Purplebricks customers love the agents. The weak point comes after they pay.
Purplebricks' agents are genuinely liked. The reviews turn negative in one place, after payment, and the same pattern shows up in any pay-upfront business.
We put 1,000 of Purplebricks' Trustpilot reviews through Sunbeam. Of the 526 reviews that mention an individual agent, 503 are positive. That is a score most estate agents, online or on the high street, would be glad to own. The reviews only turn negative in one place, and it is the same place each time - after the customer has paid.
You can read the full breakdown in the public dashboard. What follows is the short version, and why the shape of it matters for anyone who runs on reviews.
A four-star average hides where you actually lose people
Purplebricks sits on a healthy overall rating. Taken as one number, that reads as a solved problem. It isn't, and the single number is exactly why. An average blends the valuer who impressed someone in their living room with the sales progressor who went quiet for three weeks, and hands you back a figure that points at neither. To do anything useful with reviews you have to stop counting stars and start asking where in the journey each one was written.
The agents are the product, and customers know it
The warmth in the data is real and specific. Valuers get named and so do the people who held someone's hand through a sale after a bereavement or a divorce. Of the reviews about the individual agent, more than nine in ten are positive, and they read like personal thank-you notes. If Purplebricks' marketing leans on word of mouth, this is the raw material for it.
Communication drops once the money has landed
Across 160 reviews that talk about communication, the pattern is clear. One seller's review captures the whole arc in a few lines.
It is not an isolated grumble. Of 61 reviews about simply getting hold of someone, 44 are negative: ten deep in the call-centre queue, dropped calls, progressors unreachable for weeks. The early experience sets an expectation of attentiveness that the later experience does not keep.
The viewings that were logged but never happened
The sharpest version of the problem is operational rather than tonal. Of 54 reviews about viewings, 33 are negative, and several describe the same thing: a paid hosted viewing that nobody turned up to, then recorded on the system as if it had gone ahead.
Complaints follow the same after-payment shape. One customer describes paying £1,667 and being offered £650 to close a formal complaint that, in their telling, was never really answered. The numbers are small as a share of 1,000 reviews, but these are the reviews that get screenshotted and shared, and they all sit downstream of the moment the customer committed.
The lesson travels further than estate agents
Any business that takes payment upfront, then has to keep delivering, has a version of this curve. The expensive failures are not spread evenly across the journey. They concentrate at the points where the customer has already paid and is now waiting: the handover, the chase, the complaint. A blended satisfaction score will never show you that, because an average tells you the level and never the location.
The fix is simpler than a new survey. It starts with reading the feedback you already have, reviews, support tickets, NPS comments, and tagging each one by where in the journey it belongs before you count anything. Once the comments are sorted by stage rather than by score, the question stops being "are people happy" and becomes "happy when, and unhappy when", which is the version you can actually act on.
That stage-by-stage read is what Sunbeam does automatically with whatever feedback you already collect. If you want to see where your own curve dips, try it on your own data.