A data story · Singapore · 2019 to 2026

How long will Singapore's protein hype last?

Protein bubble tea. Protein chips. Meiji or Rokeby? This year, Singapore put protein in everything. I wanted to know if it's hype that burns out or a habit that stays. Five questions, and you get to answer the first one.

Scroll ↓

Question 1 of 5

Is the protein hype already over?

Search interest climbed for three years and hit Google's top score of in . What happened next? Draw your guess for May to October in the shaded area.

Draw here ↘

Question 2 of 5

Will it crash like keto did?

Unlikely. Hype burns out. Protein is shaped like a habit.

Keto peaked at in , during the circuit breaker, and scores today. Each time keto dipped, it bottomed out lower. Since 2023, each protein dip has bottomed out higher.

Is that just keto? I checked 31 more trends.

Each row is one trend, scaled to its own peak and coloured by the shape it follows. The darker the cell, the closer that month was to the trend's high. Hover to read any month.

  • Burned out
  • Steady staples
  • Slow climbers
  • New surges
  • Burned out: keto, intermittent fasting, paleo. Diets with strict rules that peaked around 2019 and 2020, then never came back. This is what hype looks like.
  • Slow climbers: gym, calorie, the high-protein diet. A little more every year. This is what a habit forming looks like.

The word "protein" sits with the new surges. But the high-protein diet, the actual eating habit, sits with the slow climbers and has been building since 2019.

Google Trends, Singapore, 3-month averages. Shapes grouped with k-means clustering.

People are searching. But is anyone actually eating more?

Question 3 of 5

Is anyone actually eating more?

Searches nearly doubled. Eating and training rose about a tenth.

Singaporeans are careful: reading about protein doesn't immediately turn into eating more of it. Buying moved more than eating, and it is moving to the easiest format. Gym powders spiked and faded in brand searches, while ready-to-drink protein held.

And the people who need it most aren't the ones buying.

%

of adults aged 18 to 49 get enough protein

%

of adults aged 50 to 69 do

National Nutrition Survey 2025, National Population Health Survey 2025, Euromonitor via CNA, Google Trends.

So the hype is cooling and the habit is quiet. What happens next? One country has already been through this.

Question 4 of 5

What happens after the hype?

Look at Japan. The diet burned out. The nutrient kept climbing.

Japan had its own keto: 糖質制限, carb restriction. In , Japanese users searched it more than タンパク質, protein as a nutrient. It peaked in and has fallen to % of that peak. The powder word プロテイン peaked in and has held at more than half its high. The nutrient word just kept climbing. The hype cooled, and the conversation moved from the shaker to the plate.

And around the region

Hong Kong

Rose in step with Singapore through 2026, and is still above last year: against .

Malaysia

Same wave, same timing, still above last year: against .

South Korea

Never had a powder craze. 단백질, the everyday word for protein, climbed steadily and is above last year: against .

United States

The early warning. Peaked first, in , and is already below last year: against .

Google Trends, 3-month averages. Japan's three words come from one download, so they share a scale. In the cards, each country is scaled to its own peak since 2019, with Singapore as the faint dashed line. Data caveats are in the case study.

Put it together.

Question 5 of 5 · So how long will it last?

The protein hype is cooling. The protein habit isn't.

The habit isn't a guess: Singaporeans are eating about a tenth more protein than before, and in Japan and Korea the everyday word for protein has kept climbing long after the hype.

My bet, which I will score against each new month rather than a fixed date, since no model I built could call the turn: interest keeps easing, the way it already is in the US, and settles on a higher floor than before, the way it has in Japan since 2020. Protein is a nutrient, not a diet with rules, so it is harder to abandon. The habit is moving to the easiest format: ready-to-drink over powder.

The bigger opportunity isn't another protein bar for gym-goers. It's reaching the adults over 50 who genuinely need more.

Read the full case study ↗

Behind the story

How I know

Google Trends downloads

search terms

data points

analysis methods

hypotheses tested

dead ends, documented

The 12 hypotheses I testedTap a card to see the verdict and the evidence.
Was it Hyrox?Causal impact: what protein should have done, judging by other food trends.
  • Protein, actual
  • Expected from other trends
  • Normal range (95%)

Without fibre, April 2026 shoots above the band and Hyrox looks guilty. Add fibre and April falls inside the normal range. Fibre jumped at the same time as protein, and nobody eats more fibre for a fitness race, so both were riding a wider health wave. Without fibre in the model, the only explanation it had was Hyrox.

Regression on matcha, air fryer, collagen and Ozempic searches (plus fibre in the second model), trained on November 2021 to March 2026. A separate five-year Trends download.

Could a model have seen the turn coming?Four forecasts trained on 2019 to 2025, tested on 2026.

    The random forest stayed flat because it can't predict a value higher than anything it has seen, and the highest month in its training data was 58. The winner was the simplest rule, "next month looks like this month". It scored best because it always arrives one month late, which means it never predicts a turn at all.

    Methods, sources and limits
    • Google Trends, monthly, January 2019 to October 2026, downloaded in November 2026, for Singapore, the US, Japan, Hong Kong, Malaysia and South Korea. Trends scales each download on its own, so I downloaded every Singapore term paired with Protein and used Protein as an anchor.
    • Topics can glitch: Google reclassified the Protein topic in November 2016, Japan's Protein topic jumps from nothing in early 2021, and Korea's protein and keto topics switch off and on in the same months. Where a topic misbehaved, I used the words people actually type instead.
    • Methods: rolling averages, year-on-year comparison, seasonality, correlation of monthly changes, k-means clustering of 32 trend shapes, forecasting with a train/test split, causal impact, cross-country timing and brand comparisons.
    • What people eat: HPB National Nutrition Survey 2025 (self-reported). Strength training: MOH National Population Health Survey 2025. Sports protein sales: Euromonitor International as reported by CNA, 8 May 2026.
    • Dead ends: SFA food supply data, archived supermarket pages, weekly Trends data and a paid market report.
    • Limits: search shows interest, not behaviour. Some health questions now go to AI chatbots instead of Google, and people often search a brand instead of the category.
    • Built in Python on Google Colab with pandas, scikit-learn, statsmodels and Lets-Plot, as a personal project for the DDDM 2026 course.