Monday, May 10, 2010

The Time of Voting: Actual Throughput of the Voting Process Via the PCOS Scanning Count

Notes:

1. (Survey was taken 10 am to 11 am Monday May 10, 2010; although taken from one school, the experience nationwide should be similar to these figures taken for 19 clusters)

Conduct your own: Ask your BEI for current PCOS count and take down exact time (Pls set your time to true and correct time; your computer clock is accurate if connected to internet). You can email me your results and I will include that in this survey. Or send to sun cell: 0922 346 1273  streetstrategist@gmail.com

2. Throughput is the ultimate measurement of the voting process as an operational management issue. How many voters pass through the system during a given period of time? This is critical because this is the point of customer contact. The government is just lucky it is a monopoly. Otherwise, in a free competition regime, only a few will go to the government as a service provider.

3. For example, in cluster 157, a total 190 voters passed through the system in first 219 minutes of operations, giving a throughput of 1.15 voters per minute. This figure already includes all the hassles, the waiting time, the setup time, etc. This is the throughput figure of the entire system given all its problems.

4. Since there are 19 clusters in this survey, this should be enough to give us an idea of what is the throughput figure for each cluster throughout the country.

5. The process is non-linear. It is probably quadratic or exponential. What is meant by this? For example, if the current throughput is 1.86 minutes per ballot, it does not mean that this speed will remain the same. It is expected to speed up because the initial time spent for setting up the PCOS and the voter's queue will have been compensated in the later part of the day.

6. Non-linear also means that if it takes 213 minutes to process 271 voters, it doesn't mean that the only 271 voters will be processed in the next 213 minutes.

7. This report includes all (100%) of all the clusters in one school.

8. The fastest time is 0.786 minutes per voter, while the slowest time is 1.86 minutes per voter.

9. The average for the entire school during the given survey period is 1.16 minutes per voter.

10. Assuming 1000 voters per cluster, and assuming the same throughput speed, it would take the fastest cluster 13.1 hours to process the 1,000 voters. For the slowest cluster, it would take 31.1 hours.

11. The ideal number of voters per cluster is 568.9 voters assuming a cluster throughput rate of 1.16 minutes per voter with opening time of 11 hours.

12. This report is 100% accurate but its implications as to total time required to process the entire 1000 voters rests on the linearity assumption, that is, assuming the the same speed.

13. However, in reality, this linearity assumption is false. Because of the learning curve and since there is no longer any setup time required in the later part of the day, then the speed is not constant. It is expected to be quadratic or exponential.

14. But at least, this report is based on "actual and real" process times, therefore very useful for future elections.

15. This report also gives the Comelec some guidance on  how to conducts the tests, and how to measure the voting process.

16.The much taunted "Time to Fill Up the Ballot" parameter is useless. Comelec must conduct a real throughput measurement of the system in the future.

17. The Comelec must conduct an entire day "walkthrough" of the entire voting process, to see how the machines behave under hot conditions of frequent scanning, to see how the battery works, what are the bottlenecks of the entire voting process etc.

This is an Operations Research (OR) and Operations Management (OM) issue.





The Time of Voting: The Actual Throughput of the Voting Process via the PCOS Scanning Results









How fast is the actual voting speed? How many ballots are actually being read per minute?









Report compiled by Thads Bentulan streetstrategist@gmail.com   http://www.streetstrategist.blogspot.com









Actual Report 100% All Clusters of Mabolo Elementary School (Cebu City) 


(Time Taken from 10:04 am to 10:47 am Monday May 10 2010)



Start Time 7:00:00 AM















Cluster No. PCOS Count  of Ballots Read (Persons) Time Taken Total Time Elapsed (Minutes) Time Per Voter (Minutes per Voter) Time to Finish 1000 voters (Hours)


157 190 10:39:00 AM 219 1.1526 19.211


158 218 10:38:00 AM 218 1.0000 16.667


159 202 10:36:00 AM 216 1.0693 17.822


160 222 10:34:00 AM 214 0.9640 16.066


161 271 10:33:00 AM 213 0.7860 13.100 Fastest

162 224 10:32:00 AM 212 0.9464 15.774


163 180 10:30:00 AM 210 1.1667 19.444


164 209 10:29:00 AM 209 1.0000 16.667


165 151 10:28:00 AM 208 1.3775 22.958


166 193 10:26:00 AM 206 1.0674 17.789


167 200 10:24:00 AM 204 1.0200 17.000


168 106 10:18:00 AM 198 1.8679 31.132 Slowest

169 164 10:12:00 AM 192 1.1707 19.512


170 122 10:11:00 AM 191 1.5656 26.093


171 189 10:10:00 AM 190 1.0053 16.755


172 133 10:09:00 AM 189 1.4211 23.684


173 194 10:04:00 AM 184 0.9485 15.808


174 165 10:47:00 AM 227 1.3758 22.929


175 197 10:44:00 AM 224 1.1371 18.951











Total Ballots Read During Survey Period 3530

Ave Time 1.1601


Total Clusters 19

Total time for 1000 voters 19.33479


Average Ballots Read Per Cluster 185.7895

Ideal no. of voters per cluster 568.9227 for 11-hour period given the ave. throughput

Average Time Elapsed for Survey 206.5263






Average Time Per Ballot Read (Minutes Per Ballot) 1.1601






Fastest (minutes per voter) 0.7860






Slowest (minutes per voter) 1.8679