WillLuckinshakeStarbucks’industrystatus?
Lastmonth,LuckinCoffee’ssaucelattebecamepopular,puttingLuckininthespotlightagain.Thelasttimewaswhenitcommittedfinancialfraud.
Thedomesticcoffeemarkethasbeenboominginrecentyears,drivingtherapiddevelopmentofmanycoffeebrandsincludingLuckin.From2013to2023,China’spercapitacoffeeconsumptionisexpectedtoincreaseby238%.Thetotalnumberofcoffeestoresinthecountrynowexceeds100,000.Andthenumberisgrowingbytensofthousandseveryyear.
TheriseofLuckinCoffeeremindsusofStarbucks,thebenchmarkinthecoffeeindustry.Starbuckshasbeenalmostsynonymouswithcoffeeinthepasttenyearsandisalsothelifestyleofurbanwhite-collarworkers.
WhatishappeningnowisthatwhereverthereisaStarbucksstore,thereisalmostaLuckinstorewithinafewhundredmeters,andsomeareevensurroundedbytwoorthree.
ThefollowingusesvisualdashboardsandPythondataanalysistocomparethedifferencesandcorrelationsbetweenStarbucksandLuckinstoresintermsofquantity,regionaldistribution.
Therearetwomainfindings:
1.StarbucksismoreconcentratedineconomicallydevelopedcoastalareassuchastheYangtzeRiverDelta,PearlRiverDelta,andBeijing-Tianjin-Hebei,especiallyfirst-andsecond-tiercities.LuckinismoredispersedthanStarbucksandhasstoresinmanythird-andfourth-tiercitiesandbelow.
2.LuckinstorelocationsareconcentratedaroundStarbucks.Datashowsthatwithinaradiusof500meters,thereareanaverageof0.6LuckinstoresaroundeachStarbucksstorenationwide.
Preparationphase
ThetoolsusedinthisanalysistaskincludeNextDataRobot,Python,andshapely.
XiamiaoDataRobotisaclouddataplatformthatintegratesdatasets,datacleaning,dataanalysis,datavisualizationandbillboardconstruction.TheStarbucksandLuckinstoredatasetsusedinthisanalysisareallstoredinXiamiaoDataRobot.
Wewilldevelopdatadashboardsbasedondatasets,andalsouseAPIinterfacestodirectlycallPythonfordataanalysisandvisualization.Pythonisusedtoconnecttothedatainterfaceoftherobotandprocessandanalyzethedata.
Shapelyisathird-partylibraryforPython,usedtoprocesslatitudeandlongitudedata,andcandeterminethedistanceandinclusionrelationshipbetweendifferentgeographicalcoordinates.
Dataset
BecauseweneedtocompareandanalyzethenumberandlocationofStarbucksandLuckinstores,themainfieldsofthedatasetincludestorename,longitude,latitude,andcity.
?
Note:Thedatasetisin2022,andthereisaquantitativeerrorofabout20%.
?
NationalStarbuckscoffeestoredataset:
Bothdatasetsarestoredonthenextdatarobotplatform.Youcandirectlyviewandusethedatasetsthroughthedataview,whichwewillusetobuildadatadashboardlater.
BecausePythonisneededtoprocessthedatalater,thedataneedstobeobtainedthroughtheAPIdatainterface.Itisveryconvenienttooperateandcanbesavedforlateruse.
importrequests
headers={"x-token":"yourauthenticationtoken"}
response=requests.get("http://app.chafer.nexadata.cn/openapi/v1/sheet/sht22nId5uouP2/records?size=1&page=1",headers=headers)
print(response.json())
Buildananalysisdashboard
ItisrelativelysimpletobuildadashboardontheNextDataRobot.First,createaprocesstaskandselecttwodataviews:StarbucksandLuckin.
ThencreateaKanbanboardandeditthedesignchart,whichissimilartowhatweusuallydoonBIsoftware.
Therearemorethanadozenchartformshere,whichcanbasicallymeetmostvisualizationscenarios.
ComparisonofthenumberofStarbucksandLuckinstoresnationwide
Asofthedatasettime(2022),thenumberofStarbucksstoresnationwideisexpectedtobe4,442,andthenumberofLuckinCoffeestoresnationwideisexpectedtobe3,904.Starbuckshas14%morestoresthanLuckin.
Judgingfromthemagnitude,thetwoareveryclose,andLuckinisexpandingitsstoresataterrifyinggrowthrate.Takingthebusinessdistrictnearmyhomeasanexample,therewasonlyoneLuckinlastyear,andtherearethreethisyear.
StarbucksismoredemandingthanRuixingintermsoflocation,storeopeningcost,storearea,andnumberofstoreemployees.Ruixingspecializesintakeout+takeout.ThisisalsothereasonwhyRuixingcanexpandrapidlyinadditiontomarketdemandfactors.
Starbucksisdistributedinthetop20citiesacrossthecountry
ThetopfivecitieswiththenumberofStarbucksstoresare:Shanghai,Beijing,Hangzhou,Shenzhen,andGuangzhou.Amongthetop20cities,thereare6intheYangtzeRiverDelta,5inthePearlRiverDelta,and2intheBeijing-Tianjin-Hebeiregion.
ThenumberofStarbucksstoresinShanghaiis668,whichistwiceasmanyasthesecond-placeBeijing.Atthesametime,ShanghaiisalsothecitywiththelargestnumberofStarbucksstoresintheworld.ItseemsthatthepeopleofShanghaihaveawell-deservedloveforcoffee.
ThenumberofStarbucksinHangzhouissecondonlytoShanghaiandBeijing,andhigherthanShenzhenandGuangzhou.Internetande-commercepractitionersinHangzhoualsopreferStarbucks.
LuckinisdistributedinTop20citiesacrossthecountry
ThetopfivecitieswiththenumberofLuckinstoresare:Shanghai,Beijing,Guangzhou,Shenzhen,andHangzhou.TheyarethesameasthetopfivecitiesforStarbucks,buttheorderingisslightlydifferent.
Amongthetop20cities,thereare6intheYangtzeRiverDelta,2inthePearlRiverDelta,and2intheBeijing-Tianjin-Hebeiregion.
Starbucksismainlyconcentratedinfirst-andsecond-tiercoastalcities,whileLuckinisrapidlyoccupyingthemarketininlandcities.Luckin’stop20citiesalreadyincludeHefei,Kunming,andZhengzhou,butthesethreeprovincialcapitalcitiesarenotinStarbucks’top20.
Therefore,thedistributionofLuckinstoresismoredispersedandnotoverlyconcentratedinfirst-tiercities.
Starbucksnationwidedistributionheatmap
ItcanalsobeseenfromtheheatmapofStarbucksstoresthattheredhigh-densityareasaremainlyconcentratedincoastalareas,whiletheinlandareasaredistributedinapoint-likemannerandarerelativelysparse.
Ruixing’snationwidedistributionheatmap
ThedistributionofLuckinstoresismoreeven.Inadditiontocoastalareas,therearealsomanystoresincentralChinasuchasHunan,Anhui,Hubei,andHunan.
StarbucksShanghaidistributionheatmap
Shanghaiisthecitywiththelargestcoffeeconsumptiondemandinthecountry.Let’stakealookatthedistributionofStarbucksstoresinShanghai.
Generallyspeaking,StarbucksstoresareconcentratedintheinnerringofShanghai,andarescatteredoutwardsindottedlines.Thefivemajornewcitiesinthesuburbs,PudongAirport,andHongqiaohubarealsorelativelyconcentratedareas.
LuckinShanghaidistributionheatmap
TheconcentrationofLuckinintheinnerringofShanghaiisnotasobviousasthatofStarbucks,andtheoverallnumberisalsomuchsmaller.
Pythondataanalysis
Previously,weanalyzedthedistributionofStarbucksandLuckinstoresacrossthecountrybybuildingavisualsignageontheNextMiaoRobot.Theregionaldifferencesarestillquiteobvious.
Let’sfurtheranalyzetherelationshipbetweenStarbucksandLuckinstores.WeknowthatLuckinCoffeeisarisingstar.ItissaidthatthelocationofmanystoresmainlydependsonwhetherthereisaStarbucksnearby.
So,onaverage,howmanyLuckinstoresaretherearoundeachStarbucksstorenationwide?Thistime,lookatthegatheringsituationofLuckinaroundStarbucksfroma500-meterradius.
WeusePythonanditsthird-partylibraryshapelytoprocessdata.Shapelyismainlyusedtoprocessgeographicalcoordinatedata.
Step1:Importtherequiredlibraries
#Importrelatedlibraries importpandasaspd importrequests importtime fromshapely.geometryimportPoint fromshapely.geometry.polygonimportPolygon
Step2:ExtractdatafromtheAPI
#ExtractStarbucksandLuckinstoredataandcallitthroughthenextrobotAPIcall
#ExtractStarbucksstoredata
headers={"x-token":"tk7a2980431688455e8976e4bad4d13d6a"}
starbucks_list=[]
foriinrange(1,10):
response_1=requests.get("http://app.chafer.nexadata.cn/openapi/v1/sheet/sht22nId5uouP2/records?size=500&page={0}".format(i),headers=headers)
starbucks=response_1.json()['data']['list']
starbucks=pd.DataFrame(starbucks)
time.sleep(1)
starbucks_list.append(starbucks)
starbucks=pd.concat(starbucks_list)
#ExtractLuckinstoredata
luckin_list=[]
forjinrange(1,9):
response_2=requests.get("http://app.chafer.nexadata.cn/openapi/v1/sheet/sht22nIeomVmYy/records?size=500&page={0}".format(j),headers=headers)
luckin=response_2.json()['data']['list']
luckin=pd.DataFrame(luckin)
luckin_list.append(luckin)
time.sleep(1)
luckin=pd.concat(luckin_list)
Step3:DeterminethenumberofLuckinstoreswithina500-meterradiusofaStarbucksstore
#DrawageographicalareawitharadiusofXXmetersbasedonthecoordinatesoftheStarbuckscoffeeshop defcircle(data,radius): #radiusrepresentsthearearadius #Givengeographicalcoordinates center_latitude=float(data['dimension']) center_longitude=float(data['longitude']) #Createacirculararea center_point=Point(center_longitude,center_latitude) circle=center_point.buffer(radius/111300) #Createpolygonarea polygon=Polygon(circle.exterior) returnpolygon #Constructcoordinatepointsbasedonlongitudeandlatitude defpoint(data): #Givengeographicalcoordinates center_latitude=float(data['dimension']) center_longitude=float(data['longitude']) #Createcoordinatepoints center_point=Point(center_longitude,center_latitude) returncenter_point #DeterminewhetherLuckinCoffeeShopiswithin500mofStarbucks defis_inside(data): polygon=data['Polygon'] #Determinewhetherthecoordinatesarewithinthearea n=0 luckin_city=luckin[luckin['city']==data['city']] forpointinluckin_city['Point']: is_inside=polygon.contains(point) #Printthejudgmentresult ifis_inside: n=n+1 returnn #Drawageographicalareawitharadiusof500metersbasedonthecoordinatesofStarbucksstores starbucks['Polygon']=starbucks.apply(circle,axis=1,args=(500,)) #ConstructcoordinatepointsbasedonthelongitudeandlatitudeofLuckinstores luckin['Point']=luckin.apply(point,axis=1) #DeterminewhetheraLuckinstoreiswithina500-meterradiusofaStarbucksstore starbucks['Luckin_numbers']=starbucks.apply(is_inside,axis=1)
Thedataafterprocessingisasfollows:
Step4:Analyzethedata
-
Withinaradiusof500meters,thereareanaverageof0.6LuckinstoresaroundeachStarbucksstorenationwide.
#Withinaradiusof500meters,thereareanaverageof0.6LuckinstoresaroundeachStarbucksstorenationwide. starbucks['Luckin_numbers'].mean()
Output:0.6
-
Withinaradiusof500meters,thereare7LuckinstoressurroundingthelargestnumberofStarbucksstoresinthecountry.
#Withinaradiusof500meters,thereare7LuckinstoresaroundthemostStarbucksstores starbucks['Luckin_numbers'].max()
Output:7
-
Withinaradiusof500meters,theaveragenumberofLuckinstoresaroundaStarbucksstoreisrankedbycity.ThelargestnumberisLinyiCity,withanaverageof1.8LuckinstoresaroundeachStarbucksstore.
#RankingofaveragenumberofLuckinstoresineachcitywithina500-meterradiusofStarbucksstores #Themostisthatthereareanaverageof1.8LuckinstoresaroundeachStarbucksstoreinLinyiCity city_list=[] forcityinpd.unique(starbucks['city']): avg_luckin_numbers=starbucks[starbucks['city']==city]['Luckin_numbers'].mean() starbucks_nums=starbucks[starbucks['city']==city]['name'].count() city_list.append([city,starbucks_nums,avg_luckin_numbers]) df=pd.DataFrame(city_list,columns=['city','starbucks_nums','avg_luckin_numbers']) df.sort_values(by=['avg_luckin_numbers'],axis=0,ascending=False)
Output:
ItseemsthatLuckindoeshaveanindissolublebondwithStarbucks.NowonderweseesomanyLuckinstoresaroundStarbucks.
Starbucksstoreshavedevelopedthehabitofdrinkingcoffeeamongsurroundingusers.Inotherwords,therearemanycoffeedrinkingusershere,soStarbucksjustcameheretoopenastore.ThenopeningLuckinCoffeenearStarbuckscanacquirealargewaveofpotentialusersatalowcost,despitecompetition.,it’sstillveryworthwhile.
Summary
WeusedvisualdashboardsandPythondataanalysistoshowtheregionaldistributionandcorrelationofStarbucksandLuckinCoffeestores.Infact,therearemanythingsworthanalyzingandexploring.
Forexample,insomecitiesthereareveryfeworalmostnoLuckinstoresaroundStarbucksstores.Whatisthereason?Isitapotentialopportunityorapittobeavoided?
Ifyouareinterested,youcangiveitatry.
Reprintedfrom:pythonbigdata
Ourmembershipcommunity WeChatconsultation:coderv5 Recommendedreading: GettingStarted:ThemostcomprehensiveproblemoflearningPythonfromscratch|LearnedPythonfor8monthsfromscratch|Practicalprojects|ThisistheshortcuttolearnPython
Essentialinformation:CrawlingDoubanshortreviews,themovie”TheNextUs”|AnalysisofthebestNBAplayersin38years|Fromhighlyanticipatedtowordofmouth!DetectiveTang3isdisappointing|WatchthenewLegendofHeavenandDragonSwordwithlaughter|Thekingoflanternriddleanswers|UsePythontomakeamassivesketchofyoungladies|Mission:Impossibleissopopular,Iusemachinelearningtomakeaminirecommendationsystemformovies
Fun:pinballgame|nine-squaregrid|beautifulflowers|twohundredlinesofPython”TiantianCoolRun”game!
AI:Arobotthatcanwritepoetry|Colorizepictures|Predictincome|Mission:Impossibleissopopular,Iusemachinelearningtomakeaminimovierecommendationsystem
Gadget:ConvertPdftoWord,easilyconverttablesandwatermarks!|Savehtmlwebpagestopdfwithoneclick!|GoodbyePDFextractioncharges!|Use90linesofcodetocreatethemostpowerfulPDFconverter,one-clickconversionofword,PPT,excel,markdown,andhtml|CreateaDingTalklow-priceticketreminder!｜60linesofcodemadeavoicewallpaperswitcherthatIcanwatcheveryday!|