created. Roles get added or need renaming partway through, so I made adding and changing them possible from the admin screen.
  • They wanted to restore an answer. After changing an answer, people sometimes wanted the previous state back. I made it keep the pre-edit state as history each time and made it possible to go back.
  • They wanted it to keep working for large events. For events with many participants and time slots, auto-assignment sometimes stopped partway. The cause was that the calculation took longer than the server's processing time limit. I changed the server plan to allow more processing time, and also made it so that if the server does stop, the calculation continues on the device you are using and the result is saved.

None of these occurred to me at the design stage. It reminded me that you only find out when people actually use it.

How it is used, in numbers

I counted how much it is used from the production data, looking at events that gathered a fair number of respondents. Names and participant details were not used in the count.

ItemRangeMedian / total
Respondents per eventAbout 10 to 30Median about 10
Number of time slots5 to the 30sMedian about 10
Number of roles1 to 6Median 3
Events with saved assignmentsAbout 70%Automatic and manual roughly half and half
Times a person edited their own answerIn a few events onlyUnder 10% per respondent

Of the assignments, those placed automatically and those placed by hand by the organizer were roughly half and half by count. The numbers are rounded so that individual events cannot be identified. The definitions and every item are in the article on usage data.

What got easier

Compared with the spreadsheet and Google Form approach, these are the three things that I feel got easier.

  • No more copying answers into a table. Answers collect as a table by slot.
  • Less trial and error assigning. It proposes a draft automatically, and I only fix the spots I care about by hand.
  • Checking imbalance is easier. There is less need to recount shifts and consecutive work myself.

I haven't measured how much time was saved well enough to give a number. It is only the impression of me and the people who helped.

Limits that remain

I'll be honest about what falls short too.

  • If too few people answer, slots stay empty. Auto-assignment arranges well within the collected availability and cannot create more.
  • It is not an optimal solution. I prioritized returning a result quickly even with many participants and slots, and did not take an approach that searches exactly for the optimum. Results can differ slightly from run to run.
  • There is no feature to close responses automatically at a set date and time. To stop accepting responses, the organizer presses “Finalize” on the admin screen.
  • There is no way to reissue an admin PIN yourself. If you forget the PIN, you need to contact us through the Contact page.

The first two remain by the nature of the approach. The last two are features that are simply missing.

When it fits

Within the scope of how it has actually been used, what I can say is that it suits one-off or short-term events of roughly 10 to 30 people. It saves the most effort for shifts where headcount differs by role and busyness changes through the day.

On the other hand, fixed-member shifts that continue every week, and management that includes attendance and payroll, are outside what this tool is for. A dedicated service is better suited to that.

A detailed explanation of how it works is in the article on automatic scheduling, and you can see what the resulting table looks like on the demo page. Comments and requests are welcome on the contact page.

Related articles

See all articles