📊 Price Tracker Dashboard

📈 Market Analyze

Auto-generated · 2026-10-16 11:21 · channel gaps / coverage live in the 📡 Coverage tab.

📌 What stands out today

Auto-generated read of this run. Channel gaps / coverage live in the 📡 Coverage tab.

• 354 of our priced listings have at least one rival cheaper in the same price & quality range (see the Head-to-head below).
• Most contested category: Accessories (5 of 5 listings with a cheaper rival) · least contested: PC (2 of 3).
• Promo pressure: 915 of 1140 listings are discounted today (median discount 27%).
• Data health: 0 dead links · 152 listings without a fresh price (detail in the Alerts tab).

📊 Snapshot

Live figures from the price database.

🏷️ Brand mix by category

Who we are up against, per category, by number of listings today.

🥊 Head-to-head — all our products vs the competition

Sub-tabs by CATEGORY (phones / tablets / …) and a filter by CHANNEL. Inside, our products are grouped by RANGE = price band × spec level (RAM+ROM from the listing name) — not by BOM — so each card is «our product vs the rivals in the SAME price and quality range». Rivals are fresh (≤14 days), one per brand. The colored chip on each row is its CHANNEL.

💰 Price bands — us vs competition

Where each side plays. A band with many rivals and few of ours is contested space.

🛒 Channel comparison — median price by category

Median of today's prices per channel. Green = cheapest channel in that category.

⚔️ Price pressure — our median vs competitors', by category

Only categories where both sides have priced products. Bar length = size of the gap.

📉 Price movement — last 7 days

Real movement from the capture history. Moves over ±60% are excluded as bad captures (see the audit note below).

📉 Price movement — last 30 days

Same rule, 30-day window: shows the trend, not the noise.

📡 Channel coverage — by family (Huawei)

✔ has listing · ✘ missing (required channel) · · not applicable (that channel does not sell the category). Edit families in familias_market.csv and exceptions in cobertura_no_aplica.csv

🏁 Coverage vs competition — tracked products by brand

➕ Agregar producto (1 a la vez)

Required fields: URL, Brand, RRP Oficial, Category. If any is missing or the link already exists, it is not added. The scraper will follow the link and fill name/price, keeping your Brand and RRP.

📄 Links existentes (0)

🔮 Forecast — Christmas 2026 season

Target window: Dec 4 – 24 (the 3 weeks before Dec 24), with two peaks: the Dec 4-6 weekend (Liverpool Venta Nocturna) and the Dec 18-24 close. The full season starts Nov 6 (Buen Fin, confirmed) and does not stop until Christmas.

⚠️ VERSION 2 — observed data + event-driven projection. Floor and ceiling are the levels each product has ALREADY hit in the history (nothing invented). The dashed line is the forecast: it only moves on the campaigns that apply to each product’s channel, and it is a projection, not a certainty.

Event weights (real discount depth). The dashed line drops ONLY on the campaigns that apply to each product’s channel, and by that product’s own measured depth scaled by the real weight of the event — not one flat average, because Hot Sale is not Venta Nocturna: El Buen Fin 1.00 · Liverpool Venta Nocturna 0.95 (Christmas) / 0.90 (anniversary) · Black Friday 0.90 · Prime Day 0.80 · Cyber Monday 0.70 · Hot Sale (AMVO) 0.65 · MELI Día Doble 0.55 · seasonal retail 0.60 · holidays 0.45. After each campaign the price recovers, so the line goes down and back up. Sources: Profeco (Buen Fin: 57.5% of products dropped, −15.5% average on those that dropped), AMVO / Hot Sale 2025 (−14% average), Liverpool Venta Nocturna (up to −40/50%), and a 5-year price-history analysis (Buen Fin wins on tech with 25–40% direct). A product with no measured campaign response stays flat — that is why e.g. an older watch does not drop while the newer one does.

Precision: 100.0% (live: predictions only mature on Dec 31, so this number says nothing yet — the honest score is below) · 0 prediction(s) scored so far
Old backtest reference: 94.9% on 1822 campaigns — inflated: it trained and tested on the same days (leakage). Kept only to compare.
Formula (price): precision = 100 − MAPE, MAPE = mean(|predicted − actual| / actual) × 100. Formula (will it hit its floor?): precision = (1 − Brier) × 100, Brier = mean((p/100 − outcome)²). Every run logs its forecast (data/forecast_ledger.jsonl) and it is scored once the horizon passes, so the score and the weights update by themselves. Self-correction on the smaller events: amz x0.6 · cop x0.6 · hua x0.6 · liv x0.6 · ml x0.6 · sea x0.61 · tel x0.6.

Backtest detail: 1822 cases in the last 150 days · mean error ±5.1% · bias 4.13% (positive = we predict a deeper drop than reality) · within ±5%: 62.8% · price actually reached its floor 43.9% of the time.

Data hygiene: Floor = 5th percentile and ceiling = 95th percentile of the observed prices, so a single junk reading cannot define them. Prices below 30% or above 200% of the official RRP are also dropped (scraper junk, inflated third-party listings). Nothing is deleted from the database — the floor is just computed on healthy data.

📅 Campaign calendar — full year

● LIVE NOW: Hot Travel (AMVO) — until Oct 18

Oct 16, 2026 · 79% of the year elapsed · ⭐ = key for the business · base: prior-year dates until the current year's official date is confirmed

📊 Category overview

Read this first. Room to floor = median % a product in that category can still fall before hitting its own lowest level. Floor vs RRP = how deep its floor already sits under the official RRP.

🎯 Room to fall — cards for every product

The same card as before, but for the whole catalog, driven by the filters above: change category, brand, channel, price, observations or minimum room and the cards rebuild instantly. Default view = the products with the most room to fall. Sorted by the same criteria as the filter bar.

Blue line = lowest price captured each day (90 days) · green dashed = its own historic floor · red dot = latest reading. ⚠ = junk reading, ⏱ = no capture today (read with care).

How to read the bar under each card (the RRP scale, 100% = list price): the bar runs from $0 to the official RRP, so 100% = the product is selling at list price. The white mark is today's price; the green mark is its historic floor; the pale green stretch (from 0 to the floor) is how far it has ALREADY come down, and the coloured stretch (from the floor to today) is the room it still has left to fall. The label repeats it in numbers: today 96% RRP · floor 40% RRP · can fall 59%. Red = more than 25% of room left, amber = 8-25%, green = almost no room left. If the price is above its RRP the bar is full and it says above RRP.

📊 Segment view — slice the data yourself

Room to fall = % a product can still drop before hitting its own historic floor (higher = more space for a campaign). Disc. vs RRP = how deep it already sits under the official RRP (negative = priced ABOVE its RRP). Charts 1-5 rebuild themselves with the filters below; the counters and the table always show exactly what is being plotted.

📉 Every product — its own price line, by category

The same reading as the TOP 15 cards, but for every monitored product, grouped by category and sorted by how much room it still has to fall. Blue line = lowest price captured each day (last 90 days) · green dashed = its own historic floor · red dot = latest reading. The bar shows how far it can still drop.

Gaps in a line = no capture those days (the width is real time). ⚠ = junk reading (below 30% or above 200% of the RRP); ⏱ = no capture today. Both mean: read that row with care.