Most employee evaluation forms produce the same result: everybody scores a 4, nobody learns anything, and the form goes in a folder until next year.
That’s not because reviews are useless. It’s because the form is built to be easy to fill in rather than hard to fudge. Ask a manager to rate “communication skills” out of five for a driver they see for twenty minutes a morning and you’ll get a 4, every time, for everyone.
A form that produces honest feedback does three things differently: it asks about observable work instead of personality, it makes the middle of the scale mean something specific, and it leaves more room for sentences than for numbers. This post is about that document: what goes on it, how to score it, and how to run the conversation it feeds. If people are leaving and you’re not sure why, the common causes of employee turnover at small businesses is the wider diagnosis; a review is one of the five fixes.
The Bottom Line
Written comments do the work. In an analysis of 11,000 reviews, 61% of the content that actually influenced change was written comments, yet 39% of reviews rely on scores rather than comments (Teamflect).
Scores say more about the rater than the employee. One study found 62% of the variance in performance ratings traced back to the rater, with actual performance explaining about 21% (Teamflect).
Reviews have earned their bad reputation: 88% of performance review mentions on Glassdoor in 2026 were negative, up from 83% in 2021, mostly over unclear criteria and pay disconnected from the outcome (4 Corner Resources).
Rate observable work, not traits. “Loads to the manifest without re-checks” is reviewable. “Attitude” is not.
Define every point on the scale in your own words, especially point 3. An undefined middle is where inflation lives.
Twice a year, twenty minutes, both parties writing beforehand. Annual reviews for hourly staff are too far apart to change anything.
Lower your delivery costs by 23%
How we reduce costs:
- No delivery vehicle expenses
- Optimized local routes
- Pay-per-delivery model
- Average 23% delivery cost reduction
What to include on an employee evaluation form
Seven blocks. Anything more and it stops getting filled in honestly; anything less and it isn’t a record.
1. The basics. Name, role, review period, reviewer, date. Name the period explicitly, as in “1 Jan to 30 Jun 2026”, because “this year” invites the recency bias where one bad week in May colours everything.
2. Three to five role-specific criteria. This is the block that decides whether the form works. They have to be things you can actually see, and they have to differ by role. For a delivery role: on-time departure, order accuracy at the door, vehicle and equipment condition, how exceptions get handled. For a packer: pick accuracy, pace against the order board, care with fragile product. For counter staff: order accuracy, customer handling, upsell on the named items.
Write them as behaviours, not qualities. “Checks every order against the manifest before loading” beats “attention to detail” because one of them can be answered with evidence.
3. A rating scale with defined points. More on this below. The point is that each number means something written down.
4. Two comment boxes per criterion. One for a specific example, one for what would move the score up. The example box is what makes the score defensible; the second is what makes the review useful.
5. Self-assessment. The same criteria, filled in by the employee before the meeting. It’s the highest-value, lowest-cost field on the form. It tells you where your view and theirs diverge, which is exactly where the conversation needs to go.
6. Goals from last time, and goals for next. Two or three, specific, with a date. “Trained on the Saturday wholesale run by 1 September” is a goal. “Be more proactive” is a wish.
7. Signatures and date. Both parties. It converts a conversation into a record, which matters if the role ever heads toward a performance plan or a dismissal.
What to leave off: personality traits, anything you heard second-hand, anything about the person’s life outside work, and any criterion you can’t give an example for. In the 11,000-review analysis, 73% of feedback-quality problems came down to feedback being about the employee rather than the work, with non-job-related personal detail the single most common issue at 36% (Teamflect).
A rating scale that resists inflation
Most forms use a five-point scale with undefined labels, which is functionally a two-point scale: 4 for everyone, 3 for someone you’re unhappy with. The fix is to write what each point means for this specific criterion.
| Point | What it has to mean | Example: order accuracy |
|---|---|---|
| 1 | Not meeting the standard; needs a plan with dates | Repeat errors on the same accounts after being shown |
| 2 | Inconsistent; right most days, wrong often enough to matter | A few wrong drops a month, caught by customers |
| 3 | Does the job to standard, reliably | Accurate; checks the manifest; errors are rare and self-caught |
| 4 | Above standard; catches problems others miss | Spots mislabelled orders before they load, flags account quirks |
| 5 | Sets the standard; others are trained on their method | The person you have write the SOP |
Two rules make this work.
Point 3 is a good score, and you have to say so out loud. Reliably doing the job is what you’re paying for. If 3 is tacitly a failing grade, every manager will award 4s to avoid a difficult conversation, and the scale dies.
5 should be rare and tied to something visible. If it means “we’d have this person teach it,” you’ll hand out very few, and they’ll mean something.
Also, don’t hide behind the number. Given that most of the influence on actual behaviour comes from written comments rather than scores (Teamflect), a 3 with a concrete example and one clear next step is worth more than a 4 with an empty box.
Should employees evaluate themselves?
Yes, and it’s the cheapest improvement available to the form.
Send the same form a week ahead and ask them to score themselves and write one example per criterion. Then compare. Three patterns come up and each tells you what the meeting is for:
- They score themselves lower than you did: Usually someone competent who’s never been told. The meeting is mostly reassurance plus one stretch goal.
- They score themselves higher: The gap is the whole agenda. Work from examples, not from the numbers, or it becomes an argument about scores.
- They score themselves the same: You agree on reality, so spend the time on the next step rather than the last six months.
For small teams, collect peer input informally: ask two colleagues what this person is best at and where they need cover. Full 360-degree processes get awkward fast in a 12-person business where everyone knows who said what.
How to run the review conversation
The form is the easy part. The meeting is where it either builds trust or burns it, and reviews are widely distrusted: 88% of Glassdoor mentions of performance reviews in 2026 were negative (4 Corner Resources).
A structure that works for hourly and operational staff:
- Twenty minutes, sitting down, not on the loading dock: Scheduled, not ambushed between runs.
- Both forms filled in beforehand: No filling it in live; that’s how you get fiction.
- Start with their self-assessment: “Tell me where you think you landed on accuracy.” You learn more in their first two minutes than in your prepared summary.
- Go criterion by criterion with examples: Specific dates and orders, not impressions.
- End with two or three goals and dates: Written on the form, both of you looking at it.
- Separate pay from the review, but not by much: Decoupling them entirely is what makes people cynical, since pay disconnected from review outcomes is one of the top complaints on record. Run the review, then be explicit: “This feeds the pay decision in March, and here’s what would move it.”
Never let a review be the first time someone hears about a problem. If a driver has been late four times, that conversation happens in week one, not in June. The review summarises; it doesn’t break news.
When a review turns into a performance plan
Sometimes the honest score is a 1, and the form’s job changes. A review records performance; a performance improvement plan is a separate document with a different purpose: specific behaviours, specific dates, specific support, and a clear statement of what happens if nothing changes.
Keep the two documents apart. Mixing them means your reviews are read as disciplinary, which makes everybody defensive and pushes every score back up to 4 again. If you need a starting structure rather than a blank page, Bit.ai publishes a performance improvement plan template you can work from. Check the current terms on their page before you commit to it.
Before you get there, rule out the two causes that look like a performance problem and aren’t: nobody ever taught them the job, and the standard was never written down. Both are fixed by affordable small business employee training rather than by a plan, and a surprising share of “bad hires” are untrained hires. Where the mismatch does go back to selection, the trial-shift step in the small business hiring process is the upstream fix.
How often should you review staff?
Twice a year for hourly and operational roles, plus a 90-day check for new hires.
Annual is too slow to change anything: a driver who picked up a bad loading habit in July shouldn’t hear about it in February. Quarterly is more than most small operations will actually sustain, and a process you abandon in month five is worse than one you do twice a year reliably.
Two practical notes on scheduling it. Put both dates in the calendar at the start of the year, including the 90-day check for anyone new. And don’t book reviews in your peak weeks. If December is when every wholesale account doubles its order, nobody is having a thoughtful conversation about goals. Check the leave calendar too, since the people you most want in the room are the ones most likely to have booked time off; the time off request form is where that visibility comes from.
Frequently asked questions
What should an employee evaluation form include? Employee and reviewer details, the exact review period, three to five role-specific and observable criteria, a rating scale with every point defined, a comment box for an example plus one for what would raise the score, a self-assessment section, goals from last period and for the next, and signatures from both parties.
Is a five-point rating scale the best option? It’s fine as long as you define each point in writing for each criterion and treat 3 as a good score. Undefined five-point scales collapse into “4 for everyone.” If you’d rather avoid numbers, three defined levels (below standard, at standard, sets the standard) work well for operational roles.
How do I avoid rating everyone the same? Define the middle of the scale as meeting the standard reliably and say so to the team, require a written example for every score, and rate observable behaviours rather than traits. It also helps to know that ratings carry a lot of rater noise: one study attributed 62% of rating variance to the rater versus about 21% to actual performance (Teamflect).
Should the review decide pay? It should inform it and the connection should be stated plainly. Reviews that are formally disconnected from pay get treated as theatre, and pay disconnected from review outcomes is one of the most common complaints people record about the process.
Do small businesses really need a written evaluation form? Yes, for three reasons: it forces you to define the standard, it gives the employee something concrete to work from rather than a vague impression, and it creates a dated record if the role later moves toward a performance plan or a dismissal.
