Before you start
- Authorize access to Lark attendance, or provide an attendance export with documented fields.
- Agree on the payroll period, hours per workday, overtime source and leave categories. Department delivery also needs a recipient map.
Month-end attendance: from three days of exports and chasing to one scheduled run
Fiitx can summarize Lark attendance for your payroll period, use agreed sources for overtime and leave, and prepare employee summaries and department exceptions. A person reviews recipients and messages before delivery. Setup requires work-hour rules, the reporting period and a recipient map; missing fields are not zero.
Requirements and verifiable outputs
What you receive
- Employee summaries and department exception lists in your template, with field sources and missing values retained.
- A recipient and message preview for an authorized person to approve before sending.
Scope and limits
- A missing statistic is not zero. Overtime without an explicit source cannot be treated as approved overtime.
- The example verifies monthly field mapping only. It does not demonstrate live Lark retrieval, a full month-end close or message delivery.
Public verification example
Synthetic fixtures processed by Fiitx's existing calculation modules. These are not client records, delivery footage or measured time savings; only the rules listed below are verified.
Example verified:
| Employee ID | Overtime hours | Leave hours | Leave days | Source |
|---|---|---|---|---|
| DEMO-001 | 16 | 16 | 2 | Monthly summary |
| DEMO-002 | 0 | 0 | 0 | Monthly summary |
| DEMO-003 | Missing | Missing | Missing | Missing |
What to check
- DEMO-001 has 16 leave hours; at 8 hours per workday this becomes 2 days.
- DEMO-002 retains an explicit zero. Missing fields for DEMO-003 stay empty rather than becoming zero.
01
Where it used to get stuck
The exported number is not the usable number
Attendance systems count every minute beyond the schedule as overtime, while the company usually only pays the part that went through approval. Someone has to filter the difference row by row, leave a trace of it, and do it again next month.
A pay period that crosses months means two exports and a manual join
As soon as the period is not a calendar month — say the 26th to the 25th — the export has to be done twice and stitched together. Get the stitch wrong and every overtime and absence figure is wrong, in a way that is hard to spot.
Exception notices take the longest and are the easiest to get wrong
Late arrivals, missing punches and short hours have to reach the right department lead. Done by hand you either miss a department or send the whole company's list to every lead — and the second is worse than the first.
02
How it runs once handed over
Pull against your pay period, splitting and rejoining across months automatically
Period boundaries are configuration, not a hardcoded calendar month. A period that crosses months is fetched in two passes and rejoined per person, so nobody has to remember that this month needs two exports.
Overtime counted as approved only; leave mapped to your own types
Only approved hours land in the overtime column, and leave categories map to the names your company already uses rather than a tool's defaults.
A monthly summary in the shape of the sheet you already use
Column names and order follow your existing template, so whoever runs payroll can use it directly instead of reformatting it first.
Exceptions graded by rule, each lead receiving only their own department
The rules are configurable — how many consecutive late arrivals, how many missing punches, what hour threshold — and leads are resolved from the directory rather than pinned in config, so a reassignment is picked up next month.
It runs on a schedule
Daily reminders, the Monday roll-up and the fixed month-end run each have their own timing. Nobody has to watch the calendar.
03
Where it stops for a person
- The full list goes to HR for review before any department lead receives anything
- A changed definition applies from the next run — no version update, no reinstall
- Sending anything outward always waits for a person to approve it
How it is accepted
- Sample a handful of people and reconcile the output line by line against the raw system records
- Re-run a month that was already done by hand; the result should match that version
- The summary exports straight into payroll with no manual column fixes
04
Where the definitions come from
06
Questions
Where does our attendance data go?
Data retrieval and deterministic statistics run on the machine with the desktop app. External model calls may transmit prompts and data excerpts required for the task to the selected model service; the scope depends on the task and configuration. This is not a promise that all data stays offline. Enterprises can use their own model or gateway and review the data scope before connecting.
Our overtime rules are unusual. Can they be supported?
Rules are configuration, not code. The common shapes — approved-only, banded conversion, holiday multipliers — are parameters. A genuinely unusual formula is calibrated once at delivery and applied every month after that.
Can each lead receive only their own department?
That is the default. Distribution is split by department and leads come from the directory; the complete company list goes to HR only.
What if a definition changes mid-month?
It applies from the next run. No reinstall, no waiting for a release.