Introductie tot Hotel Reviews Werner van Sprundel Managing Director Co-founder
Online Hotel Review Analytics Via deze korte rapportage wilt Dexam U graag op de hoogste stellen van verschillende mogelijkheden wat betreft het minen van informatie in online hotel reviews.
Paul van Loon Director of Analytics Co-founder
Dexam demonstreert middels een beknopt ’voorbeeld rapport’ de verschillende aspecten van deze analyse en gaat in op zowel achterliggende technieken, alsook de resultaten. De besproken methodieken in deze rapportage zijn deels gestandaardiseerd en kunnen tot op zekere hoogte geautomatiseerd worden toegepast; anderzijds zijn de besproken analyses maatwerk en case-specific. Dexam licht de besproken analyses graag toe en zoekt graag samen met U naar de juiste analytische oplossing.
Contents 1 Introductie
2
2 Informatie uit Numerieke Scores
3
2.1
Vergelijking van ’Uitstekende’ Scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3
2.2
Vergelijking van ’Ondermaatse’ Scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
2.3
’Uitstekend’ en ’Ondermaats’ uitdiepen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
3 Historische Ontwikkeling Numerieke Scores
8
4 Text Analyse op Reviews 4.1
10
Text Analyse op Review Titels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
4.1.1
Positief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
4.1.2
Negatief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
4.2
Sentiment Topic Analyse indien expliciet wordt gevraagd naar positieve en negatieve punten .
12
4.3
Sentiment Topic Analyse indien niet expliciet wordt gevraagd naar positieve en negatieve punten 12 4.3.1
Locatie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
4.3.2
Internet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
4.3.3
Badkamer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
4.3.4
Receptie . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
4.3.5
Eten . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
4.3.6
Staff en Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
4.3.7
Lawaai . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
5 Samenvatting Resultaten
15
6 Overzicht
16
1
Private and Confidential
1
Introductie
Voor hoteliers is klanttevredenheid van groot belang. Potentile klanten maken immers een keuze uit een veelvuldigheid van hotels in een bepaald gebied op basis van individuele beoordelingscriteria. Het zijn de klanten die de vraag bepalen, en daarmee samenhangend ook invloed hebben op de prijs van een kamer (de connectie tussen vraag en aanbod). Het veelvuldig doen van klanttevredenheidsonderzoeken, via bijvoorbeeld enqutes, is een must voor elke hotelier. Het aanpassen aan de wensen van potentile klanten blijkt een positief effect te hebben op zowel de prijs per kamer, evenals de bezettingsgraad. Daarnaast is het internetgebruik in de laatste decennia sterk gegroeid. Steeds meer mensen uiten zich op websites als TripAdvisor en Booking.com en steeds meer mensen lezen reviews op deze websites. Zo vermeldt TripAdvisor dat er momenteel 125 miljoen reviews op de website staan en dat er 260 miljoen unieke maandelijkse bezoekers zijn. Booking.com vermeldt dat er elke dag 550.000 kamers worden geboekt, en dat reviews daarbij van groot belang zijn. Internet biedt voor zowel hoteliers als mogelijke klanten een grote bron aan informatie en deze bron zal blijven groeien, zeker naarmate het aantal reviews zicht blijft opstapelen. Potentile klanten maken veelvuldig gebruik van deze informatie. Voor hoteliers is deze mogelijkheid er ook. Echter, voor hoteliers kan deze informatie worden samengevat waardoor zij geen honderden of wellicht duizenden reviews hoeven te lezen. Weet u nog wat er in review 1, 2 of 10 staat als u aangekomen bent bij review 500? Dexam biedt hiervoor een oplossing. Een oplossing om u een voordeel ten opzichte van uw concurrenten te geven. Een oplossing die aangeeft waar u actie moet ondernemen om direct de klanttevredenheid te verbeteren. Een oplossing, die uiteindelijk kan zorgen voor een hogere prijs per kamer en een hogere bezettingsgraad. Door het gebruik van complexe algoritmes herleidt Dexam wat klanten belangrijk vonden tijdens hun verblijf in het hotel. In alle reviews wordt door middel van tekst- en sentimentanalyse gekeken hoe een hotel scoort ten opzichte van concurrerende hotels. Hieruit komen bepaalde positieve en negatieve punten. Een voorbeeld: uw hotel scoort in 80% van de reviews erg goed op service bij de receptie, maar in 40% van de reviews komt naar voren dat wi-fi en kwaliteit van het diner als matig tot negatief worden ervaren. Dit zijn directe actiepunten waar u iets aan kunt doen. Daarbij kunt u ook inspelen op de zwaktes van uw directe concurrenten. Het doel van dit rapport is om u, op een korte en eenvoudige manier, inzicht te geven in hoe uw hotel wordt beoordeeld door klanten en hoe deze beoordeling staat ten opzichte van directe concurrenten. Daarbij maakt Dexam gebruik van diverse websites waar reviews worden aangeboden. Aan het einde van dit voorbeeld rapport hoopt Dexam dat u de waarde hiervan inziet.
2
Private and Confidential
2
Informatie uit Numerieke Scores
Bij Dexam geloven wij in de kracht van tekstanalyse om tot unieke inzichten te komen. Inzichten die niet, of slechts in zeer beperkte mate, naar boven komen bij traditionele enqutes. Op verschillende websites heeft de bezoeker de mogelijkheid om, naast het schrijven van zijn/haar review, een aantal numerieke scores achter te laten. Dit zijn zowel scores voor de overall-ervaring, maar ook verder toegespitst op aspecten van hun verblijf (bijvoorbeeld: tevredenheid over de kamer of het personeel). Dexam gelooft dat deze scores een uitstekend globaal beeld kunnen geven van klanttevredenheid, met name wanneer deze score in de context van de lokale markt (concurrenten) wordt gezien. Volgende hoofdstukken gaan dieper in op deze globale scores.
2.1
Vergelijking van ’Uitstekende’ Scores
Het Hilton hotel in Antwerpen staat centraal in dit voorbeeld rapport. Dit hotel opereert niet in een vacum in het stadscentrum van Antwerpen, vandaar dat Dexam altijd tracht scores en analyses weer te geven in de context van andere spelers in de markt. Alleen op deze manier vallen sterke en zwakke punten op. Het 4-sterren Hilton hotel is onderhevig aan concurrentie van onder andere ketens als ’Crowne Plaza’, ’Radisson Blu Astrid’, maar ook hotels als ’Leopold’ en het kleinere ’Julien’. Gezien de 4-sterren status van het hotel worden, met oog op de schaling van scores, ’uitstekende’ scores gedefinieerd als scores van 5 uit 5. In Figuur 1 is het percentage scores met 5 sterren weergegeven voor alle hotels, evenals per categorie. Top Scores 5***** Overall
Value
Location
Sleep Quality
Rooms
Cleanliness
Service
75%
HotelName 50%
total
Score
Crowne Plaza Antwerp Hilton Antwerp Hotel Hotel Julien Leopold Hotel Antwerp Radisson Blu Astrid Hotel
25%
0%
l l l l l l l l l l l l l l p p p p p p p p p p p p p p en en en en en en en er er er er er er er er er er er er er er ote ote ote ote ote ote ote ote ote ote ote ote ote ote uli uli uli uli uli uli uli tw ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J ntw rid H ntw rid H ntw rid H ntw rid H ntw rid H ntw rid H ntw rid H p H tel J An t t t t t t er lA lA lA lA lA lA lA st st st st st st st a A twe a A twe a A twe a A twe a A twe a A twe za Ho Ho Ho Ho Ho Ho Ho ote lu A ote lu A ote lu A ote lu A ote lu A ote lu A ote lu A ntw n n n n n n laz laz laz laz laz laz H H H H H H H A A A A A A A P P P P P P B B B B B B B n n n n n n n ne ne ne ne ne ne old old old old old old old ton ton ton ton ton ton ton so so so so so so so op op op op op op op ow ow ow ow ow ow Hil Hil Hil Hil Hil Hil Hil dis dis dis dis dis dis dis Le Le Le Le Le Le Le Cr Cr Cr Cr Cr Cr Ra Ra Ra Ra Ra Ra Ra
la eP
n ow Cr
Figure 1: Overzicht van top scores (vijf sterren) onderverdeeld naar Hotel en (sub-)categorie. Figuur 1 bevat veel informatie en kan grondig worden bestudeerd voor aanwijzingen. Wat duidelijk naar voren komt is dat: • De scores van Hilton en Julien op Locatie overal bovenuit steken
3
Private and Confidential
• Julien in 6 van de 7 categorien de meeste 5-sterren beoordelingen krijgt • Na Julien, het Hilton de beste scores behaalt • Crowne Plaza in elke categorie het laagste scoort • Leopold en Radisson Blu vergelijkbaar scoren op de meeste categorien Voor het Hilton hotel in Antwerpen is locatie dus een erg sterk punt. De (veel) kleinere concurrent Julien blijkt sterk te scoren op de meeste categorien en vormt wellicht een bedreiging voor het Hilton hotel. Echter, het analyseren van het percentage excellente scores is slechts 1 manier om inzicht te verkrijgen in klanttevredenheid. Een alternatieve manier is het analyseren van het percentage beoordelingen dat ’laag’ scoort. In het geval van een 4-sterren hotel wordt een lage score gedefinieerd als een 1, 2 of 3-sterren beoordeling.
2.2
Vergelijking van ’Ondermaatse’ Scores
Het lijkt vrij logisch dat een analyse van lage scores het tegenovergestelde laat zien van de analyse van de hoge scores. Toch hoeft dit niet altijd zo te zijn. Een 4-sterren hotel kan door klanten ook als gewoon goed worden bevonden, zonder dat het laag of hoog scoort (een 4-sterren beoordeling). Bottom Scores 1*, 2** of 3*** Overall
Value
Location
Sleep Quality
Rooms
Cleanliness
Service
50%
40%
HotelName
30%
total
Score
Crowne Plaza Antwerp Hilton Antwerp Hotel Hotel Julien Leopold Hotel Antwerp Radisson Blu Astrid Hotel 20%
10%
0%
ne row
C
l l l l l l l p p p p p p p p p p p p p p tel tel tel tel tel tel tel en en en en en en en er er er er er er er er er er er er er er ote ote ote ote ote ote ote uli uli uli uli uli uli uli Ho Ho Ho Ho Ho Ho Ho tw tw tw tw tw tw tw tw ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J ntw rp H el J p H tel J An An strid An strid An strid An strid An strid An strid An strid t t t t t t er a A twe a A twe a A twe a A twe a A twe a A twe za tel tel tel tel tel tel tel A A A A A A A Ho Ho Ho Ho Ho Ho Ho ntw laz laz laz laz laz laz Ho Ho Ho Ho Ho Ho Ho An An An An An An Blu Blu Blu Blu Blu Blu Blu nA e P ton e P ton e P ton e P ton e P ton e P ton ld ld ld ld ld ld ld o n n n n n n n n n n n n n o o o o o o o t so so so so so so so op op op op op op op ow ow ow ow ow ow Hil Hil Hil Hil Hil Hil Hil dis dis dis dis dis dis dis Le Le Le Le Le Le Le Cr Cr Cr Cr Cr Cr Ra Ra Ra Ra Ra Ra Ra
Pla
Figure 2: Overzicht van ondermaatse scores (minder dan vier sterren) onderverdeeld naar Hotel en (sub)categorie. Uit Figuur 2 blijkt dat het Crowne Plaza hotel overal bovenuit steekt. Dit hotel scoort veel vaker behoorlijk ondermaats dan de andere hotels. In combinatie met Figuur 1, leidt dit tot de conclusie dat het Crowne Plaza hotel niet alleen consequent minder excellente scores behaalt (Figuur 1), maar ook steevast meer slechte beoordelingen krijgt (Figuur 2). Voor het Hilton hotel geldt een ander beeld. Waar in figuur 1 de 4
Private and Confidential
excellente scores voor het Hilton beter waren dan alle hotels (op het kleinere hotel Julien na), blijkt uit Figuur 2 dat het Hilton vrij vergelijkbaar (met uitzondering van Crowne Plaza) scoort op slechte beoordelingen. Locatie is hierop de uitzondering, waaruit blijkt dat het Hilton hotel hier simpelweg fantastisch op scoort. Als overall conclusie geldt vooralsnog dat een verdere analyse moet worden gedaan op wat klanten aanmerken in slechte beoordelingen aan het Hilton hotel. Hoewel relatieve scores het meest informatief zijn, tracht Dexam ook informatie te halen uit absolute scores. Dit betekent dat Dexam nagaat op welk van de categorien bijvoorbeeld worden de meeste ’slechte’ scores worden gegeven. Het antwoord uit Figuur 2 lijkt Value te zijn. Andere categorien lijken erg vergelijkbaar te scoren, met uitzondering van Cleanliness, wat over het algemeen in orde blijkt te zijn en niet vaak terugkomt in reviews van klanten.
2.3
’Uitstekend’ en ’Ondermaats’ uitdiepen
Op verschillende websites met hotel reviews wordt bij elke beoordeling aangegeven tot welk marktsegment de klant behoort. Daarbij wordt vaak onderscheid gemaakt tussen vijf mogelijke profielen: 1. Business 2. Couple 3. Family 4. Friends 5. Solo Een andere mogelijkheid is om analyses op te splitsen naar type klant, waardoor U een uniek inzicht krijgt in de klanttevredenheid van verschillende typen bezoekers. Dit is weergegeven voor de 5-sterren beoordelingen (uitmuntende scores) in Figuur 3)
5
Private and Confidential
Top Scores 5***** business
couple
family
friends
solo
total
Crowne Plaza Antwerp
100% 75% 50% 25% 0% 100%
Hilton Antwerp Hotel
75% 50% 25% 0% 100%
variable Overall
Score
Hotel Julien
75% 50% 25%
Value Location Sleep Quality Rooms Cleanliness
0% 100%
Service
Leopold Hotel Antwerp
75% 50% 25%
Radisson Blu Astrid Hotel
0% 100% 75% 50% 25% 0% s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
s all lity ice ms lue ion es er Va ocat Qua Roo nlin Serv Ov a L ep Cle Sle
Figure 3: Verfijning van Figuur 1. De extra onderverdeling naar bezoekerstype is gemaakt. Het is belangrijk om in acht te nemen dat niet elk bezoekerstype even vaak voorkomt. Groepen met een klein aantal kunnen resultaten beinvloeden; groepen die zeker te klein zijn hebben wij alvast weggelaten. De laatste kolom van Figuur 3 (de totaal kolom) laat de data zien die eerder in Figuur 1 is weergegeven. Door op deze manier te groeperen per hotel, ligt de nadruk meer op het vergelijken van scores op verschillende categorien IN een hotel in plaats van bepaalde categorien vergelijken TUSSEN hotels. Hier wordt de eerste stap gemaakt naar individuele prestaties van hotels. Uit de laatste kolom blijkt dat, voor het Hilton hotel, de zeer goede score op Locatie erg opvallend is. De score op Locatie voor hotel Julien steekt ook boven andere categorien uit, terwijl dit minder het geval is voor hotel Leopold of het Radisson Blu Astrid hotel. Wat betreft het Crowne Plaza is het tegenovergestelde waar; locatie scoort laag in vergelijking met de andere categorien. Een andere conclusie met betrekking tot het Hilton hotel is de opvallend lage score op Value. Een analyse van de scores per type bezoeker toont een eentonig beeld voor het Hilton hotel. Verschillende bezoekerstypen geven in uitmuntende beoordelingen vergelijkbare scores. Een andere manier om Figuur 3 weer te geven is middels een dot-chart (Figuur 4). De dot-chart laat de gemiddelde score per categorie zien voor de 5 bezoekerstypen, opgesplitst naar hotel. Een opmerking bij Figuur 3 is dat scores voor bezoekerstypen zijn weggelaten bij onvoldoende beoordelingen. Dit is gedaan om de robuustheid van de resultaten te behouden. In dit voorbeeld rapport volstaat deze korte introductie naar bezoekerstype. Dexam gaat graag dieper in op uw specifieke vraagstukken.
6
Private and Confidential
Radisson Blu Astrid Hotel
● ●
Hotel Julien
●
●
●
●
● ●
●
●
Hilton Antwerp Hotel
●
Crowne Plaza Antwerpen
●
●
● ●
●
●
●
●
●●
● ● ●● ●
●
● ●
Hilton Antwerp Hotel
●
Crowne Plaza Antwerpen
●
●
● ●
● ●
●
●
●
●
●
Leopold Hotel Antwerp
●●
●
●
● ●
●
●
●
●
●
●●
Hilton Antwerp Hotel
●●
Crowne Plaza Antwerpen
●
●
●
●
Hotel
●
Leopold Hotel Antwerp
● ●
●
●● ●
● ●
●
●
●
●●
Hilton Antwerp Hotel
●
● ●●●
Crowne Plaza Antwerpen
●
●
Radisson Blu Astrid Hotel
●
● ●
●
● ●
● ● ●
●● ●
●
●
●
●
●
●●
●
●
● ●
●● ● ●
●
●
●
Hilton Antwerp Hotel
●● ●
●
●
Leopold Hotel Antwerp Hotel Julien
●
●●
Crowne Plaza Antwerpen
●
0
1
2
3
●
●●
4
friends
●
solo
●
●
● ● ●
●● ●
● ●
●
● ●
● ●● ●
Hilton Antwerp Hotel
family
●
Mean Service
Radisson Blu Astrid Hotel
couple
●
Mean Cleanliness
● ●
Leopold Hotel Antwerp
Crowne Plaza Antwerpen
business
●
●
Radisson Blu Astrid Hotel
Hotel Julien
●
●
●
●
Hilton Antwerp Hotel
Type
Mean Rooms
Leopold Hotel Antwerp Hotel Julien
●
●●
●
Crowne Plaza Antwerpen
●
Mean Sleep Quality
Radisson Blu Astrid Hotel
Hotel Julien
● ●●
●
Mean Location
Radisson Blu Astrid Hotel
Hotel Julien
●
Mean Value
Leopold Hotel Antwerp ●
●
●● ●●
● ●●
Radisson Blu Astrid Hotel
Hotel Julien
Mean Overall
Leopold Hotel Antwerp
● ●
● ●●
●
● ●
5
value
Figure 4: Net als Figuur 3 een weergave van de scores naar bezoekers type. Opnieuw dient U waakzaam te zijn met het intepreteren van scores voor groepen met een kleine aantal reviews.
7
Private and Confidential
3
Historische Ontwikkeling Numerieke Scores
Een belangrijke vraag om de robuustheid (correctheid) van resultaten te waarborgen is een analyse van het aantal reviews dat per kwartaal wordt gegeven. Ten opzichte van voorgaande kwartalen is er een stijgende lijn in het aantal reviews (Figuur 5). Waar in vroegere kwartalen het aantal reviews zeer beperkt was (bij alle vijf hotels), is er momenteel een overvloed aan informatie. Hieruit kan geconcludeerd worden dat websites meer en meer informatie bevatten, wat door hoteliers in hun voordeel gebruikt kan worden. Op deze potentiele vraag speelt Dexam in. Het kleinste hotel in deze steekproef (Hotel Julien) heeft een minder sterke groei dan de grotere hotels. Een andere mogelijkheid is om een analyse per kwartaal te doen (mits voldoende reviews). Deze mogelijkheid wordt realistischer, zeker omdat het aantal reviews per kwartaal stijgt. Met deze stijging kan ook een betere analyse per bezoekerstype worden gedaan, zoals is weergegeven in het voorgaand hoofdstuk. Een analyse per kwartaal geeft inzicht of een bepaald kwartaal beter of slechter scoort in vergelijking met andere kwartalen. Ook kan door middel van een dergelijke analyse worden nagegaan of een verandering of actie van de betreffende hotelier heeft geresulteerd in een betere score. Als laatste kan deze tijdsanalyse worden toegepast in combinatie met de tekstanalyse, die zal worden behandeld in het volgende hoofdstuk. Zo kan per geselecteerd woord worden nagegaan hoe het sentiment (positief of negatief) is veranderd in de loop van de tijd. Aantal Reviews per Kwartaal
50
Aantal Reviews
40
30
Group.1 Crowne Plaza Antwerp Hilton Antwerp Hotel Hotel Julien Leopold Hotel Antwerp Radisson Blu Astrid Hotel
20
10
0 3 4 1 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 _Q 3_Q 4_Q 5_Q 5_Q 5_Q 5_Q 6_Q 6_Q 6_Q 6_Q 7_Q 7_Q 7_Q 7_Q 8_Q 8_Q 8_Q 8_Q 9_Q 9_Q 9_Q 9_Q 0_Q 0_Q 0_Q 0_Q 1_Q 1_Q 1_Q 1_Q 2_Q 2_Q 2_Q 2_Q 3_Q 3_Q 3_Q 3_Q 03 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20
20
Kwartaal
Figure 5: Aantal Reviews per hotel (per kwartaal) dat wij online gevonden hebben op de reviewsite. Een voorbeeld dat inzicht geeft in een dergelijke tijdsanalyse is hoe de categorien (Location, Value, Sleep Quality, Rooms, Cleanliness en Service) veranderen per hotel per kwartaal (Figuur 6). De gekleurde lijn geeft hierin de best mogelijke lijn weer in de loop der tijd, terwijl de grijze banden de variantie aangeven (95% betrouwbaarheidsinterval). Hieruit blijkt dat het Hilton hotel, Hotel Julien en Crowne Plaze een vrij constante trend over alle categorien over tijd laat zien (met uitzondering van enkele categorien). Het Radisson Blu Astrid hotel laat een negatieve trend zien op vrijwel alle categorien (Figuur 6). Hotel Leopold 8
Private and Confidential
daarentegen laat een positieve trend voor enkele categorien. Door een tijdsaspect mee te nemen in de analyse geeft het niet alleen de huidige sterke en zwakke punten van een hotel weer (met enkele concurrenten als referentiepunt), maar ook hoe scores zich hebben ontwikkeld. Het Hilton hotel dient zich voornamelijk te richten op Value, Cleanliness en in mindere mate op Rooms. Deze gaan iets achteruit in de loop der tijd. Uit eerdere analyses is naar voren gekomen dat Cleanliness niet vaak genoemd wordt in reviews, waardoor de focus voor het Hilton hotel is op Value en Rooms. GEM Rating_location
Rating_value
Rating_sleepquality
Rating_rooms
Rating_cleaneliness
Rating_service
5.5
Hilton Antwerp Hotel
5.0 4.5 4.0 3.5 3.0 5.5 5.0
Hotel Julien
4.5 4.0 3.5
Average Rating
Radisson Blu Astrid Hotel
3.0 5.5 5.0 4.5 4.0 3.5
Crowne Plaza Antwerp
3.0 5.5 5.0 4.5 4.0 3.5
Leopold Hotel Antwerp
3.0 5.5 5.0 4.5 4.0 3.5 3.0 3 4 1 2 3 4 1 2 3 4 1 2 3 4 3 4 1 2 3 4 1 2 3 4 1 2 3 4 3 4 1 2 3 4 1 2 3 4 1 2 3 4 3 4 1 2 3 4 1 2 3 4 1 2 3 4 3 4 1 2 3 4 1 2 3 4 1 2 3 4 3 4 1 2 3 4 1 2 3 4 1 2 3 4 _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q _Q 10 10 11 11 11 11 12 12 12 12 13 13 13 13 010 010 011 011 011 011 012 012 012 012 013 013 013 013 010 010 011 011 011 011 012 012 012 012 013 013 013 013 010 010 011 011 011 011 012 012 012 012 013 013 013 013 010 010 011 011 011 011 012 012 012 012 013 013 013 013 010 010 011 011 011 011 012 012 012 012 013 013 013 013 20 20 20 20 20 20 20 20 20 20 20 20 20 20 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
Kwartaal
Figure 6: De grafiek geeft de ontiwikkeling weer van scores (numeriek) zoals die per hotel, per subcategorie, per kwartaal zijn gegeven. De extra dimensie die hier wordt toegevoegd kan van grote waarde zijn voor het stellen van toekomstige prioriteiten, alsook een bevestiging van het effect van hotel specifieke ontwikkelingen.
9
Private and Confidential
4
Text Analyse op Reviews
Dexam gelooft dat de ongestructureerde informatie, opgesloten in de daadwerkelijke review van klanten, de meeste waardevolle informatie bevat. Voorgaande analyses vatten scores samen en geven een globaal beeld. Een slechte of goede score op bijvoorbeeld de categorie Value of Service geeft niet aan wat er precies goed of slecht is. Deze informatie kan Dexam voor U analyseren uit de tekst die klanten typen in hun beoordeling. Voor een dergelijke tekstanalyse gebruikt Dexam verschillende databronnen. Het voordeel van het gebruiken van verschillende bronnen is dat het de mogelijkheid geeft gebruik te maken van eigenschappen van een bron, die een andere bron niet heeft. Op deze manier vult het elkaar aan. Ook geeft het vertrouwen wanneer dezelfde resultaten blijken uit de verschillende bronnen. Bij tekstanalyse kijkt Dexam in het bijzonder naar: 1. Titels van reviews • Titels blijken een zeer korte en bondige samenvatting te zijn van de ervaring die een klant heeft gehad tijdens zijn of haar verblijf in een hotel. Dexam heeft onderzocht dat aspecten die genoemd worden in de titel van grote invloed zijn op de totale tevredenheid (overall rating) van de bezoeker, zowel in positieve als negatieve zin. 2. Topics • Op bepaalde bronnen wordt expliciet weergegeven welk deel van de review (tekst) als negatief of als positief is ervaren. Hierdoor kan een sentiment analyse worden uitgevoerd. Aangezien zulke reviews over het algemeen kort en krachtig zijn, is deze analyse robuust. • Andere bronnen tonen geen onderscheid tussen negatieve en positieve punten in de reviews. Bij deze bronnen staat vaak een uitgebreide tekst die de ervaring van een klant beschrijft met meer diepgang. Ook hier is sentiment analyse mogelijk. Daarnaast is Dexam in staat om een sentiment score van veel meer en specifiekere topics te achterhalen. Hiervoor heeft Dexam voorgeprogrammeerde topics gedefinieerd, maar dit wordt ook graag in overleg met U gedaan als andere topics onderzocht moeten worden. 3. Diepgaande textanalyse • Diepgaande tekstanalyse gaat een stapje verder dan bovenstaande methoden; hier kan echte winst behaald worden. De diepgaande analyse trachte te achterhalen waarom een hotel goed (slecht) scoort op een bepaald topic. Een dergelijke diepgaande analyse is absoluut maatwerk en wij overleggen graag over de juiste oplossing.
4.1
Text Analyse op Review Titels
In deze paragraaf zal een analyse worden gedaan op de titels die aan elke review zijn meegegeven. Uit onderzoek van Dexam blijkt dat de titels over het algemeen een samenvatting van de review zijn en dat zij een grote invloed hebben op de totale klanttevredenheid. 4.1.1
Positief
Uit de titels die worden aangemerkt als positief, omdat zij positieve woorden bevatten (bijvoorbeeld: amazing, good, best, enz.), blijkt het woord hotel vaak voor te komen. Een voorbeeld van zo een titel kan zijn: Het hotel was goed. De algoritmes van Dexam leggen automatisch de link tussen het positieve woord goed en het bijbehorende zelfstandig naamwoord hotel. Tabel 1 laat zien dat voor het Hilton hotel ook de locatie 10
Private and Confidential
erg vaak als positief wordt opgemerkt (in 65.30% van alle reviews). In vergelijking met de andere hotels wordt value, service en business weinig genoemd. Voor het kleinere hotel Julien, zijn kenmerkende woorden in positieve titels ook boutique, design en style. Dit kleinere hotel lijkt het meer te moeten hebben van zijn onderscheidende karakter. Topic
Hilton
Radison
Julien
Leopold
Crowne Plaza
antwerp
9.50
2.70
9.80
3.80
10.30
breakfast
1.40
8.20
2.40
5.70
2.60
business
0.70
2.70
0.00
1.90
15.40
city
4.80
4.10
12.20
0.00
0.00
everything
1.40
0.00
0.00
1.90
0.00
hotel
33.30
34.20
41.50
35.80
33.30
location
65.30
49.30
26.80
32.10
5.10
service
8.20
11.00
14.60
5.70
12.80
staff
8.20
1.40
2.40
11.30
2.60
value
2.00
2.70
0.00
7.50
5.10
boutique
0.00
0.00
14.60
0.00
0.00
design
0.00
0.00
4.90
0.00
0.00
style
0.00
0.00
7.30
0.00
0.00
Table 1: Highly Pos % titels 4.1.2
Negatief
Dezelfde analyse kan worden gedaan op titels die als negatief worden aangemerkt, omdat zij bijvoorbeeld woorden als bad, horrible, terrible, enz. bevatten. Uit Tabel 2 blijkt dat voor het Hilton hotel de woorden breakfast, business en rooms vaak voorkomen in negatieve titels, zeker in vergelijking met andere hotels. Deze woorden kunnen door Dexam verder worden onderzocht door te analyseren wat er precies negatief is aan deze woorden. Bijvoorbeeld: zijn de tijden van het ontbijt niet flexibel genoeg, is er onvoldoende keuze bij het ontbijt, worden bepaalde ontbijtproducten als niet-lekker ervaren, is het ontbijt te duur of een combinatie van deze redenen? Wat verder opvalt in de tabel is dat de negatieve titels van het Leopold hotel vaak de woorden hotel, location, room, service en value bevatten. Bij dit hotel blijkt dus iets mis te zijn met zowel de locatie, de kamers, de service als de prijs. Topic
Hilton
Radison
Julien
Leopold
Crowne Plaza
antwerp
4.00
0.00
0.00
0.00
0.00
breakfast
4.00
7.10
0.00
0.00
0.00
business
4.00
7.10
0.00
0.00
3.70
hotel
8.00
14.30
0.00
14.30
14.80
location
4.00
0.00
0.00
14.30
3.70
room
4.00
7.10
0.00
14.30
3.70
rooms
4.00
0.00
0.00
0.00
3.70
service
0.00
7.10
0.00
14.30
11.10
value
0.00
0.00
0.00
14.30
0.00
staff
0.00
0.00
0.00
0.00
7.40
Table 2: Highly Neg % titels 11
Private and Confidential
4.2
Sentiment Topic Analyse indien expliciet wordt gevraagd naar positieve en negatieve punten
Indien uitdrukkelijk wordt gevraagd naar positieve en negatieve punten is een sentiment analyse vrij makkelijk om te doen. Als nadeel worden er vaak alleen steekwoorden gegeven bij de positieve en negatieve punten, waardoor diepgang ontbreekt. Voor het Hilton hotel heeft Dexam meer dan 900 beoordelingen met zulke kort en bondige punten. Van deze reviews zijn enkel de Nederlandstalige geanalyseerd. Uit Tabel 4.2 blijkt dat de locatie van het Hilton hotel erg positief is. Ook staff en service is positief. Internet wordt vaak opgemerkt als negatief, en zo ook de badkamer. Hier dient verder onderzoek naar gedaan te worden.
4.3
Topic
% Positief
% Negatief
Ratio Pos - Neg
Locatie
52.02
0.42
124.33
Internet
0.56
12.13
0.05
Badkamer
6.83
13.39
0.51
Slaapkamer
22.59
17.29
1.31
Receptie
5.02
4.88
1.03
Eten
34.31
22.73
1.51
Staff & Service
46.72
9.62
4.86
Lawaai
0.42
1.39
0.3
Sentiment Topic Analyse indien niet expliciet wordt gevraagd naar positieve en negatieve punten
Indien niet uitdrukkelijk wordt gevraagd naar positieve en negatieve punten, dan zijn deze opgesloten in de totaalbeoordeling van een klant. Dexam kan deze positieve en negatieve punten voor U uit dergelijke reviews halen. Door middel van verschillende algoritmes worden positieve en negatieve woorden, welke afkomstig zijn uit verschillende psychologische studies, geanalyseerd en wordt gekeken naar welke onderwerpend deze woorden verwijzen. Dit proces is complexer, maar geeft U uiteindelijk meer diepgang omdat de klantenbeoordelingen vaak uitgebreider zijn. Dexam heeft als voorbeeld voor U gezocht op het sentiment van de onderwerpen: locatie, internet, badkamer, receptie, eten, staff en service en lawaai. Deze onderwerpen zijn enkel een voorbeeld. Deze onderwerpen moeten blijken uit voorgaande analyses, wat hier niet helemaal het geval is. Eventueel vragen wij U ook om input om een overzichtelijke lijst te krijgen van onderwerpen waar u graag meer over zou willen weten. 4.3.1
Locatie
Met betrekking tot het onderwerp locatie, blijkt dat dit veelvuldig voorkomt in reviews bij het Hilton hotel, maar ook bij het Radisson Blu Astrid hotel (Tabel 3). Dit onderwerp wordt gemiddeld ook meer dan n keer genoemd per review omdat het percentage boven de 100% is voor deze hotels. Ook blijkt dat het voornamelijk voorkomt in positieve reviews. Voor het Crowne Plaza hotel blijkt het omgekeerde waar te zijn. Het onderwerp wordt, in vergelijking met de andere hotels, weinig genoemd en als het wordt genoemd heeft het vaker een negatief sentiment.
12
Private and Confidential
% Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
108.40
35.00
10.70
7.00
55.70
3.26
Radisson Blu Astrid Hotel Hotel Julien
111.40 64.00
31.40 20.70
15.20 6.30
5.70 2.70
59.10 34.20
2.07 3.29
Leopold Hotel Antwerp
86.20
16.30
8.40
3.40
58.10
1.94
Crowne Plaza Antwerpen
34.40
5.00
8.60
2.70
18.10
0.58
Table 3: Locatie 4.3.2
Internet
Met betrekking tot het onderwerp internet, blijkt dat dit onderwerp minder vaak wordt genoemd dan locatie (Tabel 4). Het Hilton hotel scoort hier negatief (ook in vergelijking met de andere hotels is de ratio tussen positieve en negatieve hotels het laagste bij het Hilton hotel). Dit is een aandachtspunt voor het Hilton hotel. % Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
25.40
1.20
9.10
1.20
14.00
0.13
Radisson Blu Astrid Hotel
22.30
4.90
2.30
0.80
14.40
2.17
Hotel Julien
12.60
1.80
0.90
0.00
9.90
2.00
Leopold Hotel Antwerp
34.00
2.50
10.30
0.00
21.20
0.24
Crowne Plaza Antwerpen
29.40
1.80
7.20
3.60
16.70
0.25
Table 4: Internet 4.3.3
Badkamer
Het onderwerp badkamer wordt vaak genoemd in reviews van alle hotels (Tabel 5). Als dit onderwerp wordt genoemd is het vaker negatief dan positief, met uitzondering voor hotel Julien. Voor het Hilton hotel moet worden nagegaan wat er met betrekking tot badkamers slecht bevonden wordt door klanten. % Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
90.40
10.50
18.40
9.30
52.20
0.57
Radisson Blu Astrid Hotel
95.50
10.20
21.60
2.30
61.40
0.47
Hotel Julien
99.10
12.60
9.00
9.90
67.60
1.40
Leopold Hotel Antwerp
99.00
4.90
23.60
2.50
68.00
0.21
Crowne Plaza Antwerpen
71.00
7.20
19.50
5.00
39.40
0.37
Table 5: Bathroom 4.3.4
Receptie
Het onderwerp receptie wordt minder vaak genoemd in reviews (Tabel 6). Echter, als het wordt genoemd heeft het bij elk hotel een grotere kans om in een negatieve zin voor te komen dan in een positieve. Vooral voor het Leopold hotel en het Crowne Plaza hotel lijkt dit een probleem te zijn, omdat het onderwerp receptie daar vaker voorkomt in reviews dan bij de andere hotels.
13
Private and Confidential
% Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
21.90
1.90
4.70
0.20
15.20
0.40
Radisson Blu Astrid Hotel Hotel Julien
28.80 38.70
1.50 3.60
8.30 6.30
1.10 0.90
17.80 27.90
0.18 0.57
Leopold Hotel Antwerp
56.20
2.00
11.80
2.50
39.90
0.17
Crowne Plaza Antwerpen
47.10
4.10
11.30
2.70
29.00
0.36
Table 6: Reception 4.3.5
Eten
Eten lijkt voor veel klanten een erg belangrijk onderwerp te zijn, omdat het zoveel voorkomt in reviews (Tabel 7). Als het wordt genoemd is het meestal positief, zeker voor hotel Julien en in mindere mate ook voor het Hilton hotel. Verder onderzoek kan uitwijzen wat er positief wordt ervaren door klanten, zodat deze sterkte kan worden behouden. Hoewel het percentage van de reviews waarin eten negatief voorkomt veel kleiner is dan het percentage waarin het positief voorkomt, kan ook een analyse worden gedaan wat er precies negatief wordt ervaren. % Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
403.30
183.40
42.00
43.80
134.00
4.37
Radisson Blu Astrid Hotel
328.80
116.70
39.40
18.60
154.20
2.96
Hotel Julien
300.00
134.20
9.90
18.00
137.80
13.55
Leopold Hotel Antwerp
274.90
85.20
34.50
20.20
135.00
2.47
Crowne Plaza Antwerpen
313.60
79.20
59.70
28.50
146.20
1.33
Table 7: Eten 4.3.6
Staff en Service
Het onderwerp staff en service komt vaak voor in reviews en kan zowel refereren naar een positieve of negatieve zin (Tabel 8). Opvallend is dat hotel Julien een hogere kans heeft om positief te scoren op dit onderwerp (hogere positieve-negatieve ratio) en dat het Crowne Plaza hotel een hogere kans heeft om negatief te scoren op dit onderwerp in vergelijking met de andere hotels. % Reviews
% Pos
% Neg
% Mixed
% Neutral
Ratio Pos - Neg
Hilton Antwerp Hotel
176.50
23.50
32.20
8.60
112.10
0.73
Radisson Blu Astrid Hotel
134.50
15.50
17.40
3.00
98.50
0.89
Hotel Julien
155.90
32.40
18.00
1.80
103.60
1.80
Leopold Hotel Antwerp
171.40
11.80
28.60
6.90
124.10
0.41
Crowne Plaza Antwerpen
162.00
12.20
43.00
5.00
101.80
0.28
Table 8: Staff & Service 4.3.7
Lawaai
Lawaai komt niet vaak voor in reviews. Indien het voorkomt is het logischerwijs vaak in een negatieve context. Voor het Crowne Plaza hotel loont het om dit onderwerp verder te analyseren.
14
Private and Confidential
Hilton Antwerp Hotel Radisson Blu Astrid Hotel Hotel Julien Leopold Hotel Antwerp Crowne Plaza Antwerpen
% Reviews 6.80 8.30 17.10 16.30 30.80
% Pos 0.00 0.40 1.80 0.50 1.80
% Neg 3.00 3.40 9.90 9.40 10.90
% Mixed 0.20 0.40 0.00 0.50 3.20
% Neutral 3.50 4.20 5.40 5.90 14.90
Ratio Pos - Neg 0.00 0.11 0.18 0.05 0.17
Table 9: Noise
5
Samenvatting Resultaten • Het aantal top-beoordelingen (5 sterren) is het hoogste bij hotel Julien, gevolgd door Hilton, Radisson Blue en Leopols en als laatste Crowne Plaza • Hilton en hotel Julien scoren erg hoog op de categorie locatie • Het aantal ondermaatse beoordelingen (1, 2 of 3-sterren beoordeling) is het hoogste voor Crowne Plaza • Hotel Julien heeft weinig ondermaatse beoordelingen • Betreffende Crowne Plaza: bezoekerscategorie family en friends kennen vaker een top- score toe aan de categorie service • Betreffende Hilton: bezoekerscategorie solo kent vaker een top-score toe aan de categorie slaapkwaliteit en locatie • Betreffende Julien: een klein hotel met weinig reviews, waardoor een splitsing naar bezoekerstype enkel mogelijk is voor het type business en couple • Betreffende Leopold: bezoekerstype family kent minder top-scores toe aan de categorie locatie, terwijl de categorie friends meer top-scores toekent aan alle categorien • Betreffende Radisson Blue Astrid Hotel: de categorie friends kent relatief meer top-scores toe op alle categorien, tewrijl de categorie business relatief weinig topscores toekent op de categorie waarde • Trend in de tijd voor Hilton, hotel Julien en Crowne Plaza is vrij constant, terwijl deze voor Radisson Blue negatief is en voor Leopold positief. Een verdere analyse (tekstanalyse) voor alle hotels is nodig om de oorzaak van deze constante, negatieve en positieve trends te achterhalen
15
Private and Confidential
6
Overzicht
Met betrekking van Hotel reviews kan Dexam u ondersteunen bij zowel analyse van uw hotel als een diepgaande concurrentenanalyse: • Aantal reviews met een 1, 2, 3, 4 of 5-sterren beoordeling op de categorien: alles, waarde, locatie, slaapkwaliteit, kamers, reinheid en service. – Geeft u inzicht hoe uw hotel scoort, met name in vergelijking met uw concurrenten – Geeft u inzicht hoe uw hotel op elk van de categorien scoort, welke categorien goed zijn en welke verdere aandacht nodig hebben • Aantal reviews met een 1,2,3,4 of 5-sterren beoordeling op de categorien en per bezoekerstype – Geeft u dieper inzicht of er verschillen zijn tussen de verschillende bezoekerstypes • Verdeling tussen 1,2,3,4 of 5-sterren reviews per hotel en per bezoekerstype – Geeft u een grafisch inzicht of verschillende soorten bezoekerstypes uw hotel anders beoordelen • Historische ontwikkeling van het aantal reviews – Geeft u inzicht of er over tijd meer bezoekers uw hotel beoordelen en of een tekstanalyse zinvol is • Historische ontwikkeling van de score op de categorien – Geeft u inzicht of bepaalde categorien in de loop van de tijd slechter of beter zijn geworden en of er een tekstanalyse op bepaalde categorien moet worden gedaan • Tekstanalyse van bepaalde worden in reviews – Geeft u inzicht welke woorden klanten positief en negatief aan uw hotel vinden • Diepgaande tekstanalyse – Specificeert voor u verder wat er aan bepaalde onderwerpen goed of niet goed is (bijv: gebreken in het onderwerp badkamer) • Bovenstaande is ook mogelijk voor uw concurrenten – Geeft u inzicht wat er goed en slecht wordt bevonden bij uw concurrenten zodat u hier op kunt inspelen
16
Private and Confidential