The Original blend's customer rating dropped from 4.33 to 3.95 after May 2025 and now sits at 3.77 over the last six months. The one-time tin did not move. Moderation has been removing the negative reviews, which is why this has not shown up in any dashboard.
Original is sold as a subscription Starter Kit, a one-time 30-serving tin, and a 90-serving bag. It is the same blend in all three. If the product or the formulation had degraded, all three would have fallen together. They didn't.
This is as close to a controlled experiment as we get. The variable that changed is not the product — it is the purchase experience. Subscription buyers are rating us nearly four tenths of a star lower than they did, while one-time buyers of the identical tin are unchanged.
Deduplicated across both review platforms. Five-star share has fallen from three-quarters of reviews to under half.
August 2026: average 3.54, one-star share 19.4%, five-star share 42.6%. In April 2025 those numbers were 4.46, 4.5% and 74.3%. Roughly one in five people who write a review now gives one star.
Bazaarvoice rejected 3,382 Original reviews. Those rejects average 1.46 stars and two thirds of them are one-star. Every dashboard reports the approved number.
| Since May 2025 | Reviews | Average | One-star |
|---|---|---|---|
| As customers submitted | 4,945 | 3.87 | 12.9% |
| As published on site | 3,926 | 4.47 | 0.6% |
| Gap | −1,019 | 0.60 stars | 21× understated |
Two things to check. First, what the moderation rules are actually configured to reject — the export carries no reason codes, so we cannot tell from the data why any single review was removed. Some rejection is legitimate: spam, profanity, no product content. A 1.46-star average on the reject pile means whatever the rule is, it removes negatives at scale.
Second, our internal reporting should track the submitted rating, not the published one. The published number is a marketing asset. The submitted number is the customer telling us something.
134,479 Original cancellations, July 2025 to August 2026, with the reason each customer selected.
The top reason — "Wanted to try it first, not ready to commit to a subscription" at 20.9% — plus "I don't want a subscription" at 7.6% comes to 28.5%. That is the single largest block, and it is not a complaint about the product. It is a complaint about how the product was sold to them.
Product-quality reasons are a much smaller share: taste at 6.0% combined, "haven't felt the benefits" at 2.8%. Price is 6.4%.
Timing supports the same reading. Median time to cancellation is 48 days, and 38% cancel within the first 30 days — at or before the first refill lands.
| Time from signup to cancellation | Share |
|---|
Every product line, deduplicated across both platforms. "Change" compares the ten months before May 2025 with the sixteen months after. Original's fall is two to nine times larger than any other line.
| Line | Reviews | Lifetime | One-star | Last 6 mo | Change |
|---|
Hot Chocolate is the strongest thing we sell — 4.63 lifetime and only 2.9% one-star. Its −0.19 change is the second largest on the board, but it comes off a very high base and it has since recovered to 4.68 over the last six months. Worth understanding what it does right.
Coffee's 4.04 is a comparison artifact, not a verdict. Coffee has only existed since December 2025, so its entire life sits inside the weak period, while the legacy lines are averaging in years of higher ratings. Compared like for like — same months, same SKU type — Coffee is ahead of Original. See section 06.
Nourish is healthy at 4.40 since June 2025. Rest, Matcha, Turmeric and Mushroom Boost are essentially flat, down between 0.04 and 0.14. That range of movement is normal drift.
One caveat on lifetime averages generally: they reward age. Any line that existed before 2025 banked years of higher ratings, so it will look better than a line born into the current conditions. The like-for-like table in the next section is the fairer read.
A company-wide fulfilment or quality issue would have moved every line. It didn't. But the next section shows the subscription problem is not unique to Original either — Original is simply where it is worst and where the volume is.
Comparing lifetime averages punishes new products. This table holds the window constant — December 2025 to August 2026, the only period Coffee has existed — and splits every line into its subscription kit and its one-time tin.
| Line | Subscription kit | One-time tin | Penalty |
|---|
Coffee is not our weakest product. On a like-for-like basis its subscription kit rates 3.90 against Original's 3.63, and its one-time tin rates 4.44 against Original's 4.33. Coffee is outperforming Original in the current environment. Its low lifetime number is an artifact of having no good years to average in.
Every line carries a subscription penalty, between 0.36 and 0.75 stars. Same product, same months — the only difference is how it was bought. This is not an Original problem, it is a subscription-experience problem that Original feels hardest because roughly 87% of its reviews come from the subscription SKU.
The gap between the subscription kit and the identical one-time tin, over time.
Original's one-time tin has barely moved in four years — 4.26, 4.37, 4.43, 4.40, 4.37. Over the same period its subscription kit went 4.00, 4.06, 4.10, 3.78, 3.58. The penalty widened from 0.26 stars to 0.79.
The product is not the variable. The tin proves it. Something about the subscription experience got materially worse from mid-2025 onward, and it is now costing roughly eight tenths of a star.
28.5% of cancellations are people who did not understand or want a recurring order. That is a disclosure problem, not a product problem, and it is currently being paid for in one-star reviews. The fix is explicit language at the buy box and in the confirmation email, plus a pre-renewal notice with enough lead time to act on.
3,382 rejected Original reviews averaging 1.46 stars. Find out what the filter is set to catch, whether the rejection rate changed around May 2025, and start reporting the submitted rating internally alongside the published one.
Coffee's low reviews are not about the subscription — they are about expectation. The phrase "coffee black" appears in 4.9% of its low reviews and in none of its five-star reviews. The recurring complaints are a chemical or artificial coffee note, a "washed out" flavour against real coffee, and needing milk and sweetener to be drinkable. Its five-star reviews come from people already drinking Original who wanted a coffee-leaning option — "new coffee", "morning routine", "original mud". Coffee loses when it is framed as a coffee replacement for coffee purists and wins when it is framed as a MUD\WTR variant. That is a targeting and copy decision, not a formulation one.
Beyond subscription friction, the recurring product complaint is that it does not mix — grit at the bottom of the cup, appearing in about 15% of one-star reviews. That is partly solvable with preparation guidance, since the frother already ships in the Starter Kit, but it is worth confirming nothing changed in the powder itself.