The pattern is consistent enough to predict. Month one brings forty suggestions, several of them good. Month two brings twelve. Month four brings three, two of which are complaints about parking. By month six the box is a piece of furniture, and the conclusion drawn is that staff had nothing to say.
They had plenty to say. What stopped was not the supply of ideas but the belief that submitting one leads anywhere. That belief is created or destroyed by what happens in the two weeks after a submission, which is the part that never gets designed, because the launch gets all the attention.
Silence is what kills it
An employee suggestion box is a promise, whether or not anyone says so. The promise is that a submission will be read by someone with authority and answered. Every unanswered submission is a broken promise witnessed by at least one person, and people compare notes.
Three specific failures do the damage.
No reply at all. The most common. A suggestion goes in, nothing comes out, and the submitter concludes the box is decoration. One unanswered suggestion costs that person's future submissions. A dozen unanswered suggestions cost everyone's, because the absence of visible outcomes is itself the signal.
A reply with no decision in it. Thanking someone for their input and promising to consider it is worse than a week of silence followed by a real answer. It reads as a receipt from a process that has no next step, which is usually accurate.
Decisions taken invisibly. A suggestion is implemented, and nobody connects the change to the suggestion. The submitter sees the new kettle or the revised rota and does not know the box caused it, so the box gets no credit. Six months of quiet improvement can leave staff convinced that nothing was ever done.
The corollary is uncomfortable but useful. A suggestion box that answers ten submissions properly is more valuable than one that collects two hundred and answers none. Capacity to decide, not volume of ideas, is the constraint, and the form should be opened only as wide as the capacity behind it.
Anonymous, named, or both
This is the decision that gets made by default, usually at launch, usually in the wrong direction.
| Anonymous | Named | |
|---|---|---|
| What arrives | More on managers, workload, and safety. Less on process detail | More on how work actually gets done, since it can be discussed |
| Can it be answered? | Only in public, to everyone | Directly, and follow up questions are possible |
| Can it be investigated? | Rarely. Missing context cannot be filled in | Yes |
| Typical failure | Untraceable allegations nobody can act on | The uncomfortable topics never get raised |
The workable answer for most teams is named by default with anonymity available, and with the reason for that design stated plainly. A field that reads "Leave your name if a reply would be useful. Suggestions without a name are still read, but they cannot be answered directly" does more for the quality of submissions than any amount of encouragement, because it explains the trade being made.
Two cautions. If anonymity is offered, it has to be real, which means not collecting the submitter's email, account, or device details behind the scenes. People check, and one discovery that submissions were traceable ends the scheme permanently. And anonymity is the wrong channel for allegations about individual conduct. Those need a route with a named handler and a documented process, and a suggestion box quietly becoming that route is a serious problem for everybody involved.
Ask for four things, not ten
The form should take ninety seconds. A suggestion form with a business case section attached collects suggestions only from people who enjoy writing business cases.
- What should change. One open field, unrestricted length, with a label that asks for the change rather than the problem. "What should change, and where" collects something actionable more often than "Your suggestion".
- Where it applies. A short list of teams, sites, or shifts. This is the field that makes the log sortable and it costs one tap.
- What it would improve. Three or four options: time, cost, safety, quality of the work itself, or the experience of customers. Not a free text field. This is what lets a manager scan forty submissions and find the six worth doing first.
- Name and contact, optional. With the sentence explaining what naming buys.
That is the whole form. No effort estimate, no expected saving, no priority rating. Asking a warehouse worker to estimate the cost of a change is asking them to do a job they were not given, and the usual result is a blank form.
Two build details matter more than they look. Put the form where the work happens, not in an intranet section that requires a desktop login, because the people with the most useful suggestions are often the ones without a desk. A short link or a code on a poster in the break room reaches a different population than an email to all staff. And confirm on screen what happens next, with a specific interval, because the difference between "thank you" and "every suggestion gets an answer within two weeks" is the difference between a receipt and a commitment.
Build the decision loop before opening the box
This is the part that keeps it alive, and it is entirely mechanical.
One owner for the queue, by name. Someone whose job includes reading submissions weekly. Not a committee, which meets monthly and defers.
A timebox on the answer, published. Two weeks is realistic. One week is better and harder to keep. Whatever the number, it has to be met more often than not in the first two months, because that period sets the expectation permanently.
Four statuses, visible to the submitter. Received, being looked at, decided, done. Anything more granular is maintained for a month and then abandoned.
Decisions that include a no, with a reason. A suggestion box that only ever reports successes is not believed. A short, specific no keeps the channel credible and prevents the same suggestion arriving five more times. "Not this year, because the rota software cannot support it and replacing it is not planned until the contract ends" is a real answer. "Not feasible at this time" is not.
A visible record, updated monthly. A page or a printed sheet listing suggestions, decisions, and what shipped. This is what fixes the invisible decision problem, and it is the single highest return item on the list. Attribute changes to the suggestion that caused them, by name where the submitter agreed to it.
A route for the ones that are really projects. Some suggestions are good and cost three months of engineering time. Those need to leave the suggestion queue and enter whatever process handles work of that size, with the submitter told where it went. Leaving them in the queue marked as being looked at for a year is how a working box acquires a graveyard, and the graveyard is what people notice.
One number reported to whoever sponsored the scheme. The share of submissions with a published decision, not the count of submissions. Volume rewards a launch campaign. Decision rate rewards the thing that actually makes the box work.
Where suggestions arrive through a form rather than an address, most of this is bookkeeping the tool can carry: an owner and a status attached to each response is the whole mechanism, and the monthly record is a filtered view of it rather than a document somebody maintains by hand.
One more detail decides whether the loop survives its first busy month. Put the weekly review on a calendar as a recurring block with a name on it, and treat a missed week as something that gets noticed. Schemes of this kind almost never end with a decision to stop. They end with three consecutive weeks where the review did not happen, after which the backlog is large enough to be intimidating and the queue is never opened again.
Paper box, email address, or form
All three still get used, and they are not equivalent.
A physical box is the only option that reaches staff without accounts or phones, and the only one that feels genuinely anonymous to people who do not trust the systems. It is also a box that somebody has to empty, with handwriting nobody can read and no way to reply. It works as an additional channel for a specific population and badly as the only one.
A shared email address is the cheapest to set up and the easiest to let rot. Submissions arrive in whatever shape people type them, with no consistent fields, and a mailbox that four people watch is a mailbox where everything is read and nothing is owned. It also makes real anonymity impossible, since the address is part of the message.
A form gives consistent fields, a timestamp, a queue with owners and statuses, and the option of anonymity that actually holds. The cost is that it has to be reachable in two taps from wherever people are, which usually means a poster with a short link or a code rather than a menu item in an intranet. The same intake pattern already covers other kinds of internal request, and suggestions are easier to keep alive when they land in a queue that someone already works through.
Restarting one that has already gone quiet
Relaunching with a poster and an all staff email does not work twice. The first thing people remember is that the last one went nowhere.
The sequence that does work starts before any announcement. Go through whatever was submitted last time and answer the ones that were never answered, including the ones that are now too old to act on. An honest late reply, naming the delay, repairs more credibility than a relaunch does.
Then pick two suggestions from that backlog that can be implemented within a month, do them, and say publicly which suggestions caused them. Only after something visible has shipped is it worth announcing that the box is open again, and the announcement should lead with what was done rather than with an invitation to submit.
Two things are worth changing at the same time, since the relaunch is the only cheap opportunity to change them. Move the form to where the work happens if it was previously behind a desktop login, and shorten it if it asked for more than four things. A quiet box is sometimes a box that was only ever reachable by the third of the workforce with a company laptop.
Finally, cap it. Promising to answer everything within two weeks while carrying no capacity to decide is how the first attempt died. If one person has two hours a week for this, say so, and set the timebox to match.
What to change first
Answer the oldest unanswered suggestion this week, publicly and specifically, including the ones where the answer is no. Then publish a single page listing suggestions and their decisions, because the invisible outcome is the failure that does the most damage. Halict shows what a submission looks like with an owner and a status attached to it.
Q1. Should an employee suggestion box be anonymous?
Named by default with anonymity available works better than either extreme, provided the choice and its consequence are explained on the form. Anonymous submissions cannot be answered directly or investigated, so a scheme that is anonymous only tends to fill up with things nobody can act on.
Q2. How often should submissions be reviewed?
Weekly, by one named person, with a published deadline for answering. Monthly committees are the usual arrangement and the usual reason the box dies, because a suggestion submitted on the second of the month waits five weeks for a first response.
Q3. What should happen when a suggestion is rejected?
Say no, say why, and be specific about the constraint. A concrete reason keeps the channel credible and stops the same idea arriving repeatedly. Vague refusals such as not feasible at this time are read as no reason at all, and they cost more future submissions than an honest no.
Q4. How do you measure whether a suggestion box is working?
The share of submissions that received a published decision, and the median days from submission to that decision. Counting submissions measures the launch campaign. Counting answered submissions measures the only thing that keeps people submitting.
Q5. What if all the suggestions are complaints about pay or a manager?
That is a signal that no other channel exists for those topics, not a problem with the box. Pay and conduct need their own routes with named handlers and a documented process, and a suggestion box quietly absorbing them puts serious matters into a queue that was not designed to hold them.