際際滷

際際滷Share a Scribd company logo
Codename	Raging	Bull	
Smart	Low	Latency	Price	
Availability	in	Travel	
Marcel	Dumont	,	CTO	BookerZzz	
22	September	2016
About Us
150	employees	
AcFve	in	8	countries	
500.000	bookings	
Since	2002	
We	help	more	than	
1	million	happy	
customers	per	year	
to	experience	their	
perfect	stay
Bookerzzz Smart Low Latency Price Availability in Travel
My	Background
 1,5	yrs	CTO	BookerZzz	
 Casual	Gaming,		
Spele.nl,	spiela鍖e.de	
 Co-Founder	
Schoolbank.nl	
 Online	Classi鍖eds,	
Gulfnews,	The	Times,	
BostonGlobe	
Data	Volume	
Time
Raging Bull
Part	of	our	overall	
rebuild	of	the	
plaVorm	
Refers	to	areas	
where	we	can	
increase	our	pro鍖t	
margin	by	being	
smarter.	
Work	in	progress
Cause:	High	Latency
Bookerzzz Smart Low Latency Price Availability in Travel
Some	basic	travel	principles.		
Hotel		
(PMS)	
Channel	Manager	
(CM)	
Online	Travel	
Agency	
(OTA)
Reality	
11000+	Hotels		
(PMS)	
100+	Channel	
Manager	
(CM)	
1000s	Online	Travel	
Agency	
(OTA)
Some Stats / day
1-1.5m	update	
messages	received	
75m+	permutaFons.	
40%	of	our	
infrastructure	is	Fed	
up	processing	these	
permutaFons.	
Average	processing	
Fme	60-100sec
0%	
10%	
20%	
30%	
40%	
50%	
60%	
70%	
80%	
90%	
100%	
2010	 2011	 2012	 2013	 2014	 2015	 2016	
manual	
<1	day	
<6	hr	
<1	hr	
<	5	min	
Update	Speed	Channel	Managers	over	
the	years
0	
50	
100	
150	
200	
250	
300	
0%	
10%	
20%	
30%	
40%	
50%	
60%	
70%	
80%	
90%	
100%	
2010	 2011	 2012	 2013	 2014	 2015	 2016	
manual	
<1	day	
<6	hr	
<1	hr	
<	5	min	
TTL	Sec	
Compared	to	our	processing	Time	
(TTL)
How	to	cope	with	the	鍖ood	of	updates	
	
Dedicated	
Import	
Modules	
(PHP)	
Message	
Queues	
In-Memory	DB	
(Aerospike)	
Search	
(ElasFc	
Search)	
Pricing	Engine	
(GoLang)	
Price	Package	
Availability	
(GoLang)	
ERP	
Data	
Warehouse	
DataCrunching	
(R)		
RealFme	
Reports	
(Tableau)		
External	
APIs	
RelaFonal	DB	
(MySQL)
Data		
=		
Knowledge		
=		
Opportunity's
AnalyFcs	Department
2	Data	ScienFsts	
3	Analysts	
	
Tooling	
R	
Tableau	
SQL	
Good	Old	Excel
Price	OpFmalisaFon	by	the	Hotel	
Sold-out,	1	week	prior	
1	room	lei	
mulFple	rooms	lei
Further	reducing	cancelaFon	rate	
High	volume,		high	cancel	
=	bad	habit
More info?
Marcel Dumont
CTO BookerZzz
Next Steps
 IdenFfy	under-
performers	at	an	
earlier	stage.	
 Increase	segmentaFon	
 Predict	shortages	
 Play	with	price	
elasFcity	
Lots	to	do!
More info?
Marcel Dumont
CTO BookerZzz
nl.linkedin.com/in/marceldumont

More Related Content

Bookerzzz Smart Low Latency Price Availability in Travel