Goal: Build a simple model to predict Web Transactions. This model (with more time) can get much more granular/useful by adding things such as Product, Regional Locations, PLP and PDP Views, etc.
Takeaways
1) The model is 90% accurate, though this is a small sample size.
2) Paid Web Visits were NOT significant in the Model, but have a 68.8% Correlation with Web Transactions. Try to increase here, aka room for growth.
3) Web Transactions for March 1 beat the models expectations by 123 Web Transactions.
4) Organic web visit is the most significant factor, so room to grow on Paid web visits.
5) Mondays and Fridays have the most Web Transactions.
6) Paid social across the board is down in correlations to Web Transactions.
7) One recommendation is to focus on Social, as it has the most room to grow, but also make sure that Social is getting the Credit it deserves.
You can see here Mondays and Fridays have the most Web Transactions, though it doesn’t look statistically significant.
ggplot(dat2, aes( x = DOW,y = `Web Transactions`)) + geom_boxplot() + geom_jitter(width = 0.01)
You can see here the positive correlation with Paid Web Visits and Web Transactions
ggplot(dat2, aes(x = `Web Transactions`, y = `Paid Web Visits`)) + geom_point()
And you can see here the lack of correlation from Paid Social Clicks and Web Transactions
ggplot(dat2, aes(x = `Web Transactions`, y = `Paid Social Clicks`)) + geom_point()
Here is the model at 90% accuracy.
dat4 <- lm(formula = `Web Transactions` ~ `Paid Search Clicks` + `Paid Social Impressions` +
`Paid Social Clicks` + `Paid Social Cost` + DOW + year +
`Organic Non Paid Web Visits` + `All other non paid web visits`,
data = dat)
summary(dat4)
Call:
lm(formula = `Web Transactions` ~ `Paid Search Clicks` + `Paid Social Impressions` +
`Paid Social Clicks` + `Paid Social Cost` + DOW + year +
`Organic Non Paid Web Visits` + `All other non paid web visits`,
data = dat)
Residuals:
Min 1Q Median 3Q Max
-262.532 -79.399 1.152 82.426 312.237
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.508e+03 1.913e+02 -7.884 2.99e-12 ***
`Paid Search Clicks` 5.929e-03 2.325e-03 2.551 0.012177 *
`Paid Social Impressions` 1.674e-04 7.735e-05 2.164 0.032688 *
`Paid Social Clicks` 1.766e-01 5.655e-02 3.122 0.002313 **
`Paid Social Cost` -2.185e-01 4.588e-02 -4.763 6.09e-06 ***
DOWMonday -1.979e+02 5.041e+01 -3.926 0.000154 ***
DOWSaturday -1.633e+02 4.290e+01 -3.807 0.000236 ***
DOWSunday 1.155e+02 6.222e+01 1.856 0.066174 .
DOWThursday 6.947e+00 4.204e+01 0.165 0.869065
DOWTuesday -1.156e+02 4.238e+01 -2.728 0.007451 **
DOWWednesday -2.415e+01 4.181e+01 -0.578 0.564709
year2021 2.406e+02 6.016e+01 4.000 0.000118 ***
`Organic Non Paid Web Visits` 3.403e-02 3.330e-03 10.219 < 2e-16 ***
`All other non paid web visits` 6.205e-03 1.803e-03 3.442 0.000828 ***
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 120.6 on 106 degrees of freedom
Multiple R-squared: 0.8991, Adjusted R-squared: 0.8868
F-statistic: 72.68 on 13 and 106 DF, p-value: < 2.2e-16
Finally, below you can see the prediction for March 1. Web Transactions were 2385 and the model only predicted 2262, so we overshot the models expectations that day given these factors.
day month year DOW Web Transactions predictions prediction_accuracy
60 1 3 2021 Monday 2385 2262.408 122.5924