Coffee Category Benchmark 2026

Four coffee chains, 446 verified Malaysian conversations, one month. AI-assisted tagging, checked by a human before it ships.

This is a working sample built on live public data — Google reviews across 80 outlets and YouTube conversation on four chains, anonymised for this sample: Heritage Kopitiam, a national kopitiam chain; Campaign Chain, a grab-and-go chain that ran a celebrity music-video campaign; International Chain, a global coffee brand; and App-First Chain, a fast-growing chain built around app ordering. Every figure on this page is counted from the dataset, not estimated. Your report would follow the same format.

967

Items collected

Reviews, comments and the posts that spawned them.

446

Actually on topic

After a filtering pass, checked by a human, removed 390 off-topic items.

47%

Noise removed

Nearly half of what a keyword alert would have shown you.

14%

Not in English

Bahasa Melayu, Manglish and Chinese.

01 / THE READ

What a month of Malaysian coffee talk actually says

Three things stand out this month. Campaign Chain's music-video push produced the biggest spike of the period: a single music video on 1 October drove 24 on-topic mentions in one day, and Campaign Chain's mentions overall were 76% positive. Heritage Kopitiam has the deepest review base in the category: 24,252 Google ratings at 4.40. App-First Chain's conversation centres on operations: 65% of its wait-time mentions and 43% of its taste mentions are negative or mixed, and its Google rating is 3.66 against a category average above 4.3.

Sentiment share by brand

On-topic mentions only, 18 Sep – 17 Oct 2026. Segments are counts; the figure on the right is net sentiment (positive minus negative, as a share of all mentions for that brand).

Where each brand sits

Net sentiment in social conversation against the weighted Google rating of the same 20 outlets per brand. Circle size is the number of Google ratings behind that score.

What this pair of axes shows

Two different publics, measured separately.

A Google rating is the verdict of people who walked in and paid. Social sentiment is the verdict of people watching from outside, most of whom did not. When they disagree, that gap is the story.

Heritage Kopitiam carries 24,252 ratings at 4.40, by far the largest review base in the category. App-First Chain sits lowest on both axes this month: a smaller review base (1,258 ratings), a 3.66 rating and net sentiment of +23.

Campaign Chain leads on both axes this month, but note what that is made of: 34 of its 94 mentions come from one campaign week.

Full counts behind the charts above. Weighted rating = each outlet's Google rating weighted by its number of ratings, across 20 outlets per brand. Noise removed = share of collected items for that brand that were not about the brand.
Brand	On topic	Positive	Mixed	Negative	Neutral	Net	Rating	Ratings	Noise removed

03 / ONE EVENT

One campaign, one spike: what it looked like

This campaign ran inside the reporting window. It is the kind of event a monthly summary usually reports weeks after it has happened.

Campaign Chain: a celebrity music-video campaign

Campaign · 1 Oct

Campaign Chain released a celebrity music-video campaign on 1 October. It produced the single biggest day in the dataset and the most engaged conversation of any brand: 657 likes on Campaign Chain's on-topic comments, against 153 for Heritage Kopitiam and one apiece for International Chain and App-First Chain.

Mentions on day one

+66

Net sentiment

34%

In BM or Manglish

A commenter praised the brand for hiring real performers rather than using AI.YouTube comment · Manglish · 47 likes · Paraphrased

What a dashboard would miss: a third of the praise arrived in Bahasa Melayu and Manglish, and the most-liked comment of the campaign uses playful sarcasm that reads as affection, not criticism.

04 / THE COMPLAINTS

Where each brand's complaints concentrate

Every on-topic mention is tagged to what it is about. This grid shows the share of each topic's mentions that were negative or mixed — so a dark cell means that when customers talk about this, they are usually unhappy. Taste and service carry the most mentions.

The one-line version: Campaign Chain's main friction point is wait time (30%); International Chain's most-discussed topic is service (54 mentions, 30% negative or mixed); Heritage Kopitiam's is service (36%), alongside strong taste scores (19%); App-First Chain's highest negative shares are wait time (65%) and taste (43%).

05 / IN THEIR WORDS

What customers said, paraphrased

Summarised from the same dataset and flagged during tagging as representative. Each one is paraphrased in plain English rather than quoted word for word, so no comment can be traced back to its author. No names, no handles, no links.

PositiveCampaign ChainCampaignParaphrased

Customers praised the campaign as money well spent.

YouTube · English · 3 Oct

PositiveCampaign ChainTaste · PackagingParaphrased

A customer enjoyed their latte and liked the Malaysia-themed cup design.

YouTube · BM · 2 Oct

NegativeSarcasm flaggedCampaign ChainParaphrased

Reviewers felt that less-ice orders still arrived full of ice.

Google review · English · no negative keyword in the original

MixedHeritage KopitiamTaste · PriceParaphrased

A first-time visitor liked the food but found the sambal portion small for the price.

Google review · English

NegativeApp-First ChainPriceParaphrased

A reviewer felt lattes were under-filled for the price.

Google review · English

06 / METHOD

How the number 446 was arrived at

The reason this report is worth paying for is the part that happens between collection and the chart. Here it is, in full.

Collect

Google reviews across 20 outlets per brand, plus YouTube posts and comments naming the brand. 967 items in total, including the 131 posts that drew the comments.

Filter

836 comments and reviews were read and judged on topic or not. 390 were not about the brand at all and were dropped before any counting.

Tag

Each surviving item gets an AI-assisted sentiment, a language, up to eight topic tags, and flags for sarcasm and quotability. Sarcasm is marked separately because it inverts sentiment.

Check

AI-assisted tagging, checked by a human before it ships: the tags, the spikes and the paraphrased comments are reviewed before delivery.

What filtering removed

The gap between the first and last bar is the difference between a keyword alert and a report.

App-First Chain's keyword pull was the most affected: 115 of its 217 collected items turned out to be comments on unrelated travel videos, in Japanese and Chinese, that mentioned the brand only in passing. Heritage Kopitiam lost 214 of 361 to unrelated commentary picked up alongside the brand name. Campaign Chain lost 7 of 101, the result of a tight query.

Language mix of on-topic talk

446 on-topic items by the language they were written in.

9 items were flagged as sarcasm. Read literally, each one lands in the wrong sentiment bucket, which is how a keyword-based tool would read them.

What this sample does and does not cover

Sources in this sample. Public Google reviews and public YouTube posts and comments, only. Threads, Xiaohongshu, TikTok, Instagram, Facebook and news coverage are available as added scope and are not included in these figures.
Handling. Brand names are anonymised in this sample, and customer comments are paraphrased. Nothing in this sample identifies an individual. Author names, handles and post links stay in the working dataset and go to the client privately; they are never printed into a document that circulates.
Review dates. The review sample is the most recent reviews per outlet, so it reads as a standing snapshot of each brand rather than a 30-day flow. 320 of 360 on-topic reviews were written in 2026.
Small cells. Any topic with fewer than 10 mentions for a brand is marked on the grid. Price for International Chain is 3 mentions; treat it as a flag to watch, not a finding.
Sentiment is tagged in context, not by keyword. AI-assisted tagging reads each item in context, including Manglish, code-switching and sarcasm, and a human checks the result before it ships.
Figures refreshed 17 October 2026, 09:47 MYT. Every number here is reproducible from the source dataset.

07 / WORKING TOGETHER

What your brand would get, every month

Same method, scaled to your size: from one cafe against 3 nearby rivals up to your brand and the three competitors you name. Larger tiers are delivered as a page like this one, plus the underlying tagged dataset so your team can slice it themselves.

Starter

RM199 ~~/ month~~

Your cafe vs 3 nearby rivals you pick
Google reviews compared on wait, price, taste, service and ambience
Top 3 praises and complaints, yours vs theirs
3 reply drafts for your toughest reviews
Monthly one-page summary, human-checked

Pulse

RM399 ~~/ month~~

Your brand only
Google reviews across your outlets, plus YouTube
Monthly report, human-checked
Tagged dataset included