Two things haunt category picks: guessing the wrong peak month, and having no feel for how big a niche really is. Snapshots only show how many listings and what prices look like today — not which months sell, or how many units a category can absorb. 辉光ERP Category Analytics now defaults to a Demand Insights tab that turns platform-wide review history into seasonality, volume and opportunity scores, then back-calculates prep dates to the day.

Same entry: left nav Data Analytics → Category Analytics (Advanced tier). Demand Insights opens first; the older Analysis / Comparison tabs sit beside it. The four-level cascade filter stays in sync with ranking-row clicks, so the demand overview follows whichever level you drill into.

What this feature does for you

Six overview cards at a glance: estimated monthly units (with a range), estimated monthly GMV, growth vs the whole platform, active-sales ratio, newcomer share, and top-10 concentration. These are estimates meant for cross-category comparison, not exact sell-through — the page also states the data-through date up top.
Opportunity score 0–100: five peer-percentile inputs — per-SKU demand, newcomer opportunity, growth, peak approaching, and competition dispersion — weighted about 30% / 25% / 15% / 15% / 15%, with a progress bar per part and your rank among scored peers at the same level. Categories with too few recent reviews (under 20 on a leaf, under 50 higher up) get a dash, not a made-up score.
Season calendar + prep countdown: a 12-month demand index normalised to a mean of 1, peaks highlighted; the right panel shows "N days until the next peak" and three latest dates back-calculated from peak start — leadtime listing ~3 weeks early, air freight ~4 weeks, sea freight ~10 weeks. Dates already past are struck through, with a nudge to switch to a faster lane or trial a small leadtime batch.
Monthly demand trend: estimated units by month for about three years. Because the review API only reaches still-listed products, earlier months skew low — the chart notes to trust the last 1–2 years more than early absolute levels.
Sub-category / L4 opportunity boards: under All or an L1 department, sort by opportunity score, estimated monthly units, estimated monthly GMV, growth or newcomer share; filter Peak coming (within 10 weeks) or Flat year-round. Click a row and the cascade jumps there, with the hot-products table refreshing in lockstep.
Hot products, last 90 days: who is actually selling in this category lately, with products that first appeared in the past year flagged — so a category opportunity becomes a concrete list of rivals.

Where the numbers come from — and where they stop

The model reads reviews, not vibes. Each Takealot review has a date; many also say how long after purchase they were written — the system back-dates to a purchase month and builds the within-year shape from those months. Last-90-day review counts times a per-L1 calibrated units-per-review factor (default about 18, auto-calibrated from our exclusive-SKU orders, clamped 8–60) become estimated monthly units. Pilot checks matched common sense: pool stands ~4.6 in December vs ~0.01 in July; school bags ~2.7 in January; ink cartridges flat all year.

tip: the newest purchase months are often incomplete — months under 50% completeness are hidden; 50%–98% are completeness-adjusted and drawn lighter. Sparse leaf categories shrink their season shape toward the parent and are labelled as mostly referencing the parent. While the site-wide review backfill is still running, the tab shows Demand data preparing — no filler numbers. Category Analytics itself is an Advanced feature; Basic plans do not see the entry.

Real seller scenarios

Peak prep countdown: open a sharp seasonal niche like pool stands — December lights up as peak, and the right panel names the latest sea-freight ship-by date. Miss it and switch to air or a small leadtime trial instead of guessing "we should move around October."
Back-to-school window: filter Peak coming; school-bag and stationery leaves rise. Drill into a high-score L4 with decent newcomer share, scan last-90-day hot products, and decide whether there is still room to enter.
Flat categories for cash flow: filter Flat year-round, find ink / consumable niches hugging an index of 1, and pace restock from estimated monthly units — less capital locked in one seasonal sea shipment.
Cut "busy-looking" traps: a category with lots of listings and reviews, but an opportunity score dragged down by top-10 concentration and weak newcomer share — active-sales ratio makes it clearer: most SKUs barely sold in three months. Do not pile more SKUs in first.

Category Demand Insights is part of 辉光ERP — the all-in-one management platform for Takealot cross-border sellers in South Africa. Inspiration can pick the niche; the calendar and the numbers should pick the ship date.