6,865 People at "Taste of Buffalo" at 3:30 PM Saturday
Below are the raw figures for a count of the "Taste of Buffalo" crowd made by Longacres, Inc between 3PM and 4PM on Saturday, July 6^{th}.
The results of a survey last year done by Focused Marketing Associates, and published in the Buffalo News on Friday, July 5^{th} indicated that the average length of stay at "Taste" is 3.3 hours. That figure correlates closely with a study we did ourselves at Allentown Art Festival this year, showing average stay at that event to be 3.53 hours.
CROWD TURNOVER  Based on the above length of stay figures it is reasonable to estimate that the counted crowd of 6,865 people turned over three times on Saturday, resulting in a total of about 20,600 different people attending the "Taste" on Saturday. If Sunday is about the same, the total "Taste of Buffalo" crowd is about 40,000. (Or one tenth the claimed 400,000  very nearly the same crowd inflation percentage that we found when we studied "Allentown Art Festival" earlier this spring. Click here for our study of the Allentown Crowd.
How we Counted:
The festival area was divided into grids bounded by the rapid transit line green power poles.
Each grid was approximately 150 feet long and covered one half of Main Street.
Two enumerators (counter A and counter B) counted the people in each grid, and those counts are included in the table below. Although the differences between the two counts occasionally approachs 25%, the total of all grids done by each counter is nearly identical (5 people off). Each counter feels certain that their margin of error was never great than 25%, and usually much better.
Late tonight, we will calculate an exact margin of error. Initially we assign an estimated margin of error of plus or minus 15% to this count.
NOTES ON MARGIN OF ERROR: Although the below table shows significant differences between two enumerators counting the same block (or grid) of people, those differences taken over a number of grid blocks tend to be self canceling. Note that the total counts of the two enumerators differed by only 5 people.
We are good counters, but not formally trained in statistics. There are different ways to calculate "margin of error", based on averages, "standard deviations", etc. The science of that discipline is beyond the scope of this study. For our own use, we announce a margin of error based on the average difference between the two enumerator's counts, expressed as a percentage of the average "grid count".
The average counted # in each grid cell at this event was 182.6 people
The average difference in count between two enumerators counting the same grid cell was 28.33
This calculates as a 15.52% margin of error counting each cell. As stated above, the more cells that are counted the more this error tends to cancel out.
The two counters backed each other up and counted together one side of each "grid section" (one side of main street)
The "times two" column shows the total for that section, doubling the halfgrid count to show the total for that section of the full width of Main Street.
There were three sections of Main street near the stages where the two sides of the street had very different crowds, and those were counted individually, the figures showing near the bottom of the following table.
Questions may be addressed to Thomas Kranz at 7166529495 or emailed to this website at: email link
counter A 
counter B 
Total count for grid 
average 
times two 

250 
260 
510 
255 
510 

260 
300 
560 
280 
560 

150 
160 
310 
155 
310 

195 
230 
425 
212.5 
425 

160 
155 
315 
157.5 
315 

165 
230 
395 
197.5 
395 

140 
120 
260 
130 
260 

175 
120 
295 
147.5 
295 

290 
300 
590 
295 
590 

185 
130 
315 
157.5 
315 

300 
230 
530 
265 
530 

145 
160 
305 
152.5 
305 

180 
180 
360 
180 
360 

90 
120 
210 
105 
210 

50 
45 
95 
47.5 
95 

total= 
5475 

following are counts for grids done separately by only one counter 

220 

165 

395 

170 

220 

220 

1390 
total for single counts 

grand total: 
6865 

total for dual counts for counter "A" 
2735 

total for dual counts for counter "B" 
2740 