Uso da Lógica Fuzzy no estudo de Health Score de clientes de uma solução digital para delivery
Carregando...
Arquivos
Data
2022-07-21
Autores
Suzuki, Andressa Yuri
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Este trabalho teve como objetivo criar um sistema de diagnóstico por pontuação
usando a Lógica Fuzzy, denominado Health Score, para avaliar a saudabilidade de
clientes em relação ao uso da Solução para Delivery da empresa Goomer
Tecnologias S.A. e agir preventivamente nas lojas com menor pontuação a fim de
evitar um cancelamento da assinatura do produto, reduzindo assim, suas taxas. A
Solução para Delivery da Goomer foi lançada no começo da pandemia de COVID-19
no Brasil e utiliza uma estratégia de mercado chamada Product-Led Growth que é
considerada nova no mercado, mas que é utilizada por grandes empresas com o
modelo de negócio SaaS (Software as a Service) como, por exemplo, a Netflix,
Slack, Nuvemshop, Loja Integrada entre outras que se tornaram referência neste
modelo. Essa solução possui uma versão freemium que, neste trabalho, foi chamada
de Plano A e planos de assinatura mensal chamados aqui de Plano B e Plano C.
Para esse diagnóstico prévio, duas versões de um sistema de modelagem
matemático foram criadas usando a Lógica Fuzzy e 6.600 lojas foram analisadas em
ambos. As variáveis escolhidas para comporem esses sistemas foram: faturamento
da loja nos últimos 7 dias, quantidade de pedidos recebidos nos últimos 7 dias,
quantidade de dias que a loja teve pedidos nos últimos 7 dias, quantidade de meses
que a loja teve pedidos nos últimos 6 meses e pontuação pelo uso das
funcionalidades disponíveis na solução. Ambas as versões resultam em um
diagnóstico que classifica a saudabilidade de cada uma das lojas em Ótima saúde,
Saudável, Estado de atenção e Doente. Nota-se que as duas versões foram
eficientes para prever uma maior tendência de cancelamento do Plano B entre as
lojas classificadas com uma saudabilidade pior. A 2a versão obteve uma acurácia de
52,17%, mais de 10 pontos percentuais abaixo da versão 1, com 63.83%. A Lógica
Fuzzy foi uma solução muito pertinente para a criação de uma metodologia Health
Score, mas percebe-se que ainda existe muita oportunidade de melhoria da acurácia
do sistema.
This work had the objective of creating a system of diagnoses by score using Fuzzy Logic, denominated Health Score, to evaluate healthiness of clients that use the Delivery Solution by Goomer Tecnologias S.A. and to act preventively in stores with lower score for the purpose of avoid a subscription cancellation of the product, reducing the churn rate. Goomer’s Delivery Solution was launched in the beginning of COVID-19 pandemic in Brazil and has a market strategy called Product-Led Growth that is considerate new in market, but it’s used by large SaaS (Software as a Service) companies such as Netflix, Slack, Nuvemshop, Loja Integrada and others that have become a successful cases on this model. This solution has a freemium version which at work was called Plan A and monthly subscription plans called Plan B and Plan C. For this previous diagnosis, two versions of a mathematical model system were created using Fuzzy Logic and 6,600 stores were analyzed in both. The variables chosen to the system were: store revenue in the last 7 days, orders received in the last 7 days, days with orders in the last 7 days, months with orders in the last 6 months and a score for the use of available features in the solution. Both versions result in diagnoses that classify the store healthiness in Very healthy, Healthy, Alert state and Sick. Both versions were efficient in predicting a greater tendency to cancel Plan B among stores classified with the worst healthiness. Second version has 52.17% accuracy, 10 points lower than the first version, with 66.83%. Fuzzy Logic was a very pertinent solution to create a Health Score methodology, but there are still many opportunities to improve the accuracy of the system.
This work had the objective of creating a system of diagnoses by score using Fuzzy Logic, denominated Health Score, to evaluate healthiness of clients that use the Delivery Solution by Goomer Tecnologias S.A. and to act preventively in stores with lower score for the purpose of avoid a subscription cancellation of the product, reducing the churn rate. Goomer’s Delivery Solution was launched in the beginning of COVID-19 pandemic in Brazil and has a market strategy called Product-Led Growth that is considerate new in market, but it’s used by large SaaS (Software as a Service) companies such as Netflix, Slack, Nuvemshop, Loja Integrada and others that have become a successful cases on this model. This solution has a freemium version which at work was called Plan A and monthly subscription plans called Plan B and Plan C. For this previous diagnosis, two versions of a mathematical model system were created using Fuzzy Logic and 6,600 stores were analyzed in both. The variables chosen to the system were: store revenue in the last 7 days, orders received in the last 7 days, days with orders in the last 7 days, months with orders in the last 6 months and a score for the use of available features in the solution. Both versions result in diagnoses that classify the store healthiness in Very healthy, Healthy, Alert state and Sick. Both versions were efficient in predicting a greater tendency to cancel Plan B among stores classified with the worst healthiness. Second version has 52.17% accuracy, 10 points lower than the first version, with 66.83%. Fuzzy Logic was a very pertinent solution to create a Health Score methodology, but there are still many opportunities to improve the accuracy of the system.
Descrição
Palavras-chave
Modelos matemáticos, Serviço de alimentação, Software, Negócios, Mathematical models