Our review section shows the reviews rated four stars and above, which is the honest way to run a marketing page and a dishonest way to learn anything. So we went back to the source table and read all 875 rated reviews with written feedback, from February 2020 to March 2026. That includes the twenty-four we would rather not have.
| Rating | Reviews | Share |
|---|---|---|
| 5 stars | 800 | 91.4% |
| 4 to 4.9 | 51 | 5.8% |
| 3 to 3.9 | 8 | 0.9% |
| 2 to 2.9 | 8 | 0.9% |
| Below 2 | 8 | 0.9% |
Twenty-four reviews at 3.6 stars or below, 2.7 percent of the rated set. Small enough to read every word of, which is what makes them worth something. A survey would have given us percentages. Reading them gave us the pattern.
The failures are not technical
Classified by hand, by the main thing the client is actually angry about:
| What the review is about | Reviews | Of 24 |
|---|---|---|
| Pace: missed dates, slippage, work taking far longer than agreed | 8 | 33% |
| Silence: chasing for updates, unanswered messages, not being listened to | 6 | 25% |
| Nothing delivered, or far less than the scope described | 4 | 17% |
| Integrity: a specification quietly not followed, or worse | 3 | 13% |
| Technical capability | 2 | 8% |
| Rating and written text pointing in opposite directions | 2 | 8% |
Two of twenty-four are about whether the person could do the work. Fourteen of twenty-four, well over half, are about pace and silence: how long it took, and how hard the client had to work to find out how long it was going to take.
The phrasing repeats almost word for word across six years and several different teams. Chasing constantly. Had to remind. Needs to update clients more often. Responses often delayed. Ignored for more than 48 hours. One review, generously, reduces the whole thing to a single instruction: update clients more often.
The happy clients name the same axis
If pace and communication were only a complaint, you could argue they are what people reach for when they are annoyed about something else. They are not. Across the 850 reviews rated four and above, here is how often each theme is named:
| Named in a positive review | Reviews | Share of 850 |
|---|---|---|
| Quality or skill | 279 | 32.8% |
| Communication, responsiveness, availability | 226 | 26.6% |
| Speed, promptness, hitting the date | 205 | 24.1% |
| Would hire again or work long term | 152 | 17.9% |
| Revisions, flexibility, going further than asked | 78 | 9.2% |
| Understood the brief | 66 | 7.8% |
Quality leads, but only just. Communication and speed together are named more often than the work itself. Clients are describing the experience of being kept informed at roughly the same rate as they describe the thing they actually bought.
Which gives an uncomfortable but useful conclusion. Technical quality is close to table stakes. It is necessary and it is not what the review is going to be about. The review is going to be about whether the client ever had to wonder what was happening.
What we changed because of this
Reading your own bad reviews is only worth the discomfort if something operational changes. Four things did, and all of them are about making silence visible before the client notices it:
- 01Every project has a thread, and hours are logged into that thread rather than into a private timesheet. A client asking what happened this week is answered by scrolling, not by asking someone.
- 02Every milestone carries its own dated status, so pending, in progress and completed are timestamps rather than opinions. Slippage shows up as a gap on a record instead of as a feeling.
- 03When an update is given, the next update has a date attached to it. The single most common request buried in those bad reviews is not faster work, it is a date.
- 04A quiet thread raises a flag internally rather than waiting for the client to raise it. The question a project manager asks is not is this finished, it is has anyone spoken to them, and if the answer is no, that is the problem to fix today.
None of that makes the work better. It makes the work legible, which turns out to be what most of the difference between a five star review and a one star review is made of.
What to take from this if you are hiring anyone
This is not really a finding about us. It generalises, and it generalises in a way that is directly useful when you are choosing between proposals that all look competent.
- Ask what the update rhythm is, in days, and who owns it by name. A proposal that cannot answer that has not thought about the part most likely to go wrong.
- Ask what happens when a date is going to be missed. Everyone misses dates. The difference between the good and bad outcomes in this dataset is entirely in whether the client heard about it first or found out later.
- Read the middling reviews of anyone you are considering, not the five star ones and not the one star ones. Three stars with a paragraph attached is where the actual information is.
- Get the specification in writing, including anything technical you care about, such as the language or platform something must be built in. Where that was verbal, it is the one failure mode in this dataset that no amount of good communication rescued.
Where these figures come from
- 875 rated client reviews with written feedback, collected across our delivery teams between February 2020 and March 2026, held in our own project-management database.
- The theme counts for the positive reviews are keyword matches against the review text and overlap, so they do not sum to 100 percent.
- The twenty-four low-rated reviews were classified by hand rather than by keyword, because at that sample size a keyword pass is less accurate than reading them.
- The review section on this site publishes only reviews of four stars and up. This article deliberately works from the unfiltered set, which is why the counts here do not match the counts there.