Boulder County Watcher Information

The following watcher accounts relate to Boulder County where I, Harvie Branscomb, also watched on two occasions. One topic is the quality of reference signatures and captured signatures. Another is the policy for handling the witnessed signatures. Another is the result of failure of the SCORE statewide voter registration database access statewide, and also server failures or overloads within Boulder County. Also described is the very strong pressure applied to and by election judges to accept discrepant signatures based on a very minor resemblance of perhaps only one feature. I have comments on all these topics and will insert them into the following three reports in green and [square brackets].

Harvie Branscomb

From: Russ Boehm Sent: Sunday, November 13, 2016 2:43 PM To:Cc:Subject: Re: REMINDER: Boulder County Canvass Board - Audit and Canvass Schedule

Election Friends,

This may or may not be pertinent to your current cause & task but I have a few comments on the process for signature verification. Some of these are comments on the process for Election Integrity members interested in following up with the SOS  and some are questions of C.R.S.1-7.5-204  available here and the administration thereof in the actual process.

On the last point. As part of the training of signature validation judges, we were again advised that a witnessed signature is accepted without question because "that's the law" according to the Clerk's team lead. I asked for the legal citation which of course couldn't be immediately produced. It took a few days for the team lead to get the citation and it was originally given to me as Rule 2.6.2 and 7.2.9. I checked the rules, found  and read as follows:

Rule 2

2.6.2 If an elector is unable to sign, another person must witness the elector’s mark. An elector may use a signature stamp because of age, disability, or other need. The stamp is treated as a signature and does not require a witness. And,

Rule 7

7.2.9 On all ballot-return envelopes printed after April 1, 2016, the county clerk must provide a space on the ballot-return envelope for a witness to the elector’s mark to provide his or her full legal name.

Obviously, neither of these rules says that a witnessed signature is automatically accepted. I reported this back to the team lead and requested some further investigation. A day or two later I was referenced to C.R.S.1-7.5-204 referenced above. It reads as follows:


(1)(a) Before opening any mail ballot, one of the receiving judges, in the presence of a majority of the receiving judges, shall inspect the self-affirmation on the return envelope.

(b) The self-affirmation is valid if:

(I) The self-affirmation was completed by the elector;

(II) The self-affirmation was signed by the elector or, if the elector is unable to sign, marked by the elector with or without assistance and witnessed by another person; and

(III) In any election conducted by a county clerk and recorder, the signature on the self-affirmation matches the signature stored in the statewide voter registration system, or the eligible elector's marks on the application and the self-affirmation were witnessed by other persons.

(c) If the self-affirmation is valid, the receiving judge shall open the envelope without defacing the self-affirmation or mutilating the enclosed ballot.

(d) For the purposes of subparagraph (III) of paragraph (b) of this subsection (1), the signatures on an eligible elector's self-affirmation and stored in the statewide voter registration system shall be compared in the manner prescribed by section 1-7.5-107.3.

- See more at:

[FYI at present the format for a return envelope provided on the SOS website in a portion accessible only to officials contains a witness line that is marked: “Witness’s Legal Name”. The various counties have described this line as “Witness’s Legal Name” (Eagle, Archuleta, Rio Grande, Douglas); “Witness Legal Name” (Huerfano); “WITNESS’S LEGAL NAME” (Jeffco); “WITNESS” (Weld); “PRINT WITNESS FULL LEGAL NAME” (Denver); but Arapahoe prints two lines, the one above: “Witness”, and below: “WITNESS PRINT NAME”.]

My understanding of the Rules and Law would be as follows:
  • As per Rule 7, the witness must provide their full legal name without question. They might also be required to provide their signature by inference from 1-7.5-204 section (1) paragraph III(d) since the reference is to signatures (plural) on the self-affirmation (singular). I don't specifically remember but I don't think there was a note or provision for the witness to provide a full legal name.

[Of about 10 counties’ envelopes and affirmations sampled only one implies the need for a signature of the witness – Arapahoe.]

  • Further, the elector's marks or unintelligible signature (my emphasis) must be matched to the SOS database.

[In reality there are numerous successful signatures that also are accompanied by a printed witness name. In one case I witnessed an election judge at in-person voting in a VSPC offer to witness by signing a voter’s envelope and including the role – election judge as part of the signature.]

  • Further by inference, if there are no elector's marks or attempted signature, the ballot is not accepted without further verification by the elector as per 1-7.5-107.3.2(a)

[ I believe this situation is responded to variably by county. In some cases the envelope might be accepted for counting. In others it might be subject to a cure for no signature. ]

  • Further I would infer that the witness' name must be verified in the SOS database as a registered voter and perhaps, if the witness of a signature is required, it must be verified.

[In some counties the witness name is checked and if in the SCORE database the envelope is counted without further attention. If the name is not in the database then the envelope is counted and also documented and held aside for later review. Counties generally interpret the law such that no check of the witness name is called for in law or rule. ]

Regarding the validation process. Much time was wasted looking up signatures in SCORE only to find there were no signatures on file or the only signature on file was from a finger-pad signing devise. In most cases this led the judges to finding an excuse to accept the signature rather than verifying as valid. It was suggested to me at one point that a dotted "I" looked good! Nothing else did.

[ My experience in Boulder was similar. Boulder has three layers of human verification and also a software layer. In each step, one decision or one judge can approve the signature – first by software, then a single judge on first tier “New Review” then either of a pair of judges on 2nd tier and then either of a pair of judges on a 3rd tier can approve the signature for counting. When I challenged a very discrepant pair of signatures that was accepted on first tier, it was then handled by the 3rd review team who lookd for other signatures in SCORE, then both judges said they see “nothing there” to compare.  It was when another judge pointed out that one or two characters in the printed name on a SCORE stored registration form resembled the same characters in the signed name on the envelope that the challenge was denied and the envelope once again approved for counting. The watcher minder with me seemed as surprised as I was to see this turnaround in the decision that to me suggested that challenging decisions in Boulder would not likely be fruitful. I said to her it was like pulling a rabbit out of a hat, and also like my experience in fighting a parking ticket when carrying good photographic evidence but the judge refuses to let the system admit to a mistake.]

I would suggest [to] the SOS:

It is inconceivable to me that a banking system (BITSY signature verification software as noted by the Clerk's team lead) would only be able to verify 11% of elector's signatures. The R.O.I. on such a system would seem to be marginal and hardly worth their development time. My graphoanalysis training also taught me that signatures are practiced and are generally recognizable and difficult to forge.

[BITSY is the familiar name of the Bell and Howell envelope scanner sorter and signature verification subsystem used in Boulder. A similar system with two sorter scanners is used in Denver.]

  • Not accept finger-pad signatures as validation of ballot signatures.
  • Examine entries for documents that have no signature for inclusion in signature verification Perhaps a sub database of documents suitable for election signature verification should be available

[SCORE is filling with non-signature records as well as both good wet signatures signed in person after ID is shown, poor tablet signed exemplars that are signed in person after ID is shown, and incredible signatures signed remotely on ballot envelopes, signature cure forms, mailed in voter registration forms, etc. Yes, SCORE must be updated to classify preferable trustworthy signatures to distinguish against those that are of unknown provenance. And also the non-signature forms such as outgoing cure letters entered by some counties do not need to be seen by signature verifiers. In addition, there is unnecessary proprietary demographic information shown to signature verifiers that is unnecessary. Once bounding boxes on the various digitized forms have been created, the verifiers need only see the signatures and not the party name, age, sex and address of the presumed voter.]

  • Too many electors seem to have a variety of signatures with strong characteristics that nullify another signature(s) on file. If we are going to use mail-in ballots, I suggest the Clerk's Office solicit a verified elector signature annually using a fold-over, returnable, secure postcard. If not annually then perhaps in January of an election year.

[This suggestion is in the correct direction but insufficient. Another remotely signed card will not add credibility to the signatures if mimicry of the voter is involved or possible. We need in-person attestation or demonstration of identity and periodic collection of a wet, on-paper signature to refresh the database and to replace previous questionable or poorly captured signatures. Some SCORE records are as mentioned here filled with disparate signature exemplars and could be focused on first. But all records are subject to degradation as aging takes place and as mimicked signatures are accepted as future exemplars for comparison. The occasion of in-person drop off of a mail-in ballot packet is the correct moment to ascertain identity and refresh the database with a new, observed signature on the envelope, now accredited by an election judge or other official.]

Signature verification is tedious and repetitive, things not normally handled well by people. The team leads were very concerned about the "acceptance rate" and the fact that it needed to be higher. The position description is Signature Verification not acceptance but "good work" is determined by acceptance level and you quickly learn that higher verification levels are based on people that accept signature no matter how difficult it might be to verify them against the seven rule characteristics. And of course, no voter is contacted unless there is no signature on file or the highest level verification can't agree.

[can’t both agree that the signature must be challenged with a cure process. If there is disagreement, the ballot will be counted.]
[Yes, the system itself is designed to err in the direction of acceptance of the signature on the envelope- through the fact that any disagreement leads to acceptance, through the multiple layers of examination, through the administrative pressure to maintain expected acceptance rates, through peer pressure by the adjacent judges some of whom are more outgoing and dominant. In El Paso County each day a representative from the DA’s office reviews potential challenges of signatures with 2nd tier judges. This is a remarkable integrity feature, but it fails to check questionable mismatches that have been determined by judges after some consternation to match. If our signature verification system were more balanced, these probable mismatches that had been adjudicated as matches would also be reviewed, preferably by experts. In Garfield County signature verification judges became aware that an acceptance of a signature in error led to disenfranchisement of the voter who came to vote in person and could only then vote provisional.]

Enough for now


[Cathy Jarrett is also a canvass board member.]

From: C JARRETT   Sent: Sunday, November 13, 2016 11:59 PM To:Cc:Subject: Re: REMINDER: Boulder County Canvass Board - Audit and Canvass Schedule


Here is some of the information that I have learned about some of the items in the emails below.

[now meaning above]

I believe that it was during the Logic and Accuracy Test that Hillary Hall herself told us that the ballot envelope opening machines would send a puff of air into the envelopes to separate the  secrecy sleeve (which contained the ballot) from the envelope.  (I believe that this year Harvie witnessed this and reported that the operator quickly put the envelopes into a pile.)  Then the sleeves containing the ballots were sent to two person teams to have the ballots removed.

[ A robotic desk called “Opex” sequences envelopes through a milling device that opens the top and side of the envelope and then two suction cups pull the sides of the envelope apart such that the operator cannot see either side. The operator quickly removes the contents and stacks ballots, presumably in secrecy sleeves, in quick succession. It is unlikely that a missing ballot or extra ballot will be detected at this point, but the scanner sorter used in a prior step has the ability to check thickness of envelope for this purpose and outstack such anomalies.  When the election contains two cards, such as Boulder, there is much more difficulty with this determination. I observe that the system can longer really successfully distinguish double ballots or missing ballots and that voters returning envelopes with such anomalies will get credit for voting (meaning they cannot vote again). If there is an extra card or two front cards in the envelope, both will likely be counted, even as the mismatches of front and back are caught in visual verification. It is possible that the process does remove two front cards from the count successfully. Boulder has some processes that are difficult to watch and evaluate and questions about process are often left unanswered]

This is unlike in the past when one person read the precinct number and DS number from the envelope out loud and another person pulled the secrecy sleeve back and read the same from the ballot.  Only when the numbers matched was the ballot removed from the secrecy sleeve.  The secrecy sleeves were placed in one pile and the ballots were opened up, flattened and placed in another pile.  Before and after this process the envelopes and ballots were counted.  Occasionally a ballot was in the envelope without a secrecy sleeve. A few times the wrong ballot was in the envelope.  ( I remember once when I was doing this job that the ballot in the envelope at a general election was from a previous primary election.)  This was reported to the Lead (worker) and noted on the tracking slip.  Then the ballots were sent to the Visual Inspection people to look for anything that might prevent the ballot from being read by the scanning machines.  Those ballots were put in colored folders on the top of the pile of ballots.

[In this General election (2016) there are three outstack folders used by the visual verification teams- one for the usual verification and the others for “mismatch” and “orphan”. The last two are the result of a two card ballot. Ballots are marked during visual verification by election judges in red ink sequentially as “O1, O2” or “V1”, “V2” etc. and put into colored folders (red, green yellow). This way they can be handled specially in the scanning room. My understanding is that orphan ballots gain a blank alternate page so that scan counts are uniform. Mismatch ballots are isolated and rebatched for later counting in the hopes that the match can be reestablished. However it is also the case that the style of the incoming ballot is NOT checked with the envelope style as the separation happens in the OPEX machine. In this case, privacy of the voter intent is being treated as preferential to the integrity of the style of the returned ballot.]

I was hired as a VSPC Election Judge for the last few days of this election.  I attended training from Oct. 31st until Nov. 3rd.  Then I worked at the Longmont Fire Station #5 on Nov. 4th, 7th and 8th.  All of the equipment worked well until the morning of Nov. 8th.  Shortly after the poll opened at 7:00 AM and after printing two ballots as usual, one of the two printers stopped working.  The IT person who was assigned to our location all day tried unsuccessfully to get it working again.  She called for help and the head IT person came and got it working again.  Meanwhile one of the laptop computers would not let anyone sign in to use it.  Another replacement laptop was brought to our location and it also did not work at first.  Finally it was fixed and it continued to work for the rest of the day.  Then about 10:00 the whole system went down.  I suggested that we go into Contingency Mode.  I was told that we had it wait for a decision from the main office.  We could not process voters for in person voting, but people could still bring in their mail ballot envelopes and put them in that ballot box.  About 30 or 45 minutes later we were told to go into  Contingency Mode which meant using Provisional ballots for all in person voters.  We could use the laptop computers to look up the voter's address and find their correct ballot style.  Then we could use a different program to print their correct ballot.  We used that system for the rest of the day.  I did learn later that some of the larger polling places did get to go back on the system and print the ballot the regular way.  I understood that they did not want to overload the system and that is why our smaller location continued to use the provisional ballots.

[ The statewide voter registration system went down 29 minutes, 2:47 p.m. to 3:16 p.m. on election day. SOS IT DirectorTrevor Timmons reported to me on 11/9 in person that there were actually two outages of the voter registration database SCORE- another at 10:32 to 10:56AM.   Similar slowdowns and stoppages have happened with access to the state’s registration data in two previous elections including on election day. The outages created lines in major county polling centers some of which never recovered and voters were voting between 9 and 10PM. ]

As for witness signatures I can confirm that when I was doing signature verification in the past, we were told to accept all witness signatures without questioning them unless the witness put the letters POA (Power of Attorney) after their name.  Those were not to be accepted.  When I worked as a Health Care Facility election judge for the 2016 Primary Election, if the voter wanted help to  fill out their ballot, my partner from another party and I filled out the ballot according to the voter's instructions.  Then if the voter could not sign his or her own name , he or she would make a mark and I would witness it and sign my name on the witness line.

Also I was told that we were to work very quickly when doing signature verification.  In fact sometimes our Lead (supervisor) told us that two seconds was enough time to check a signature on the first tier of checking.  When working on the second tier (checking signatures that did not pass on the first tier), we worked in bipartisan teams and were allowed more time because we had to look for additional signatures in SCORE.  I know that questionable signatures were sometimes passed.  These included those from the touch pad or tablets.

Mary and I watched the "test" of the first 150 signatures that the BITSY machine had passed.  Two bipartisan judges were accepting the signatures very quickly.  Sometimes I could only look at one signature and part of the reference signature before they went on to the next one.

[This is a defect of the way watching of signature verification is set up. Watchers can only watch the process being performed by election judges. Election judges are moving as fast as they can. When working alone, there is no conversation and little indication of the decision of the judge other than watching the mouse cursor on the screen and which screen button is clicked. Sokme counties have all such adjudication done in bipartisan pairs and in that case there may be conversation that facilitates the watcher’s understanding. In most large counties two judges are involved as a team to adjudicate at 2nd tier (or later).]

I asked my host to ask them to slow down and they did for a short while and then they went back to the previous pace.  I did challenge two signatures one did not look the same to me and the second did not have a reference signature.  I filled out the challenge forms.  The next day  Justine showed me screen shots of more signatures for those voters.  (She did not let me look at the signatures on SCORE myself as I had requested.)  The screen shots looked like they came from SCORE and I withdrew my challenges.  She also said that the BITSY machine had "deeper information" from another server than what had showed up on the screen of the election judges who were accepting the signatures.

[I doubt that Bitsy the sorter has access to more signatures in SCORE. If so that would be I think contrary to rule. The sorter’s autosignature (ASV) comparison should be similar to a human first tier – capable of comparing only two signatures, one from the envelope and one from the latest example uploaded to the SCORE system – usually an envelope signature from the previous election, or possibly the signature from a cure letter.]

I believe that the "deeper information" was additional signatures from SCORE.  Later Mary sent me a copy of the form that the judges were filling out to accept the signatures.   She had requested it from Justine.  On the form containing the signatures accepted by BITSY, some (rows 1-44 and 88-150) had the code 50 (which I believe from previous experience means accepted) next to the ballot envelope ID numbers and under the column headed "ASV Results" (Automatic Signature Verification Results).  The rest of the rows did not have anything printed in that cell, but the judges did printed a  red letter "A" (Accepted?) in most, but not all, of those cells.  They also printed a red letter "A" (Accepted?) in all of the cells under the heading Judges Result.  They also wrote the word "challenged" next to one of the signatures I challenged and and "OK" with something else scribbled out next to another challenged one.  Another one has a red line next to that row.  (The judges use only red pens in the ballot processing areas.)  I knew that the 150 signatures had been accepted by BITSY and I believe that the two judges also knew that.  It is likely that they were under pressure to accept all of those signatures.  I never saw them go deeper into SCORE to look at additional signatures before accepting them.

[I think Cathy here is discussing the “audit” of the ASV specifically.  I have only once happened into a facility when the ASV audit was taking place. The reason to go into SCORE during this audit would be if the reference signature was suspect and all the remaining signatures in SCORE appeared consistent and probably belonging to the voter (or signed in person and thus higher credibility). Indeed in no county is such a SCORE lookup occurring during the ASV audit to my knowledge.]

Thank you for reading all this.

From: Mary Eberle Sent: Sunday, November 13, 2016 7:28 AM To:Cc:Subject: Re: Fw: REMINDER: Boulder County Canvass Board - Audit and Canvass Schedule

Hi Cathy, Thank you for the schedule. I'll be thinking of what CB [Canvass Board] might do. Maybe we should get together separately from Hillary's meeting on Monday. Here's something that Canvass Board might look into: Boulder County's computer problems were much worse than the half-hour SCORE crash. Ask Aaron for a written report. Here is what I heard and watched on Election Day in the Houston Room VSPC: Hillary was using a single server to support the VSPCs, and it could not keep up on Election Day. It crashed multiple times. In 2012 we had more than 1,000 provisional ballots cast, and Denver had more than 10,000 (if memory serves). Other counties were similar. Processing provisional ballots is very costly and time-consuming. The large numbers were part of the county clerks' arguments for going to all mail ballots. Guess what! This time Boulder County has more than 4,000 provisional ballots to process because of the multiple crashes that prevented verification of in-person voters. I think Hillary will use this occurrence--likely avoidable had her IT department used its second server, as was mentioned by IT personnel--to push getting the Dominion system instead of waiting to get the new Hart or Clear Ballot system. During the crashes, the VSPCs went into "contingency" mode. (That means using provisional ballots--but the clerks don't use the word "provisional" because it scares the voters. I was told that, and it is true that it scares voters.) I saw IT personnel loading software into the laptops the election judges use to check voter eligibility and issue the ballot with the correct style. The computers also were rebooted. I got the two VSPC training manuals. I will give them to Peg so that they can be scanned and distributed. There is a section on "contingency." Here are two more things that the Canvass Board can be concerned about in addition to the things that you have noted regarding signature verification: As far as I know, Clerk Hall did not send election judges to pick up ballots in ballot return envelopes at the Post Office, the Voter Service and Polling Centers, or the standalone drop-off boxes. Instead, she sent "couriers" who are not sworn election officials to collect and transport the ballots. I did watch one pickup at the drop-off box on Broadway at Iris, and the process seemed good. However, there is no way that any video would be able to confirm that, I believe. Here are the critical questions: Were all the ballots collected? (Yes.) Was the box sealed? Well, I was standing right there in the dark at 7:00 p.m. on Election Night, and I could hardly detect whether the seals were actually locked. I did take photos, which I'll send at some point. Second, as far as I know, Clerk Hall did not have the judges check that the returned ballot is the correct style. I meant to watch that process, but I was never present when the machines to open the envelopes were in use. Sorry. Maybe you watched them and can confirm how the process worked.

[I can confirm that the incoming style was not being checked, but card consistency was being checked using the semi-unique lower left numbers and bar codes.]

Also, remember that although Hillary's schedule calls this meeting on Monday the "Canvass Board Kick-Off," the Canvass Board has existed ever since Peg appointed its members. It exists outside of the clerk's control. She does not bring it into existence. She likes to think that she does, however.

Best, Mary

Hi Harvey,

Here is the web link to the Boulder County Schedule of Activities:

[Schedule ]

Look at Election Resources underneath the 2016 General Election Information section. Activities performed by judges (Signature Verification, Visual Inspection, Resolution, Tally) and Canvass dates are listed in the activity schedule.  The activity schedule is the document that is responsive to your CORA request.

Regards, Justine

Go back to 2016 Watching Page