Logistic regression perfect fit but non-significant coefficients

I have a data set for customers with some columns like sign-up-date, address-added-date, credit-card-added-date and order-placed-date. Only a portion of signup customers added address, a portion of them added credit card information, and at the end most of them do not place orders. So most of the cells in this data set is NA. The goal is to predict whether a sign-up customer will place an order. I convert all the dates,except the sign-up-date into binary variable, to indicate whether a sign-up customer will add address or add credit card. I then use these two binary variable to fit the binary outcome of whether a customer has placed order. I use logistic regression, and have perfect fit. However, I have very high p-values. How do I interpret this result?

• Why do you say that you have a prefect fit? You parameter estimates all have very large standard errors. May 20, 2017 at 16:42
• @MichaelChernick I am new to logistic regression. The predicted values are all +/-26.56607, which are far off 0/1. But I used this for classification which perfectly match the binary result +26.56607 for 1 and -26.56607 for 0. May 20, 2017 at 17:12
• Can you show us the results of running confint(fit.logistic)? (maybe you have to do library(mass) first). May 20, 2017 at 17:49
• what is the correlation between mydata3[6] and mydata3[7]? May 20, 2017 at 21:09
• It looks to me like either multicollinearity or perfect separation about which there are many questions and answers here. What does cor(as.matrix(mydata3[,c(11,6,7)])) show? May 21, 2017 at 1:17