WEBVTT
1
00:00:01.300 --> 00:00:04.980
So the log log model is one adaptation
2
00:00:04.980 --> 00:00:09.780
of the regulation model that is necessary
when you apply it into marketing.
3
00:00:09.780 --> 00:00:15.220
Now we are going to look at many more of
those adaptations that are necessary.
4
00:00:15.220 --> 00:00:18.710
The main thing is regulations
are used in marketing for
5
00:00:18.710 --> 00:00:21.199
something called marketing mix models.
6
00:00:22.210 --> 00:00:26.110
This is where you're trying to
find out how does marketing,
7
00:00:26.110 --> 00:00:29.650
the different aspects of marketing,
affect sales.
8
00:00:29.650 --> 00:00:34.290
Now one of the things we saw in
multiple regression is, it's not just
9
00:00:34.290 --> 00:00:38.880
about what you include in the model,
it's also about what you have missed out.
10
00:00:38.880 --> 00:00:42.690
So what are the common variable
to consider when including
11
00:00:42.690 --> 00:00:44.440
in marketing mix models?
12
00:00:44.440 --> 00:00:46.720
So I go back to the basics here.
13
00:00:46.720 --> 00:00:55.080
I always say remember to include the four
Ps, product, price, place, distribution.
14
00:00:55.080 --> 00:01:01.660
So what you have up here are different
aspects of these 4Ps, product quality and
15
00:01:01.660 --> 00:01:07.270
brand lifecycle, whether it's a new
product or the first P, product.
16
00:01:07.270 --> 00:01:13.980
Now you got distribution which is about
place, we got price and promotion.
17
00:01:13.980 --> 00:01:19.040
So one of the things about promotion
to understand is that promotions have
18
00:01:19.040 --> 00:01:23.410
carry over, advertising today has
an effect on sales tomorrow or
19
00:01:23.410 --> 00:01:25.200
the day after, and so forth.
20
00:01:25.200 --> 00:01:29.640
Because people remember the advertisement
even when the advertisement stops.
21
00:01:29.640 --> 00:01:32.690
So you need to remember
to include the carryover,
22
00:01:32.690 --> 00:01:36.770
the effect of advertising
yesterday on sales today.
23
00:01:36.770 --> 00:01:42.134
So here are the four Ps, product,
price, place, promotions.
24
00:01:42.134 --> 00:01:47.694
A nice little tidbit from recent research
is that when you include these four Ps,
25
00:01:47.694 --> 00:01:49.507
what's more important?
26
00:01:49.507 --> 00:01:52.780
Turns out the first thing is product line.
27
00:01:52.780 --> 00:01:56.270
Product is the most important
factor determining sales.
28
00:01:56.270 --> 00:01:57.623
Second is distribution.
29
00:01:57.623 --> 00:02:00.230
How widely you distribute the product.
30
00:02:00.230 --> 00:02:04.010
Third is price and fourth is promotion.
31
00:02:04.010 --> 00:02:08.290
You get a lot of play about promotion,
a lot of thinking about promotion.
32
00:02:08.290 --> 00:02:12.770
Mainly because that's the thing you
can change much more quickly and
33
00:02:12.770 --> 00:02:15.820
the one thing that is more
prominent among the consumers.
34
00:02:15.820 --> 00:02:19.200
But the fact that it is actually,
product, place,
35
00:02:19.200 --> 00:02:22.440
and price that are effects on sales.
36
00:02:22.440 --> 00:02:25.670
Now let's turn to another
thing that is important
37
00:02:25.670 --> 00:02:27.700
when you use regression in marketing.
38
00:02:28.920 --> 00:02:32.670
This is the difference between
statistical and economic significance.
39
00:02:34.870 --> 00:02:39.880
Statistical significance is something
that we are commonly used to, right?
40
00:02:39.880 --> 00:02:42.950
This is the relationship
observed in the sample
41
00:02:42.950 --> 00:02:46.210
likely to be observed in
the population as well.
42
00:02:46.210 --> 00:02:48.890
So this is what e saw in p-value.
43
00:02:48.890 --> 00:02:55.220
We looked at p-values and said look is the
p-value of a coefficient is less than 10%.
44
00:02:55.220 --> 00:03:00.220
Then the coefficient in the model is
actually significant which means that
45
00:03:00.220 --> 00:03:04.220
this coefficient is likely to
be observe to have an effect
46
00:03:04.220 --> 00:03:07.600
even if you look at
another sample of data.
47
00:03:07.600 --> 00:03:12.570
Now economic significance is likely
different than statistical significance.
48
00:03:13.620 --> 00:03:19.040
You can have statistical significance,
but still not have economic significance.
49
00:03:19.040 --> 00:03:24.300
So what economic significance is,
is that does the benefit
50
00:03:24.300 --> 00:03:29.680
from a marketing intervention justify
the expense of that intervention.
51
00:03:29.680 --> 00:03:34.080
So that actually looks at what
is called effect size, right?
52
00:03:34.080 --> 00:03:37.570
How big is the effect of a coefficient, so
53
00:03:37.570 --> 00:03:41.150
that it is worth
the investment put into it.
54
00:03:41.150 --> 00:03:43.730
So now, let's look at an example here.
55
00:03:43.730 --> 00:03:45.790
Let's go back to the first example.
56
00:03:45.790 --> 00:03:50.868
Let's look at number of promotions in
the x axis and y axis is dollar spend
57
00:03:50.868 --> 00:03:55.864
by the consumer, and we saw all those
red dots, and the regression line
58
00:03:55.864 --> 00:04:00.633
going through, and by now you know
what these numbers are, right?
59
00:04:00.633 --> 00:04:01.939
This is the intercept.
60
00:04:04.506 --> 00:04:11.010
And this is the coefficient of x.
61
00:04:12.200 --> 00:04:13.010
You know that by now.
62
00:04:14.680 --> 00:04:19.412
And when we looked at the regression you
saw that the number of promotions has
63
00:04:19.412 --> 00:04:27.130
a p-value less than 0.1,
so it is likely that
64
00:04:27.130 --> 00:04:31.490
number of promotions will have an effect
even when you look at another sample.
65
00:04:31.490 --> 00:04:36.780
So you know that number of promotions
has statistical significance.
66
00:04:36.780 --> 00:04:42.826
Now what you need to know is does number
of promotions have economic significance.
67
00:04:42.826 --> 00:04:44.259
Let's see how we do that.
68
00:04:46.486 --> 00:04:51.752
So, a unit increase in number of
increase in number of promotions
69
00:04:51.752 --> 00:04:57.991
increases units purchased by 1.42,
that you get from the coefficient.
70
00:05:00.169 --> 00:05:06.100
Now assume the company making this
product has a gross profit of $5.
71
00:05:06.100 --> 00:05:08.590
Cost of promotion is $0.50.
72
00:05:08.590 --> 00:05:12.790
Now what you want to do is
construct an equation for profit.
73
00:05:12.790 --> 00:05:14.280
So how do we get profit?
74
00:05:14.280 --> 00:05:18.570
Profit is units purchased
times gross profit
75
00:05:18.570 --> 00:05:23.140
minus cost of promotion
times number of promotions.
76
00:05:23.140 --> 00:05:28.670
With the numbers we have up here,
we can say units purchased
77
00:05:28.670 --> 00:05:33.600
is 1.42 that you get from here for
a single promotion.
78
00:05:33.600 --> 00:05:36.115
If you have promotion of one unit,
79
00:05:36.115 --> 00:05:42.030
the coefficient gives you that value
is 1.42, that is the predicted sales.
80
00:05:42.030 --> 00:05:47.528
Times gross profit, which is $5,
minus cost of promotion,
81
00:05:47.528 --> 00:05:52.417
which is 50 cents,
times you are doing 1 promotion.
82
00:05:52.417 --> 00:05:59.244
So 7.1, that's 1.42 times
5 minus 50 cents times 1,
83
00:05:59.244 --> 00:06:03.224
that is 0.5 that gives you 6.6.
84
00:06:03.224 --> 00:06:09.680
So for a single promotion
you make a profit of 6.6.
85
00:06:09.680 --> 00:06:14.532
This means that promotions have
86
00:06:14.532 --> 00:06:19.230
economic significance.
87
00:06:19.230 --> 00:06:25.520
The cost of the promotion is less
than the benefit from the promotion.
88
00:06:25.520 --> 00:06:32.749
If the coefficient was not 1.42,
what if the coefficient was 0.01?
89
00:06:35.480 --> 00:06:39.305
What if the coefficient of 0.01 and
not 1.42?
90
00:06:39.305 --> 00:06:41.690
Then, it is not worth it.
91
00:06:41.690 --> 00:06:45.560
The cost is equal to the benefit,
you don't do the promotion.
92
00:06:45.560 --> 00:06:50.230
So, that's what I meant by
saying that even when you have
93
00:06:50.230 --> 00:06:54.610
statistical significance,
you may not have economic significance.
94
00:06:54.610 --> 00:06:56.510
You need to look at the effect size.
95
00:06:56.510 --> 00:06:59.380
You need to plug this into
a profit equation and
96
00:06:59.380 --> 00:07:02.850
see whether it makes sense to
invest into the promotion.