Riba:
- Ikon yin bayanin yaduddukan bayanan sirri na kasuwanci (tushen, ETL, ma'ajin bayanai, rahoto) da ma'anar ma'anar ma'auni na kasuwanci daidai (KPIs).
- Ikon yin amfani da hankali na wucin gadi don samar da ma'anar awo, SQL daftarin aiki da bayar da rahoto da bayar da sakamakon tare da ainihin bayanai
- Ƙarfin gane haɗari na rikice-rikice-haɓaka dangantaka da ma'auni na yaudara a cikin bayanan bincike na goyon bayan AI.
Harkokin Kasuwanci (BI) shine horo wanda ke tattara bayanan da aka tarwatsa na kungiyar, ya sa shi a shirye don bincike, kuma yana samar da bayanai masu goyon bayan yanke shawara daga wannan bayanan. Ga ƙwararren MIS, BI shine Layer inda "bayani ya juya zuwa yanke shawara." Rubutun oda kawai ba su da ma'ana; amma "a wanne yanki ne aka samu raguwar canjin kudi a wannan watan, me ya sa?" Yana samar da ƙima lokacin da ya zama rahoton da zai iya amsa tambayar. A cikin wannan rukunin, za mu ga yadudduka na BI, ƙirar ma'auni daidai, da kuma inda hankali na wucin gadi shine mai haɓakawa da tarko a cikin wannan tsari.
Tsarin gine-ginen BI gabaɗaya ya ƙunshi yadudduka masu zuwa. Tsarin tushen: Wuraren da bayanai suka samo asali, kamar ERP, CRM, kasuwancin e-commerce. Tsarin ETL (Turanci Extract-Transform-Load): Tsarin da ke fitar da bayanai daga tushen (Extract), tsaftace shi kuma ya canza shi zuwa tsarin daidaitaccen tsari (Transform) kuma ya loda shi zuwa manufa (Load). Ma'ajiyar bayanai: Wurin ajiya na tsakiya inda ake tattara bayanan tarihi da daidaitattun bayanai da aka tsara don bincike. Layer na rahoto/gani: dashboards, rahotanni, da tambayoyin ad-hoc. A cikin wannan sarkar, ingancin kowane Layer yana ƙayyade na gaba; Idan majiyar ta kazanta, rahoton kuma ya yi kazanta.
Ma'anar Ma'auni da KPI daidai
Ma'auni shine ƙimar ƙima da aka auna: jimlar juyawa, adadin umarni. A KPI (Maɓallin Ayyukan Maɓalli) ma'auni ne mai mahimmanci wanda ke auna aiki a kan manufa: " ƙimar abokin ciniki na wata-wata ƙasa da 5%". Ba kowane ma'auni ne KPI ba; KPI ma'auni ne wanda ke daure da burin kasuwanci kuma yana haifar da yanke shawara.
Mafi girman matsalar ayyukan BI shine rashin ma'anar ma'auni. Menene ma'anar "abokin ciniki mai aiki"? An yi oda a cikin kwanaki 30 na ƙarshe ko kwanaki 90? Ana ƙidaya waɗanda suka dawo? Idan ƙungiyoyi biyu suna nufin abubuwa daban-daban ta "yawan abokan ciniki masu aiki," wannan dashboard ɗin yana nuna gaskiya guda biyu daban-daban. Shi ya sa kowane KPI ya kamata ya kasance yana da jumla ɗaya, ma'anar da aka yarda da ita. AI da sauri yana samar da zane-zane na waɗannan ma'anar; amma ya rage ga sashin kasuwanci don yanke shawarar wane ma'anar "daidai".
Tukwici: Lokacin zayyana KPI, rubuta abubuwa uku: (1) dabara (menene ainihin ƙididdiga / ƙima), (2) taga lokaci, (3) abubuwan da ba a haɗa su ba. Samun AI ta ce "cire shubuha a cikin ma'anar wannan KPI a matsayin tambayoyi" yana bayyana zato na ɓoye.
Mataki-mataki: Ƙarfafa Rahoto Mai ƙarfi AI
Mataki na 1 - Bayyana tambaya. Wace shawara rahoton zai yi aiki? Maƙasudi mai mahimmanci kamar "za mu yanke shawarar yankin da za mu canza kasafin zuwa", ba "ya kamata ya yi kyau ba".
Mataki 2 - Ƙayyade ma'auni. Rubuta KPIs da ake buƙata tare da dabaru, windows da keɓancewa. Hankalin wucin gadi na iya samar da daftarin ma'anar.
Mataki 3 - Ƙirƙirar daftarin SQL. Ba da bayanin tsari ga haƙƙin ɗan adam kuma samar da daftarin tambaya. Amma karanta kuma ku fahimci tambayar kafin gudanar da ita.
Mataki 4 - Tabbatar da ƙananan bayanai. Gudanar da tambayar farko akan ƙaramin samfurin tare da sanannun sakamako; Duba jimlar da hannu. SQL na AI na iya zama daidai daidai amma a hankali ba daidai ba.
Mataki na 5 - Ƙara labari, da'awar gwadawa. AI na iya samar da rubutun labari don rahoton; amma tabbatar da kowane da'awar dalili ("shi yasa tallace-tallace ya fadi").
Tarkon Haɗin Kai da Sanadin
Kuskure mafi haɗari a cikin BI shine fassara ma'auni biyu waɗanda ke aiki tare azaman "ɗayan yana ƙirƙirar ɗayan". Daidaitawa shine lokacin da dabi'u biyu suka canza tare; Sanadin shine lokacin da wani ya haifar da wani. Maganar "Yayin da tallace-tallacen ice cream ya karu, lokuta masu nutsewa sun karu" gaskiya ne, amma ice cream ba ya haifar da nutsewa; Dalilin gama gari shine lokacin rani (yanayin zafi). Hankalin wucin gadi na iya yin sauƙi a sauƙaƙe jumlar dalili yayin samar da labarin rahoto; Masanin MIS ya amsa waɗannan da'awar ta hanyar tambayar "akwai wani bayani?" Ya kamata ya gwada shi. In ba haka ba, za a yanke shawarar da ba daidai ba bisa ga dalili mara kyau.
Karamin Harkoki guda uku: Ta Lambobi
Case 1 - Kudin awo da ba a bayyana ba. A cikin wani kamfanin sadarwa, adadin "masu biyan kuɗi" da aka gabatar wa kwamitin gudanarwa ya kai miliyan 2.1, kuma rahoton ƙungiyar kuɗi ya kasance miliyan 1.7. Bambancin shine ɗayan ya ƙidaya kwanaki 90 a matsayin "aiki" yayin da ɗayan ya ƙidaya kwanaki 30. An tattauna ƙimar girma mara kyau na makonni biyu har sai an fayyace ma'anar gama gari. Ma'anar KPI mai jimla ɗaya zai guje wa wannan ruɗani.
Case 2 - SQL na AI ba daidai ba. A wani dillali, AI ya ƙara layin dawowa zuwa jimlar lokacin samar da tambayar "matsakaicin kwandon kowane abokin ciniki"; Sakamakon ya kasance 12% sama da ainihin ƙimar. SQL ya kasance cikakke daidai gwargwado. Lokacin da gwanin ya tabbatar da adadin ranar da aka sani da hannu, ya kama karkacewar kuma ya sa aka ƙara matatar dawowa.
Case na 3 - Falacy na causality. A wani kamfani na e-commerce, dashboard ɗin yana cewa "tallace-tallace sun fi 18% girma a ranakun da aka aika kamfen imel" kuma ƙungiyar tana gab da haɓaka kasafin kuɗin kamfen. Binciken ya nuna cewa an riga an tsara kamfen ɗin don dacewa da manyan kwanakin yaƙin neman zaɓe (lokacin rangwame). Lokaci ne, ba imel ba, wanda ya jagoranci tallace-tallace. Idan an ƙara kasafin kuɗi ba tare da gwadawa tare da ƙungiyar kulawa ba, za a yi asarar kuɗin.
Rauni Mai Sauƙi / Ƙarfi Mai ƙarfi
Rawanin faɗakarwa:
Rubuta rahoton tallace-tallace SQL daga wannan tebur.
Ƙarfi mai ƙarfi:
Matsayinku: Kai mai nazari ne na BI mai hankali. Rubuta DRAFT na tambayar SQL bisa ga zanen da ke ƙasa. Dokokin: - Yi amfani da tebur/filayen da aka bayar kawai; filin da bai dace ba.- EXCLUDE ya dawo (status = 'Komawa') daga jimlar.- Tagar lokaci: kwanaki 30 na ƙarshe.- Layin sharhi ta layi abin da tambayar ke yi. - Ba da shawarar layin samfurin 1 wanda za'a iya inganta shi da hannu don gwaji a ƙarshe sashi.
Ƙarfin faɗakarwa yana iyakance makirci, yana sanya dokar kasuwanci (ban da dawowa), yana ƙayyade taga, kuma yana buƙatar fitarwa mai iya tabbatarwa.
Samfura guda huɗu masu Kwafi
1) Bayanin ma'anar KPI:
Rubuta cikakken bayanin KPI mai zuwa: dabara (mai ƙididdigewa/ ƙididdiga), taga lokaci, abubuwan da ba a haɗa su ba. Ƙara kowane shubuha a cikin ma'anar azaman tambaya.KPI: [suna, misali. "ƙimar abokin ciniki"]
2) SQL duba dabaru:
Bincika tambayar SQL mai zuwa: shin akwai haɗarin kurakurai masu ma'ana, JOINs na kuskure, ɓacewar tacewa, ko ƙidaya biyu? Rubuta hujja don kowane bincike. Kar a canza tambayar, kawai a duba ta. SQL: [tambaya]
3) Ba da rahoton labari + sarrafa da'awar:
Rubuta taƙaitaccen taƙaitaccen zartarwa daga teburin sakamako a ƙasa. Lakabi [SHAIDA DA AKE BUKATA] kusa da kowane da'awar dalili kuma bayar da shawarar madadin bayani. Kawai dogara da bayanan da ke cikin tebur. Table: [bayanai]
4) Duba daidaiton awo:
A cikin rahotanni guda biyu da ke ƙasa, ma'auni masu suna iri ɗaya suna ba da ƙima daban-daban. Akwai bambance-bambance masu yuwuwa a cikin ma'anar (tagar lokaci, tacewa, lissafi). Rahotanni: [A] [B]
Chart kwatanta: KPI mai kyau da mara kyau
fasali
KPI mara kyau
KPI mai kyau
Bayani
"abokin ciniki mai aiki"
"Abokin ciniki tare da ≥1 kammala oda a cikin kwanaki 30 na ƙarshe"
dangantaka da manufa
Babu
"Zama a kasa 5% asarar"
aunawa
m
dabara bayyananne
banda
rashin tabbas
Banda dawowa
Shin yana haifar da yanke shawara?
a'a
Ee
Kuskuren gama gari
- Barin awo ba a fayyace ba. Idan ana amfani da kalmomi kamar "aiki," "nasara," "cikakke" ba tare da dabara ba, kowace ƙungiya ta ƙidaya daban.
- Gudun SQL na AI ba tare da tabbatarwa ba. Madaidaicin tambaya na iya zama daidai ba daidai ba; kirga biyu da JOIN ba daidai ba abu ne na kowa.
- Daidaita ruɗani tare da dalili. Tunanin cewa "ya karu da wannan" yana nufin "wannan ya haifar da hakan" zai haifar da yanke shawara mara kyau.
- Vanity metric bin. Kuskuren zato amma marasa ma'auni kamar "jimlar dannawa" kamar KPIs.
- Gabatar da lambobi ba tare da mahallin ba. "Juya miliyan 4.2" kadai ba shi da ma'ana; Ana buƙatar yanayi bisa ga watan da ya gabata, manufa, ko kasafin kuɗi.
Tsanaki: Rahoton rahotannin da aka samar ta hanyar basirar wucin gadi suna da gamsarwa da ruwa; Wannan daidai yana haɓaka haɗarin. Hukuncin da ya dace na iya ɗaukar da'awar ƙarya na dalili. Gwada kowace magana "saboda" da "don haka" tare da shaida.
a takaice
Ilimin kasuwanci shine Layer wanda ke canza tarwatsa bayanai zuwa yanke shawara kuma ya ƙunshi tushe, ETL, ma'ajiyar bayanai da sarkar rahoto. KPI wani ma'auni ne mai mahimmanci wanda ke da alaƙa da burin kasuwanci, tare da ƙayyadaddun tsari da keɓancewa; awo wanda ba a bayyana shi ba shine mafi yawan kuskuren BI. Artificial intelligence provides significant speed in producing KPI definition, SQL draft and report narrative; amma kowane SQL dole ne a tabbatar da gaskiya, kowane lamba dole ne a goyi bayan bayanan da aka sani, kuma kowane da'awar dalili dole ne a gwada shi da shaida. Daidaitawa ba dalili ba ne; Labarin ruwa baya bada garantin daidaito.
Aikin aikace-aikace
Ƙirƙirar "ƙididdigar kammala" KPI don dandalin kwas ɗin kan layi. (1) Rubuta bayanin jumla ɗaya tare da dabara, taga lokaci, da keɓancewa (misali, an soke rajistar rajista?). (2) Ƙirƙiri tsari mai sauƙi (Rijista, Course, Ci gaba) kuma sami daftarin SQL da aka samar don wannan KPI tare da faɗakarwa mai ƙarfi. (3) Nemo aƙalla haɗari guda ɗaya mai yuwuwar ƙirgawa biyu ko tacewa ba daidai ba a cikin tambayar. (4) Buga taƙaitaccen taƙaitaccen bayanin ƙarshe kuma yi alama ga kowane da'awar dalili a cikinsa. (5) Ƙirƙiri misali na tarkon haɗin kai da kuma bayyana yadda za ku gwada shi.
jerin abubuwan dubawa
- [ ] Formula, taga lokaci da banda kowane KPI an rubuta su.
- [ ] Na karanta kuma na fahimci SQL da layin AI ya samar ta layi.
- [ ] Na tabbatar da tambayar da hannu tare da bayanan da ba a sani ba.
- [ ] Na gwada kowane da'awar dalili a cikin rahoton tare da shaida.
- [ ] Na tsara kowace lamba tare da maƙasudi (manufa/lokacin ƙarshe).
- [ ] Na cimma matsaya tsakanin ƙungiyar kan ma'anar awo.