When it applies, it is shown that access increases if the number of nontop pages is increased.
Now, when adding contents to the site, it assumes that a keyword is diversified enough and the number of accesses of the visitor (s) who has searched assumes that it is proportional to the number of nontop pages in a site by this.
The number of nontop pages of the Japanese site of a Sonobe Laboratory (it is hereafter described as P) was 39 pages.
The ratio R of those (s) who have searched among visitors is function [ of P ] R (P).
Actual measurement R of P= 39 (39) :
NPJ:0.71,
APJ:0.71,
NPE:0.81,
APE:0.65,
Since it has gathered very much, it is assumed that it is R(39) =0.72 here.
Since ratios other than , s, and s (namely, d) are having assumed that a visitor's (s) number of accesses was proportional to the number of nontop pages in a site in a top to general P although it was R(39):1R (39) at the time of P= 39 and it is set to R(39) xP / 39:1R (39), R (P) is.
R(P) =1  (1/(1+ (R (39) /(1R (39))) (xP/39)))
becoming ., for example, :
R(0)=0,
R(1)=0.062,
R(10)=0.40,
R(39)=0.72,
R(100)=0.87,
R(390)=0.96,
R(1000)=0.985,
R(10000)=0.9985,
R(100000)=0.99985,
.
Next, the ratio of what came to the nontop page among the visitors (s) who have searched (sN) is from statistics. :
NPJ:0.98,
APJ:0.98,
NPE:0.98,
APE:0.96,
since  here, it is assumed that it is 0.98
Furthermore, the ratio of the visitor (what contains t or T in an access pattern) who will look at a top page among the visitors (sN) who have searched the nontop page is the abovementioned
table 1
.
Like (h) and (the ratio which showed the top page to the visitor who has searched the nontop page) of a (main index statistical table), since it is 0.892 in APJ, it is assumed that it is 0.89 here.
If amount of increases deltad=delta[ of top page access by addition display ] d (P) is expressed with the ratio which set the whole number of visitors to 1
deltad(P) =R (P) x0.98x0.89=0.87xR (P)
coming out  for example
deltad(0) =0,
deltad(1) =0.054,
deltad(10) =0.35,
deltad(39) =0.63,
deltad(100) =0.76,
deltad(390) =0.84,
deltad(1000) =0.857,
deltad(10000) =0.869,
deltad(100000) =0.870,
.
If the above formula is used
"
top page access after applying ‚Ì‹Zp  before application  comparing  the number of deltad(P) x visitors  it increases
"
It is predicted.
Therefore, it is if a site scale is set as six kinds as follows (this).
Model A
It carries out.
A nontop page

Annual visitor

Site

The number of the increases in access

1 page

300 persons

Small scale site

16 times
increase

10 pages

3,000 persons

Smallscale site

1,000 times
increase

39 pages

12,000 persons

Minor scale site

7,600 times
increase

100 pages

30,000 persons

Middlescale site

23,000 times
increase

1,000 pages

300,000 persons

Largescale site

260,000 times
increase

10,000 pages

3,000,000 persons

It is overly a largescale site.

2,600,000 times
increase

It becomes.
The model of the abovementioned trial calculation is changed and it is the site (this) where the number of annual visitors is the same and which had a nontop page 10 times but, respectively.
Model B
It carries out.
Let's also make the trial calculation of the install effect.
Increment is the number of deltad(P) x visitors.
A nontop page

Annual visitor

Site

The number of the increases in access

10 pages

300 persons

Small scale site

105 times
increase

100 pages

3,000 persons

Smallscale site

2,300 times
increase

390 pages

12,000 persons

Minor scale site

10,000 times
increase

1,000 pages

30,000 persons

Middlescale site

26,000 times
increase

10,000 pages

300,000 persons

Largescale site

260,000 times
increase

10,000 pages

3,000,000 persons

It is overly a largescale site.

2,600,000 times
increase

‚Æ,  "
To the site below middlescale, a big effect can be acquired especially.
" Things are understood.
The above "every year" is . which can be transposed also to arbitrary fixed periods, such as "monthly."
Since it is actually influenced, of course by various elements including the individuality of a site, this trial calculation is a standard.
