I have added some more analysis on IPS meta using (give or take) two weeks of data. This can be found here.
Let's compare this Wednesday to Wednesdays in the past:
Week `page views`
<dbl> <int>
1 25 31861
2 26 29288
3 27 26134
4 28 50706
5 29 49773
6 30 32966
7 31 41723
8 32 30800
9 33 19670
10 34 35570
11 35 20137
12 36 28289
13 37 27842
14 38 29110
15 39 40004
16 40 20902
17 41 24496
18 42 13698
Lower, but those values are kinda all over the place, so let's compare it with what we might expect based on how this week started.
I am comparing Monday to Wednesday, because Monday is the last complete day of this week with HNQ, and Wednesday is the first complete day afterwards.
Week Monday Wednesday RatioWedToMon
<dbl> <int> <int> <dbl>
1 26 22842 29288 1.28
2 27 28242 26134 0.925
3 28 39476 50706 1.28
4 29 43726 49773 1.14
5 30 44029 32966 0.749
6 31 42273 41723 0.987
7 32 34528 30800 0.892
8 33 31826 19670 0.618
9 34 26157 35570 1.36
10 35 30651 20137 0.657
11 36 23212 28289 1.22
12 37 24193 27842 1.15
13 38 25534 29110 1.14
14 39 44303 40004 0.903
15 40 26644 20902 0.784
16 41 24323 24496 1.01
17 42 40497 13698 0.338
That's a stark difference. Looking at the summary statistics of the ratio before this week, we can see that week 42 is a definite outlier.
Ratio
Min. :0.6180
1st Qu.:0.8651
Median :0.9971
Mean :1.0061
3rd Qu.:1.1678
Max. :1.3599
On average, the traffic is equal between Monday and Wednesday, this week that was a third.
Fitting a simple linear model on the previous weeks: lm(data = train, formula = Wednesday ~ Monday)
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6185.6933 7634.4636 0.810 0.43136
Monday 0.7978 0.2316 3.445 0.00395 **
It's not a perfect model, but I don't have an awful lot of data to work with. Based on this, we would predict a traffic of ~38500. In reality, this was 13698, or 65% lower than expected.
Gist with my code. I don't think I'm allowed to share the Google Analytics data sadly.
Let's compare this Wednesday to Wednesdays in the past:
Week `page views`
<dbl> <int>
1 25 31861
2 26 29288
3 27 26134
4 28 50706
5 29 49773
6 30 32966
7 31 41723
8 32 30800
9 33 19670
10 34 35570
11 35 20137
12 36 28289
13 37 27842
14 38 29110
15 39 40004
16 40 20902
17 41 24496
18 42 13698
Lower, but those values are kinda all over the place, so let's compare it with what we might expect based on how this week started.
I am comparing Monday to Wednesday, because Monday is the last complete day of this week with HNQ, and Wednesday is the first complete day afterwards.
Week Monday Wednesday RatioWedToMon
<dbl> <int> <int> <dbl>
1 26 22842 29288 1.28
2 27 28242 26134 0.925
3 28 39476 50706 1.28
4 29 43726 49773 1.14
5 30 44029 32966 0.749
6 31 42273 41723 0.987
7 32 34528 30800 0.892
8 33 31826 19670 0.618
9 34 26157 35570 1.36
10 35 30651 20137 0.657
11 36 23212 28289 1.22
12 37 24193 27842 1.15
13 38 25534 29110 1.14
14 39 44303 40004 0.903
15 40 26644 20902 0.784
16 41 24323 24496 1.01
17 42 40497 13698 0.338
That's a stark difference. Looking at the summary statistics of the ratio before this week, we can see that week 42 is a definite outlier.
Ratio
Min. :0.6180
1st Qu.:0.8651
Median :0.9971
Mean :1.0061
3rd Qu.:1.1678
Max. :1.3599
On average, the traffic is equal between Monday and Wednesday, this week that was a third.
Fitting a simple linear model on the previous weeks: lm(data = train, formula = Wednesday ~ Monday)
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6185.6933 7634.4636 0.810 0.43136
Monday 0.7978 0.2316 3.445 0.00395 **
It's not a perfect model, but I don't have an awful lot of data to work with. Based on this, we would predict a traffic of ~38500. In reality, this was 13698, or 65% lower than expected.
Gist with my code. I don't think I'm allowed to share the Google Analytics data sadly.