Accuracy by Segment, Date & Grain
A blended accuracy number is an alibi - the truth lives at date, segment and day-of-week grain.
stairsThe Grain Ladder
Accuracy measured at the total-house monthly level looks stellar - and means nothing for daily pricing. The number is smooth because the grain hides everything that matters.
Descend the ladder - month to week to date, house to segment to rate code - and accuracy falls while relevance rises. Every step down costs comfort and buys usefulness.
The right grain is the grain of the decisions being fed: pricing lives at date and segment, so that is where accuracy must be scored.
grid_viewSix Grain Cuts That Matter
Six cuts expose where accuracy actually lives and dies. Each answers a different operating question, and each is cheap to score once the error data exists.
By Date
Each arrival date scored on its own - the grain at which price is actually set and where errors bite first.
By Day-of-Week
Monday and Saturday accuracy differ wildly - blended weekly numbers hide both problems.
By Segment
Corporate runs precise, leisure runs volatile, groups run lumpy - one house number hides all three behaviors.
By Booking Window
Accuracy at 90, 30 and 7 days out - the curve that tells you when the forecast becomes trustworthy.
By Room Type
Suite forecasts run noisier than standards - low-volume inventory carries wider error bands by nature.
By Season
Peak weeks versus shoulder periods compared - each season graded on its own terms, not averaged away.
visibilityBlended vs Disaggregated
A blended accuracy number is an alibi - it flatters the forecast by averaging away exactly the failures that need attention.
A 4 percent overall error reads as healthy while hiding a 12 percent Saturday error that is quietly mispricing every weekend in the quarter.
Disaggregated accuracy exposes the problem and makes it fixable: Saturday 12 percent, weekday 2 percent tells the team precisely where to spend its attention.
blur_onBlended Accuracy
4 percent overall error - a number that hides the 12 percent Saturday error doing the real damage. Comforting, and useless.
view_moduleDisaggregated Accuracy
Saturday 12 percent, weekday 2 percent - the weekend problem exposed, sized and fixable. The truth at the grain of the decision.
Score where you decide: price is set by date and segment - so accuracy must be scored by date and segment. Report the blended number to no one, or accept that it will flatter exactly the forecasts that need scrutiny.
translatePenjelasan dalam Bahasa Indonesiaexpand_more
Grain akurasi (accuracy grain) adalah tingkat pemecahan tempat akurasi forecast diukur, misalnya per tanggal, per segmen, atau per tipe kamar. Angka akurasi gabungan total-house sering tampak mulus justru karena menyembunyikan kegagalan yang penting. Semakin dituruni tangga grain dari bulan ke tanggal dan dari house ke rate code, akurasi menurun tetapi relevansinya naik.
- check_circleEnam potongan grain: per tanggal, day-of-week, segmen, booking window, tipe kamar, dan musim; masing-masing menjawab pertanyaan operasional berbeda.
- check_circleAkurasi Senin dan Sabtu bisa berbeda jauh; angka mingguan gabungan menyembunyikan keduanya sekaligus.
- check_circleSegmen corporate berjalan presisi, leisure volatil, dan group tidak teratur; satu angka house menyembunyikan ketiga perilaku itu.
- check_circleKurva booking window di 90, 30, dan 7 hari menunjukkan kapan forecast mulai layak dipercaya.
- check_circleContoh lesson: error gabungan 4 persen menyembunyikan error Sabtu 12 persen dan weekday 2 persen yang salah harga setiap akhir pekan.
- check_circleUkur akurasi di grain tempat keputusan dibuat: harga ditetapkan per tanggal dan segmen, maka di situlah akurasi dinilai.
helpPertanyaan Umum (FAQ)
Apa itu grain dalam akurasi forecast?
Grain adalah level detail tempat akurasi diukur, dari bulanan total-house hingga per tanggal, segmen, dan rate code. Semakin halus grain, akurasi tampak memburuk sementara relevansinya terhadap keputusan meningkat. Grain yang tepat adalah grain dari keputusan yang disokong forecast.
Mengapa angka akurasi gabungan menyesatkan?
Angka gabungan meratakan sekaligus menyembunyikan kegagalan yang paling perlu diperbaiki. Error keseluruhan 4 persen terlihat sehat padahal menyembunyikan error Sabtu 12 persen yang salah harga setiap akhir pekan. Angka terpecah seperti Sabtu 12 persen versus weekday 2 persen langsung menunjukkan lokasi masalah.
Apa saja potongan grain yang penting?
Enam potongan utama meliputi per tanggal, per day-of-week, per segmen, per booking window, per tipe kamar, dan per musim. Tiap potongan menjawab pertanyaan operasional berbeda dan murah dihitung begitu data error tersedia. Suite misalnya selalu lebih berisik daripada kamar standar karena volumenya kecil.
Di grain mana akurasi seharusnya dinilai?
Di grain tempat harga benar-benar ditetapkan, yaitu per tanggal dan per segmen. Angka bulanan total-house nyaris tidak berguna untuk pricing harian. Laporkan angka gabungan hanya dengan sadar bahwa ia akan memuluskan justru forecast yang paling butuh pengawasan.