Friday, 02 October 2026 · Daily Market Intel

Market IntelFri, 02 Oct

Kelantan trader who sold US$240 a day shuts three shops in a year.

Footfall focus · power bills · spreadsheet forecasts

Someone's payday
- He once sold US$240 a day, so why did Syed Abu Bakar shut three Kelantan shops? — Malay Mail
  Syed Abu Bakar Syed Daud closed two grocery shops and an eatery in Rantau Panjang, a Kelantan town on the Thai border, in the past year. Sales that once topped US$240 a day took two or three days to reach.
  Positive-sum angle: Malaysia and China have agreed to a proposal to extend the East Coast Rail Link, a new line from Gombak near Kuala Lumpur to Kota Bharu, on to Rantau Panjang. A station at the border would bring shoppers to the traders who stayed and freight to both countries.
  What's the impact: A Kelantan trader weighing a second shop should wait for a signed opening date; the extension is still a proposal. Fellow trader Fu Chee Kat says shoppers stay away for fear of border tax on a single item, so a plain guide to what is taxed comes first.
Your costs
- Singapore power just fell 10.6% for businesses, so why does SP Group warn of a January rise? — AsiaOne
  From 1 October to 31 December, Singapore's electricity tariff drops by an average 10.6%, or US$0.026 a kilowatt-hour, the unit on a power bill. SP Group, the grid operator, expects a rise next quarter because fuel prices have climbed since September.
  Positive-sum angle: A cheaper quarter lands on every bill at once: the bakery's oven, the laundrette's dryer, the cold room behind a hawker stall. The saving arrives in the year-end months, when small shops spend most on stock.
  What's the impact: A Singapore café or laundrette on the regulated tariff should ask electricity retailers for fixed-price quotes this month, before the January rate is set. A 12-month plan signed now keeps the next rise off the bill.
The tool
- Guessing which customers will reorder? NVIDIA's free model predicts from 60,000 spreadsheet rows, no training — MarkTechPost
  NVIDIA released Kumo Tabular, a free family of models that reads a table of past records with known outcomes and predicts the outcome for new rows. It needs no training step and was built for tables of up to 60,000 rows and 100 columns.
  Positive-sum angle: Prediction from a spreadsheet used to need a data scientist and weeks of tuning. The licence allows commercial use, so bookkeepers, agencies and software makers can build it into the tools small firms already run.
  What's the impact: A Kuala Lumpur or Jakarta wholesaler with two years of orders in a spreadsheet can test which customers will reorder, on one PC with an NVIDIA graphics card. The release is code for developers, with no upload-a-file page; that wrapper is a solo founder's opening.

Never miss a SEA deal.