The product holds a 4.3-star average across 3,104 reviews. That number hides the most important thing about it: it is being held up by build quality and dragged down by battery life.

Reviews that mention build quality average 4.7 stars. Reviews that mention battery life average 3.1. Same product, same customers — a 1.6-star spread depending on which feature the reviewer happened to focus on. Battery is mentioned in 742 reviews (24%), more than any other single theme, and it is the one theme that consistently pulls the rating below 4.

The second story is demand, not defect. The companion app is requested in 498 reviews — not a complaint about something broken, but a clear, repeated ask for something that does not exist yet. That is the cleanest feature signal in the dataset.

What is working: hardware build, display, and the in-box setup experience. What is not: battery life — and the v4 "battery optimisation" release made the sentiment worse, not better.

Six themes account for the entire dataset, but three carry 64% of all mentions: battery life (24%), build quality (21%), and display (19%).

Two of those three are strengths. Build quality shows up in praise — "feels solid", "premium", "no creaks" appear across 884 reviews and average 4.7 stars. Display is similar: bright, sharp, no complaints about colour accuracy. These themes are doing exactly what you want a product strength to do — showing up often and pulling the rating up.

Battery life is the inverse. It is mentioned as often as build quality but in the opposite register. The word "all day" appears in 211 battery reviews — almost always negated: "doesn’t last all day", "barely makes it through the day". This is not a fringe complaint. It is the single most common thing a customer says about the product, and it is negative four times out of five.

The companion app sits apart from all of this. It is not a complaint about a defect — it is 498 customers asking for something that does not exist. That distinction matters: defects erode trust, missing features create upgrade demand.

Build quality reviews average 4.7 stars. Battery reviews average 3.1. The product average sits at 4.3 — a weighted blend of a strength and a weakness, not a description of either.

This is the trap with a single headline rating. A 4.3 reads as "good with room to improve". The feature-level breakdown reads as "one specific thing is broken". The second framing is the one that tells you what to do.

The companion-app number is the subtle one. Reviews that mention the missing app average 3.8 stars — below the product average. These are not unhappy customers; they like the hardware. But the absence of the app is enough to cost roughly half a star from people who would otherwise rate 4 or 5. A missing feature is not free. It shows up in the rating whether or not you have built it.

The ANOVA confirms the feature you mention genuinely predicts your rating (F=31.9, p<0.001, η²=0.21 — a large effect). This is not reviewers being noisy. Which feature a customer cares about explains a fifth of the variance in how they rate the product.

The v4 release shipped with a "battery optimisation" in the release notes. The reviews disagree that it optimised anything. Battery sentiment fell from 3.4 to 2.9 stars between v3 and v4 — the steepest single-release drop in the dataset.

Reading the v4 battery reviews explains why. The word "background" appears in 38% of post-v4 battery complaints, up from 11% before v4. The optimisation appears to have throttled background activity — customers report notifications arriving late and the companion features they rely on going silent. The battery may genuinely last longer; the experience of using the product got worse, and reviews measure experience, not milliamp-hours.

This is the kind of finding a star average will never surface. v4’s overall rating barely moved — the build-quality and display praise absorbed it. But the battery sub-population moved half a star, and battery is already your weakest theme. You shipped a fix that the people who needed it most experienced as a regression.

The regression on release number is strong (t=-5.43, p<0.001, R²=0.74). Battery sentiment has declined every release, and v4 accelerated it. This is a trend, not a blip — and it is pointed the wrong way.

The 4.3 average reads as "consistently good". The distribution says otherwise. 48% of reviews are five stars and 16% are one star — the two extremes hold 64% of the dataset between them. The middle is hollow: only 10% of reviews land on three stars.

This is a polarised product, not a uniformly good one. An average is only a fair summary when ratings cluster around it. Here they do not — they cluster at both ends. The 4.3 is the arithmetic midpoint of two crowds who feel very differently, and it describes almost nobody’s actual experience.

The shape matters because it tells you the one-star reviews are not noise to be averaged away. 486 customers had an experience bad enough to rate the lowest possible score, and the theme breakdown already told you what most of them are about: battery life. Fix the bottom of this distribution and the average moves more than any amount of marketing to the top.

The chi-squared test confirms the distribution is genuinely bimodal, not uniform (chi2=1284.6, p<0.001, Cramer’s V=0.44 — a large effect). The polarisation is real and structural.

The frequency chart told you how often each theme comes up. This one tells you which way it leans. They are different questions, and battery is where the answers diverge.

Five of the six themes lean positive. Build quality is 91% positive, display 88%, setup 74%. Even price — the theme people most expect to be a complaint — runs 56% positive against 22% negative. Customers mention these themes because they like them.

Battery life is the exception, and it is not close: 79% negative, 18% positive. It is the only theme in the entire dataset where negative comment outnumbers positive. Connectivity leans negative too (61%), but it is a small theme. Battery is both large and lopsided — the worst combination.

That is the case for sequencing the v5 sprint the way the recommendation does. Build quality and display do not need defending; they already win their split decisively. Battery is the single theme actively pulling customers toward the one-star end of the distribution, and the sentiment split shows there is no positive counterweight inside the theme to soften it.

These are representative quotes, not cherry-picked extremes.

On battery (the v4 problem): "After the latest update my notifications come through an hour late. I think they tried to save battery by killing background stuff. I’d rather charge it twice a day and actually get my alerts." The "background" complaint, stated plainly. 38% of post-v4 battery reviews say a version of this.

On build quality: "This feels like it costs twice what I paid. No creaks, no flex, the buttons are crisp. Genuinely premium." "Premium" and "solid" appear in 280+ reviews. This is a real, defensible strength — lead with it.

On the missing companion app: "Great hardware, but why is there no app? I want to change settings from my phone, not dig through menus on the device. Knocking off a star until that exists." The half-star penalty for a missing feature, said out loud.

On setup: "Out of the box to working in under five minutes. The instructions actually matched the product." Setup is quietly one of your best themes — 4.2 stars and rising. Do not break it.

Three things, in priority order:

1. Roll back or fix the v4 background throttling. This is the highest-leverage fix in the report. Battery is already your weakest theme, and v4 made it worse for the exact customers who care most. Restoring reliable background behaviour should recover the 0.5-star v4 drop and stop the trend. Do this before anything else — it is a regression, not a roadmap item.

2. Commit to the companion app for v5. 498 customers asked for it, unprompted, and the absence already costs roughly half a star from people who otherwise rate you well. This is the rare feature request backed by a measurable rating penalty. It is not a nice-to-have; it is leaving rating on the table.

3. Put build quality and display at the front of your store copy. 4.7 and 4.6 stars, 90% positive. These themes are doing the work of holding your average up. Marketing should be repeating the customers’ own words — "premium", "solid", "no creaks" — because the reviews prove they land.

Note what is not here: connectivity (3.6 stars, 6% of reviews) is a real issue but small. Fix battery and ship the app first; connectivity is a v6 line item.

If the v4 background throttling stays as shipped, battery sentiment continues its decline — the trend line points at 2.5 stars by v6, and at that point battery stops being a weak theme and becomes the headline of every negative review. The 4.3 average starts to slip.

If v5 ships the background fix plus the companion app: the battery sub-population recovers toward 3.4 (its v3 level), and the 498 app-requesters lose their reason to dock a star. Modelling those two moves against current review mix lifts the overall average from 4.3 toward 4.5 — not from a marketing push, just from closing the two gaps the customers already told you about.

The build-quality and display strengths are stable and require no investment to maintain — only the discipline not to compromise them while fixing battery. The whole v5 thesis is simple: stop the regression, ship the obvious missing feature, and protect what already works.