MomijiPerform
The merit cycle, without the HR department.
Goals and a two-question check-in all period. A review written from what is on file, sealed the moment the person acknowledges it, in English or in French. An approved increase waiting on the pay run itself.
An add-on to any plan, priced per person: $6 for Essentials, $12 for the Suite. No seat minimum, and no headcount at which it starts.
Review recordJanuary to June
Dana Whitfield
Site Supervisor, Fieldstone North
Sealed 4 JulyAcknowledged by Dana
How the record is made
Six months of evidence, then one conversation.
A review written the week it is due is written from memory, and everybody in the room knows it. MomijiPerform keeps the small things as they happen, in the same directory as everything else, so the person, the manager and the goals are already the same records.
Goals, and what they do not cover
A goal carries a success measure and, just as usefully, what it does not cover. Individual, team and company goals sit in one place, and a goal points at the one above it as a reference rather than a cascade: retiring a company objective is a status change, and the work underneath keeps both its history and its pointer. There is no percentage complete, because a number nobody can defend is worse than a sentence somebody wrote.
Two questions, once a quarter
What went well and what needs attention, answered in a minute, once a fiscal quarter. Two fields and not a form, because the form is the thing nobody fills in. A manager marks it read, which records that they read it, closes the period, and decides nothing. An update on a goal can be ticked as blocked, which is what puts it on the list of what is stuck.
One-on-ones both people can see
One list of talking points, the same one on both screens. An open point carries forward until somebody closes it, pointing at the conversation it continues rather than being copied into it. Momiji schedules nothing. A point is kept for the record only if somebody says so: most of a one-on-one is not evidence about anybody, and a tool that harvested all of it would teach people to stop writing things down.
Feedback with a name on it
Feedback and praise, with who may read it chosen at the moment it is written: the person and their manager, the person alone, or the whole team, which is praise only. Nothing here is anonymous and nothing is a rating. A manager can ask a colleague for feedback about somebody, and that request is pinned to the visibility that lets the manager read the answer, rather than leaving it to be chosen later.
The record, read back by quarter
Check-ins, goal updates, the one-on-one notes that were kept, feedback received, and training completed from MomijiComply, in one stream per person, newest first, grouped by quarter. An empty quarter is shown as an empty quarter, because the gap is the finding. Nothing in it is weighted, averaged or comparable across people: a number out of this stream is the one thing the design exists to prevent.
A review both people write
Twice a year, in five phases that only move forward. The person writes first, answering the cycle's questions in their own words, and a manager cannot read one until it is sent. The manager then writes from what is on file, citing it line by line, with the evidence on the same screen rather than a click away, and a panel that names what the review has not looked at: the goals nothing was written about, and the months it never read.
Then it has to hold up.
The scale, beside the choice
Four levels, each with the behaviour it means written out, and all four printed next to the control at the moment the level is chosen rather than in a policy document. The rating is the last field on the page on purpose: a level chosen first becomes the thing every paragraph underneath is written to justify. Four and not five, because a five point scale has a middle, and a middle is where a manager who does not want the conversation puts everybody. The four ship written and you can rewrite them, as a set.
Sealed by the person it is about
Acknowledging is the seal. There is no separate seal button and no administrator who can press it, and the person may put their own words on the record in the same act, kept beside the assessment exactly as they wrote them. Acknowledging records that they read it, not that they agreed with it. Afterwards nothing on the record can be changed by anybody, the administrator and us included: a database trigger refuses every later update and every delete on a sealed review, and the seal is a fingerprint that covers the rating and the scale that gave it meaning.
Both languages, and no machine in the middle
Every label, heading and legend renders in English or in Canadian French, chosen by whoever is reading rather than by the company, and the whole product is built for Quebec's French-language requirements. What the two people actually wrote is printed as they wrote it, with a line saying which language that was. Nothing is translated at print time. A manager can produce a rendition in the other language while writing and approve it there, and it is marked as a rendition on the record.
Momiji records what was written and when. It ranks nobody, averages no person, and makes no judgement about whether a standard is reasonable or an outcome is right. The module is built for the way Canadian employment law reads a written record and for Quebec's French-language requirements, which is a statement about how it is built and not a conclusion about your organization. Nothing here is legal advice.
- 4levels, each one written out
- 2questions in a check-in
- 5fields a plan needs to start
- 0sealed reviews that can be edited
Where the raise actually goes.
The one thing a best-of-breed performance tool cannot do, because the pay run is in somebody else's software.
One decision each
Every person, their salary, their last sealed rating and their tenure, on one roster with one effective date. An increase carries the reason it was given, and a cycle will not approve without one.
Staged, and inert
Approving writes an effective-dated staged change and moves no money. Nothing in the database reads it into a pay calculation, and the performance module writes nothing into payroll at all.
Waiting on the run
It appears beside that run's own variances, naming the cycle and the day it was approved, with what the salary was and what it becomes. A person takes it up, and Momiji refuses to record that unless the two figures already agree.
Two approvals, in two modules, by two acts, kept apart on purpose. A staged change is a record of a decision, not an instruction to the engine; the run pays whatever the employee record says on its pay date. Nobody is paid differently because a review was written.
The review folderA form filled in the week it is due, from memory.
MomijiPerformA record written over six months, cited line by line as the review is written.
The review folderA rating chosen first, and paragraphs written to justify it.
MomijiPerformThe level chosen last, against four written definitions printed beside it.
The review folderA document in a folder that can be reopened, re-dated or quietly improved.
MomijiPerformSealed by the person it is about, unchangeable afterwards, fingerprinted.
The review folderAn approved increase re-typed into payroll from an email.
MomijiPerformStaged, effective-dated, and waiting on the pay run's own screen with the cycle named.
$6 or $12 per person per month, on any plan.
No seat minimum, no contract, and no headcount at which it starts.
The things people ask first.
Does approving a raise pay it?
No, and it is built that way on purpose. Approving a compensation cycle writes a staged, effective-dated change and moves no money. The change then appears on the payroll prepare screen, beside that run's own variances, naming the cycle it came from and the day the cycle was approved.
A person still sets the salary on the work profile, and a payroll administrator still takes the change up on the run. Momiji refuses to record that a run carried a change unless the person's live salary already matches the figure that was approved, so taken up means the cash record and the decision agree rather than that somebody pressed a button.
Two approvals, in two modules, by two acts, kept apart deliberately. The performance module writes nothing into payroll at all: the migration that built it verifies that about its own code before it will install.
Does anything here score, rank or average a person?
No. Goals, check-ins, feedback and the evidence stream carry no scores at all. A review cycle carries exactly one rating, chosen by a named manager against written behavioural definitions, and never by a machine. A survey is averaged and is the only thing here that is: never a person, only a question, and only across five people or more.
There is no forced distribution and no quota. A compensation cycle can carry a budget as guidance, and going over it is allowed: the figure is shown so the decision is a deliberate one, and nothing is blocked. Nothing multiplies a rating by anything; there is no suggested increase and no formula from a rating to a number.
Nothing ranks a team against itself. Calibration draws a 9-box, and it is a grid for the conversation rather than a label: moving somebody on it writes nothing anywhere, and the dated snapshot a room can choose to keep is refused without a written reason. A fairness lens is available over a proposed set of increases, cut by department, tenure and rating, with small groups suppressed. It refuses nothing and flags nobody, and it says which cut it cannot make rather than letting a reader assume the axis was checked.
What does the AI actually do?
Four things, all of them inside the review a manager is writing, and each of them refused outright if the box is empty. It can read the period back as bullets that each cite something on file. It can tighten a paragraph you wrote. It can show you how your own wording reads to somebody else. It can carry one field into the other language.
It never drafts from nothing, never proposes a rating, and never touches a compensation figure. It cannot: the assistant has no path to write to a review at all, and there is no assistant anywhere on the compensation screen. A suggestion that comes back longer than what you wrote is refused rather than shown, and so is one that introduces a number your own words did not have.
Nothing is ever applied automatically. Every suggestion sits beside your own words until you press to take it, and the record notes which fields were tightened or translated, on the screen and on the printed record: “Parts of this review were tightened or translated with AI assistance. Every statement was written and approved by the people named.”
What does an improvement plan have to carry?
Five things written down: the standard being asked for, what will be measured, what the organization is providing, what happens if the standard is not met, and the day it will be reviewed. Momiji will not put a plan into effect without all five, and that refusal is enforced on the table itself, so no script and no administrator can produce an active plan missing one.
A draft is invisible to the person it is about, which is the only state in which that is true. Putting it into effect makes it visible to them and asks them to record receipt, which is receipt and not agreement; anything they write travels with the record whatever it says. Checkpoints are written up on the day they happened and can never be edited or removed, and a later one answers an earlier one.
A plan that runs past its own review date with nothing recorded is surfaced as exactly that, rather than being given a state of its own that would make the silence look deliberate. The plan and a written warning both print, in English or in French, with a signature block. That structure is built for the way Canadian employment law reads a written record; whether a given plan is sound is a judgement about your situation, and nothing here is legal advice.
What do calibration and surveys add?
Both are in the Suite. Calibration is the room where managers compare the ratings they have drafted, before anybody has been shown one, and it runs on written justification rather than on a shape: every card leads with what the manager wrote, the level sits below it as the conclusion, and under both is a count of the evidence the period actually holds, which is a count and never a score. Moving a rating costs a sentence, recorded against the person and against whoever moved it, in a log that only ever grows. Momiji refuses the change without one.
There is no target anywhere in it. The board counts how many people sit at each level and says, above the counts rather than under them, that the count describes what has been written and is never a shape to hit. It also refuses to move a rating the person has already read, and says so in place of the control: that is a conversation with them, not an edit behind them. The 9-box is part of the same conversation and stores nothing at all; a room that wants to keep the grid takes a dated snapshot, which Momiji will not store without a written reason saying what the two axes meant.
Surveys are confidential and never anonymous, and the difference is the design. Your answers are stored with your name so nobody can answer twice, and they are only ever shown in groups of five or more. Under five people overall, or under five answers to one question, or under five in a department, and that figure does not come back at all, and the next smallest group is withheld alongside it so nobody recovers it by subtraction. Written answers come back as bare strings, shuffled, with nothing to line them up by. Results are read by an administrator, never by a manager looking at their own team.
And there is a wall. Nothing anybody writes in a survey reaches their goals, their review, their rating or their pay. A survey answer is not evidence and cannot be cited in anything, which is a rule the database keeps rather than a policy we hold: no key joins an answer to a judgement about the person who wrote it, and the migration refuses to install if one is ever added.
How does French actually work here?
Every screen and every document label renders in English or in Canadian French, and the language follows the reader rather than the company. A review record and an improvement plan each offer both languages as a straight choice, so a francophone employee reads a French document whichever language it was written in.
What the two people wrote is a different matter and is treated as one. Momiji does not translate authored sentences at print time, ever. What was written is printed as written, with a line naming the language it was written in. A manager can produce a rendition in the other language while the review is open and approve it there, and the record says it is a rendition.
That is the honest half of being built for Quebec's French-language requirements: the chrome is offered in both languages in full, and a machine translation of a jeopardy sentence nobody approved is a different jeopardy sentence, so it is not produced.
What is not in it?
No forced ranking, no forced distribution, no curve and no quota, in calibration or anywhere else. There is no formula from a rating to a number, no suggested increase, no salary bands and no merit matrix.
Feedback comes only from inside your own organization and is never anonymous. Goals align by reference and never cascade, and one-on-one notes are shared rather than private.
The module sends nothing. No reminder emails, no nudges, no calendar invitations. It records the conversations you have; having them is still yours.
How is it sold?
As an add-on rather than a tier, so it is switched on for one organization at a time and it never depends on how many people you employ. Performance Essentials is $6 per person per month and Performance Suite is $12, on any plan, Budget upwards, with no seat minimum and no contract.
Essentials is goals, the check-in, one-on-ones, feedback, the per-person record, the review cycle end to end and the monthly people report. The Suite adds compensation cycles with the fairness lens and the payroll hand-off, improvement plans, calibration sessions, and surveys.
A person judges. Momiji keeps the record.
On any plan, per person, in English and in Canadian French.