Tableau in kiri, gaske?

Lokacin bayar da rahoto a cikin Excel yana ɓacewa da sauri - yanayin zuwa ga kayan aikin da suka dace don gabatarwa da nazarin bayanai yana bayyane a duk yankuna. Mun daɗe muna tattaunawa a cikin gida don ƙididdige rahoto kuma mun zaɓi tsarin hangen nesa na Tableau da tsarin nazarin ayyukan kai. Alexander Bezugly, shugaban sashen nazari da bayar da rahoto na M.Video-Eldorado Group, yayi magana game da kwarewa da sakamakon gina dashboard na fama.

Zan ce nan da nan cewa ba duk abin da aka shirya ba ya tabbata, amma ƙwarewar ta kasance mai ban sha'awa, ina fatan zai kasance da amfani a gare ku kuma. Kuma idan wani yana da ra'ayi kan yadda za a yi shi mafi kyau, zan yi matukar godiya da shawararku da ra'ayoyinku.

A ƙasa yanke shine game da abin da muka ci karo da abin da muka koya game da shi.

A ina muka fara?

M.Video-Eldorado yana da ingantaccen tsarin bayanai: bayanan da aka tsara tare da zurfin ajiyar da ake buƙata da adadi mai yawa na ƙayyadaddun rahotanni (duba ƙarin cikakkun bayanai). wannan labarin). Daga waɗannan, manazarta suna yin ko dai tebur na pivot ko tsararrun wasiƙun labarai a cikin Excel, ko kyawawan gabatarwar PowerPoint don masu amfani da ƙarshe.

Kimanin shekaru biyu da suka gabata, maimakon rahotannin ƙayyadaddun tsari, mun fara ƙirƙirar rahotanni na nazari a cikin SAP Analysis (wani ƙari na Excel, ainihin teburin pivot akan injin OLAP). Amma wannan kayan aikin bai iya biyan bukatun duk masu amfani ba; yawancin sun ci gaba da amfani da bayanan da manazarta ke sarrafa su.

Masu amfani da mu na ƙarshe sun faɗi cikin rukuni uku:

Babban gudanarwa. Yana buƙatar bayani a cikin ingantaccen gabatarwa kuma a bayyane hanyar fahimta.

Gudanar da tsakiya, ci-gaba masu amfani. Mai sha'awar binciken bayanai kuma yana iya gina rahotanni da kansa idan akwai kayan aikin. Sun zama manyan masu amfani da rahotanni na nazari a cikin SAP Analysis.

Masu amfani da yawa. Ba su da sha'awar nazarin bayanai da kansu; suna amfani da rahotanni tare da iyakacin yanci, a cikin tsarin wasiƙun labarai da tebur mai mahimmanci a cikin Excel.

Tunaninmu shine don rufe bukatun duk masu amfani kuma mu ba su kayan aiki guda ɗaya, dacewa. Mun yanke shawarar farawa da babban gudanarwa. Suna buƙatar dashboards masu sauƙin amfani don tantance mahimman sakamakon kasuwanci. Don haka, mun fara da Tableau kuma mun fara zaɓar hanyoyi guda biyu: dillalai da alamun tallace-tallace na kan layi tare da iyakataccen zurfin bincike da faɗin bincike, wanda zai rufe kusan 80% na bayanan da manyan gudanarwa suka nema.

Tun da masu amfani da dashboards sun kasance babban gudanarwa, wani ƙarin KPI na samfurin ya bayyana - saurin amsawa. Babu wanda zai jira 20-30 seconds don sabunta bayanan. Ya kamata a yi kewayawa a cikin daƙiƙa 4-5, ko mafi kyau tukuna, an yi nan take. Kuma mu, kash, mun kasa cimma wannan.

Wannan shi ne yadda tsarin babban dashboard ɗin mu ya yi kama:

Mahimmin ra'ayi shine haɗa manyan direbobin KPI, waɗanda akwai 19 a cikin duka, a hagu kuma suna gabatar da ƙarfinsu da rushewar su ta hanyar manyan halayen dama. Ayyukan yana da sauƙi, hangen nesa yana da ma'ana da fahimta, har sai kun nutse cikin cikakkun bayanai.

Dalla-dalla 1. Girman bayanai

Babban teburin mu na tallace-tallace na shekara yana ɗaukar kusan layuka miliyan 300. Tun da yake wajibi ne a yi la'akari da abubuwan da suka faru na bara da kuma shekarar da ta gabata, yawan adadin bayanai akan ainihin tallace-tallace kawai shine kusan layin biliyan 1. Ana kuma adana bayanai kan bayanan da aka tsara da kuma toshe tallace-tallacen kan layi daban. Saboda haka, ko da yake mun yi amfani da columnar in-memory DB SAP HANA, gudun tambaya tare da zaɓin duk alamomi na mako guda daga ajiya na yanzu a kan tashi ya kasance game da 15-20 seconds. Maganin wannan matsala yana nuna kanta - ƙarin kayan aiki na bayanai. Amma kuma yana da ramummuka, ƙarin game da su a ƙasa.

Dalla-dalla 2. Alamomin da ba ƙari ba

Yawancin KPI namu suna daura da adadin rasit. Kuma wannan alamar tana wakiltar COUNT DISTINCT na adadin layuka (cack headers) kuma yana nuna adadi daban-daban dangane da halayen da aka zaɓa. Misali, yadda yakamata a lissafta wannan mai nuna alama da abin da aka samo ta:

Don daidaita lissafin ku, kuna iya:

• Yi lissafin irin waɗannan alamun akan tashi a cikin ajiya;
• Yi lissafi akan ɗaukacin adadin bayanai a cikin Tableau, watau. akan buƙata a cikin Tableau, samar da duk bayanai bisa ga zaɓaɓɓun tacewa a cikin ƙimar matsayin karɓa;
• Ƙirƙirar nunin kayan aiki wanda a cikinsa za a ƙididdige duk masu nuni a cikin duk zaɓuɓɓukan samfur waɗanda ke ba da sakamako daban-daban waɗanda ba ƙari ba.

A bayyane yake cewa a cikin misali UTE1 da UTE2 halayen kayan aiki ne da ke wakiltar tsarin samfurin. Wannan ba wani abu ba ne; gudanarwa a cikin kamfani yana faruwa ta hanyarsa, saboda Manajoji daban-daban suna da alhakin ƙungiyoyin samfura daban-daban. Muna da sauye-sauye na duniya da yawa game da wannan matsayi, lokacin da duk matakan suka canza, lokacin da aka sabunta alaƙa, da canje-canjen ma'ana akai-akai, lokacin da ƙungiya ɗaya ta ƙaura daga wannan kulli zuwa wancan. A cikin rahotanni na al'ada, duk wannan ana ƙididdige su a kan tashi daga halayen kayan aiki; a cikin yanayin samar da wannan bayanan, ya zama dole don samar da wata hanya don bibiyar irin waɗannan canje-canje kuma ta atomatik sake shigar da bayanan tarihi. Aiki mara nauyi sosai.

Detail 3. Data kwatanta

Wannan batu yayi kama da na baya. Maganar ƙasa ita ce, lokacin da ake nazarin kamfani, al'ada ne don samar da matakai da yawa na kwatanta da lokacin da ya gabata:

Kwatanta da lokacin da ya gabata (rana zuwa rana, mako zuwa mako, wata zuwa wata)

A cikin wannan kwatancen, ana ɗauka cewa ya danganta da lokacin da mai amfani ya zaɓa (misali, sati na 33 na shekara), yakamata mu nuna ƙarfin gwiwa ta mako na 32; idan muka zaɓi bayanai na wata ɗaya, misali, Mayu. , to wannan kwatancen zai nuna abubuwan da ke faruwa a watan Afrilu.

Kwatanta da bara

Babban abin lura anan shine idan aka kwatanta da rana da mako, ba kuna shan rana ɗaya ta bara, watau. ba za ku iya sanya shekarar da muke ciki kawai ba. Dole ne ku kalli ranar mako da kuke kwatanta. Lokacin kwatanta watanni, akasin haka, kuna buƙatar ɗaukar daidai wannan ranar kalanda na bara. Hakanan akwai nuances tare da shekarun tsalle. A cikin wuraren ajiyar asali, ana rarraba duk bayanai da rana; babu wasu filaye daban-daban masu makonni, watanni, ko shekaru. Sabili da haka, don samun cikakken ɓangaren giciye a cikin panel, kuna buƙatar ƙidaya ba lokaci ɗaya ba, alal misali a mako, amma makonni 4, sannan kwatanta waɗannan bayanan, kuyi la'akari da kuzari, karkacewa. Don haka, ana iya aiwatar da wannan dabarar don samar da kwatancen a cikin motsin rai ko dai a cikin Tableau ko a gefen shago. Ee, kuma ba shakka mun sani kuma mun yi tunani game da waɗannan cikakkun bayanai a matakin ƙira, amma yana da wahala a hango tasirin su akan aikin dashboard na ƙarshe.

Lokacin aiwatar da dashboard, mun bi doguwar hanyar Agile. Ayyukanmu shine samar da kayan aiki tare da mahimman bayanai don gwaji da sauri. Saboda haka, mun shiga sprints kuma mun fara daga rage girman aiki a gefen ajiyar yanzu.

Sashe na 1: Imani a Tableau

Don sauƙaƙe tallafin IT da aiwatar da canje-canje cikin sauri, mun yanke shawarar yin dabaru don ƙididdige alamomin da ba ƙari ba da kwatanta lokutan baya a cikin Tableau.

Mataki na 1. Komai yana Live, babu gyare-gyaren taga.

A wannan mataki, mun haɗa Tableau zuwa manyan shaguna na yanzu kuma mun yanke shawarar ganin yadda za a ƙididdige adadin rasit na shekara guda.

Sakamako:

Amsar ta kasance mai ban tsoro - minti 20. Canja wurin bayanai akan hanyar sadarwa, babban nauyi akan Tableau. Mun fahimci cewa ana buƙatar aiwatar da dabaru tare da alamomin da ba ƙari ba akan HANA. Wannan bai tsoratar da mu da yawa ba, mun riga mun sami irin wannan gogewa tare da BO da Analysis kuma mun san yadda ake gina abubuwan nunawa cikin sauri a HANA waɗanda ke samar da daidaitattun ƙididdiga marasa ƙari. Yanzu abin da ya rage shi ne a daidaita su zuwa Tableau.

Mataki na 2. Muna kunna lokuta na nuni, babu kayan aiki, duk abin da ke tashi.

Mun ƙirƙiri wani sabon nunin nuni wanda ya samar da bayanan da ake buƙata don TABLEAU akan tashi. Gabaɗaya, mun sami sakamako mai kyau; mun rage lokacin samar da duk alamomi a cikin mako guda zuwa 9-10 seconds. Kuma da gaske muna tsammanin cewa a cikin Tableau lokacin amsawar dashboard zai kasance 20-30 seconds a farkon buɗewa sannan kuma saboda cache daga 10 zuwa 12, wanda gabaɗaya zai dace da mu.

Sakamako:

Buɗe dashboard na farko: mintuna 4-5
Kowane danna: 3-4 mintuna
Babu wanda ya yi tsammanin irin wannan ƙarin haɓaka a cikin aikin kantin sayar da kayayyaki.

Part 2. Nutse cikin Tableau

Mataki na 1. Taswirar aikin bincike da saurin daidaitawa

Mun fara nazarin inda Tableau ke ciyar da mafi yawan lokutan sa. Kuma akwai kayan aiki masu kyau don wannan, wanda, ba shakka, ƙari ne na Tableau. Babban matsalar da muka gano ita ce, tambayoyin SQL masu sarkakiya da Tableau ke ginawa. An danganta su da farko da:

- canja wurin bayanai. Tun da Tableau ba shi da kayan aikin watsa bayanai, don gina gefen hagu na dashboard tare da cikakken wakilcin duk KPIs, dole ne mu ƙirƙiri tebur ta amfani da akwati. Girman tambayoyin SQL a cikin bayanan ya kai haruffa 120.

- zabin lokaci. Irin wannan tambaya a matakin ma'ajin bayanai ya ɗauki ƙarin lokaci don haɗawa fiye da aiwatar da:

Wadancan. aikace-aikacen aikace-aikacen 12 seconds + 5 seconds kisa.

Mun yanke shawarar sauƙaƙa dabarun lissafi a gefen Tableau kuma mu matsar da wani ɓangaren lissafin zuwa wurin ajiya da matakin bayanai. Wannan ya kawo sakamako mai kyau.

Na farko, mun yi juzu'i a kan tashi, mun yi shi ta hanyar cikakken haɗin waje a matakin ƙarshe na lissafin VIEW, bisa ga wannan hanyar da aka bayyana akan wiki. Transpose - Wikipedia, encyclopedia na kyauta и Matrix na farko - Wikipedia, encyclopedia na kyauta.

Wato, mun yi tebur saitin - matrix transposition (21x21) kuma mun karɓi duk alamun a cikin rugujewar layi-bi-jere.

Ya kasance:


Ya zama:


Kusan ba a ɓata lokaci akan jujjuyawar bayanan da kanta. An ci gaba da aiwatar da buƙatun ga dukkan alamu na mako a cikin kusan daƙiƙa 10. Amma a daya bangaren kuma, an yi asarar sassauci ta fuskar gina dashboard bisa wani takamaiman nuni, watau. ga gefen dama na dashboard inda aka gabatar da kuzari da cikakkun bayanai na takamaiman nuni, a baya yanayin nuni yayi aiki a cikin daƙiƙa 1-3, saboda Bukatar ta dogara ne akan mai nuna alama guda ɗaya, kuma a yanzu ma'aunin bayanai koyaushe yana zaɓar duk alamomi kuma ta tace sakamakon kafin a mayar da sakamakon zuwa Tableau.

Sakamakon haka, saurin dashboard ɗin ya ragu da kusan sau 3.

Sakamako:

1. 5 sec – tantance dashboards, abubuwan gani
2. 15-20 seconds - shirye-shirye don tattara tambayoyin tare da yin ƙididdigewa kafin a cikin Tableau
3. 35-45 sec - harhada tambayoyin SQL da aiwatar da aikinsu na layi daya a Hana
4. 5 sec - sakamakon sarrafawa, rarrabawa, sake ƙididdige abubuwan gani a cikin Tableau
5. Tabbas, irin wannan sakamakon bai dace da kasuwancin ba, kuma mun ci gaba da ingantawa.

Mataki na 2. Mahimman tunani a cikin Tableau, cikakke kayan aiki

Mun fahimci cewa ba shi yiwuwa a gina dashboard tare da lokacin amsawa na daƙiƙa da yawa a kan kantin sayar da kayayyaki wanda ke gudana na daƙiƙa 10, kuma mun yi la'akari da zaɓuɓɓuka don ƙirƙirar bayanai a gefen bayanan musamman don dashboard ɗin da ake buƙata. Amma mun ci karo da wata matsala ta duniya da aka kwatanta a sama - alamomi marasa ƙari. Ba mu sami damar tabbatar da cewa lokacin da ake canza matattara ko faifai ba, Tableau a hankali ya canza tsakanin shagunan shagunan daban-daban da matakan da aka riga aka tsara don jerin samfuran samfuri daban-daban (a misali, tambayoyin uku ba tare da UTE ba, tare da UTE1 da UTE2 suna haifar da sakamako daban-daban). Saboda haka, mun yanke shawarar sauƙaƙa dashboard, watsi da tsarin samfurin a cikin dashboard kuma mu ga yadda sauri zai iya kasancewa a cikin sassauƙan sigar.

Don haka, a wannan mataki na ƙarshe, mun haɗa wani ma'aji na dabam inda muka ƙara duk KPI a cikin sigar da aka canza. A gefen ma'ajin bayanai, duk wani buƙatu ga irin wannan ma'ajiyar ana sarrafa shi a cikin 0,1 - 0,3 seconds. A cikin dashboard mun sami sakamako masu zuwa:

Buɗewar farko: 8-10 seconds
Kowane danna: 6-7 seconds

Lokacin da Tableau ya kashe ya ƙunshi:

1. 0,3 dakika - nazarin dashboard da harhada tambayoyin SQL
2. 1,5-3 seconds. - aiwatar da tambayoyin SQL a cikin Hana don manyan abubuwan gani (yana gudana a layi daya da mataki na 1)
3. 1,5-2 seconds. - ma'ana, sake lissafin abubuwan gani
4. 1,3 dakika - aiwatar da ƙarin tambayoyin SQL don samun ƙimar tacewa masu dacewa (Brand, Division, City, Store), tantance sakamakon.

Don takaita shi a takaice

Muna son kayan aikin Tableau daga hangen nesa. A mataki na samfuri, mun yi la'akari da abubuwa daban-daban na gani kuma mun same su duka a cikin ɗakunan karatu, gami da hadaddun rarrabuwar matakan matakai da ruwa mai direba da yawa.

Yayin aiwatar da dashboards tare da mahimman alamun tallace-tallace, mun ci karo da matsalolin aiki waɗanda har yanzu ba mu sami damar shawo kansu ba. Mun shafe fiye da watanni biyu kuma mun karɓi dashboard ɗin da bai cika aiki ba, saurin amsawa yana kan hanyar karɓuwa. Kuma mun yanke shawara kan kanmu:

1. Tableau ba zai iya aiki tare da adadi mai yawa na bayanai ba. Idan a cikin ƙirar bayanan asali kuna da fiye da 10 GB na bayanai (kimanin layuka miliyan 200 X 50), to, dashboard ɗin zai ragu da gaske - daga 10 seconds zuwa mintuna da yawa don kowane dannawa. Mun yi gwaji tare da haɗin kai-tsaye da cirewa. Gudun aiki yana kwatankwacinsa.
2. Iyakance lokacin amfani da ma'ajiya da yawa (saitin bayanai). Babu wata hanya ta nuna alakar da ke tsakanin bayanan bayanan ta amfani da daidaitattun hanyoyi. Idan kuna amfani da wuraren aiki don haɗa saitin bayanai, wannan zai shafi aiki sosai. A cikin yanayinmu, mun yi la'akari da zaɓi na kayan aiki na bayanai a cikin kowane ɓangaren ra'ayi da ake buƙata da yin sauyawa akan waɗannan bayanan da aka yi amfani da su yayin adana abubuwan da aka zaɓa a baya - wannan ya zama ba zai yiwu a yi ba a cikin Tableau.
3. Ba zai yiwu a yi sigogi masu ƙarfi a cikin Tableau ba. Ba za ku iya cika ma'aunin da ake amfani da shi don tace saitin bayanai a cikin tsantsa ko yayin haɗin kai tare da sakamakon wani zaɓi daga saitin bayanai ko sakamakon wata tambayar SQL ba, kawai shigar da mai amfani na asali ko na dindindin.
4. Iyakance masu alaƙa da gina dashboard tare da OLAP|Abubuwa na PivotTable.
   A cikin MSTR, SAP SAC, SAP Analysis, idan kun ƙara saitin bayanai zuwa rahoto, to duk abubuwan da ke kan sa suna da alaƙa da juna ta tsohuwa. Tableau bashi da wannan; dole ne a saita haɗin da hannu. Wataƙila wannan ya fi sassauƙa, amma ga dukkan dashboards ɗin mu wannan wajibi ne ga abubuwa - don haka wannan ƙarin farashin aiki ne. Haka kuma, idan ka yi matattarar da ke da alaƙa ta yadda, alal misali, lokacin da ake tace yanki, jerin biranen ya iyakance ga biranen wannan yanki kawai, nan da nan za ku ƙare tare da ci gaba da tambayar ma'ajin bayanai ko Extract, wanda a bayyane yake yana rage gudu. dashboard.
5. Iyakoki a cikin ayyuka. Ba za a iya yin sauye-sauyen taro ko dai akan abin da aka cire ko, MUSAMMAN, akan saitin bayanai daga Live-connecta. Ana iya yin wannan ta hanyar Prep Tableau, amma ƙarin aiki ne da wani kayan aiki don koyo da kulawa. Misali, ba za ku iya tura bayanai ko haɗa su da kanta ba. Abin da ke rufe ta hanyar sauye-sauye a kan ginshiƙai ko filayen, wanda dole ne a zaɓa ta hanyar harka ko kuma idan, kuma wannan yana haifar da tambayoyin SQL masu sarƙaƙƙiya, wanda ma'ajin bayanai ke ciyar da mafi yawan lokutansa wajen tattara rubutun tambaya. Dole ne a warware waɗannan rashin daidaituwa na kayan aiki a matakin nuni, wanda ke haifar da ƙarin hadaddun ajiya, ƙarin zazzagewa da canji.

Ba mu daina kan Tableau ba. Amma ba mu la'akari da Tableau a matsayin kayan aiki da ke iya gina dashboards masana'antu da kayan aiki da abin da za a maye gurbin da kuma ƙididdige tsarin tsarin rahoton kamfani gaba ɗaya.

Yanzu muna haɓaka dashboard irin wannan a cikin wani kayan aiki kuma, a lokaci guda, muna ƙoƙarin sake fasalin gine-ginen dashboard a cikin Tableau don sauƙaƙe shi har ma. Idan al'umma suna sha'awar, za mu gaya muku sakamakon.

Muna kuma jiran ra'ayoyinku ko shawarwarin ku kan yadda a cikin Tabeau zaku iya gina dashboards masu sauri akan manyan tarin bayanai, saboda muna da gidan yanar gizon da ke da bayanai da yawa fiye da na dillali.

source: www.habr.com