Auditing Voicemail Disposition Codes Against Actual Dial Events

Research question and evidence scope

Does a voicemail disposition code match an actual voicemail event in the phone system? The scope is one outsourced appointment-setting queue during a defined review window, not an industry benchmark. The audit compares the code a caller selected after a dial with the event the dialer or phone system recorded for that same attempt. It asks a narrow question: when the record says a voicemail was left, can the team show a matching dial event and a recorded message?

The review does not judge whether leaving the voicemail was a good decision or whether the prospect called back. It also does not measure answer rates, appointment rates, or the quality of the spoken message. Those outcomes depend on the list, the offer, the script, and the campaign rules. Separating code accuracy from call quality keeps a data-entry problem from being blamed on caller persuasion, or the reverse.

The unit is one dial attempt tied to one record, not one customer and not one shift. A record dialed three times creates three attempts, each with its own disposition and dial event. A voicemail left on the second attempt is a separate event from a voicemail left on the third. Keep the attempt identifier beside the record identifier so a reviewer can follow the sequence without merging repeated calls into one vague note.

Operating scenario and evidence

Begin with a dated export from the dialer or phone system and a second export from the system of record that stores dispositions. Include the attempt identifier, record identifier, dial start time with time zone, duration, result code, the disposition selected by the caller, the note text, the caller identity, and the campaign. Join the two exports on the attempt identifier rather than on a phone number that may repeat across records. A manual join on customer name is not strong enough for this audit.

Define the dial event as the outbound attempt recorded by the phone system, not the moment the caller opened a record on screen. Define the voicemail disposition as the code the caller saved after the attempt, not a later edit made during a shift review. Record both timestamps and flag any disposition saved more than a few minutes after the dial ended. A late edit is not automatically wrong, but it should be visible to the reviewer.

Method and review criteria

A matching voicemail needs more than a code that says voicemail. The dial event should show a connected call with a duration long enough to leave the message the campaign permits, and the note should state what was left and when. If the campaign forbids voicemail, a voicemail code is itself a finding even when the dial event matches. The audit should read the campaign rule before scoring any attempt.

Classify each attempt as matched, unmatched, missing, or conflicting. Matched means the code and the dial event agree on a voicemail attempt. Unmatched means the code says voicemail but the dial event shows no connected call. Missing means a connected call has no disposition or note. Conflicting means the note describes a conversation while the code says voicemail, or the code and the note disagree about what happened.

Exclude test attempts, internal calls, duplicate records created by a single click, and attempts the campaign was not permitted to make. State the excluded count and review a sample of excluded rows before accepting the filter. Do not drop an unmatched voicemail merely because the cause looks obvious, such as a system outage. Label the cause instead, because the pattern of causes is what the repair will target.

Authority and practical implications

Use only the fields needed to compare the code with the dial event, and restrict access to the joined file. Attempt identifiers and call outcomes can be linked to a named person, so treat the working export as personal data even when the report is aggregate. Store it in the approved location, remove the working copy when the review closes, and follow the organization's retention rule. The audit does not need message audio unless a separate quality review has approved access.

In a typical queue, several patterns appear quickly. One caller leaves voicemail on almost every attempt because the dialer connects to an answering machine and the code auto-fills. Another records voicemail when a call rings out without connecting. A third writes a detailed message in the note but selects the wrong code. Each pattern needs a different correction, which is why a single percentage is not enough.

Build a coding sheet with one row per attempt and columns for the dial result, duration, code, note summary, campaign rule, and final classification. Freeze the classification rules before reading the sample, and write the rule beside each category. Where the dial event and the code disagree, quote the relevant event rather than paraphrasing it. The coding sheet should let a second reviewer reach the same answer without asking the first reviewer what happened.

Limitations, conclusion, and sources

Use a census for a small queue and a documented sample for a larger one. Oversample attempts at the start and end of each shift, because handoffs and closing routines are where codes are most likely to be auto-filled or skipped. Also oversample attempts with very short durations, since a brief connection is the most common source of a false voicemail label. Report the selection rule and the number of attempts added after the cutoff.

Record the duration and the message start point the dialer exposes, because a connected call of three seconds cannot contain the campaign message. If the system provides an answering-machine detection flag, include it as context but do not treat it as proof. A detected machine can still reach a live person, and a live answer can route to a voicemail box. The reviewer's job is to compare evidence, not to certify the carrier's classification.

Have a second reviewer recode a subsample and compare results before discussing causes. Disagreement usually comes from a different definition of a connected call, not from arithmetic. Resolve it by writing the duration and event rule more precisely, then recode the disputed rows. If the systems cannot supply a clean attempt identifier, narrow the audit to attempts with two reliable timestamps rather than guessing.

The audit supports a specific repair, not a general warning. If unmatched codes cluster around one dialer result, the integration or the auto-code mapping needs attention. If missing dispositions cluster around one caller or shift, the closing routine needs a check. If voicemail is being left on a campaign that forbids it, the script and the campaign brief need immediate review by the owner.

Regulatory context shapes what the repair may permit. The Federal Trade Commission's Telemarketing Sales Rule requires sellers and telemarketers to keep records of telemarketing activity for twenty-four months, and the rule restricts calls to numbers on the National Do Not Call Registry. The Federal Communications Commission's guidance on the Telephone Consumer Protection Act covers automated calls and texts, including rules for artificial or prerecorded messages. These sources frame record-keeping and calling conditions; they do not validate any single voicemail code.

The measure has limits. A dialer can log a connected event that the caller experiences as a ring, and a carrier can route a call to a voicemail box without a clean signal. A code can be technically accurate and still describe a poor contact attempt. The audit shows whether the record and the phone event agree, not whether the prospect heard or understood the message. Pair it with answer and callback outcomes before changing a script.

Run the audit after any change to the dialer, the disposition list, the campaign brief, or the caller closing routine. A monthly cadence suits a stable queue, while a weekly check may fit a campaign with frequent changes. Archive the exports, coding sheet, definitions, and repair log so another reviewer can reproduce the result. Close the loop by naming the owner who accepted the repair and the date the next check will run.

Repeat the audit on a stable sample before announcing a trend. A single review can show that codes are inaccurate, but it cannot show whether a new training or a dialer update fixed the issue. Keep the same unit, window, and classification rules for the next check so the comparison means something. If the repair changes the disposition list, label the break and report the new baseline separately.

Sources: FTC guidance on complying with the Telemarketing Sales Rule, https://www.ftc.gov/business-guidance/resources/complying-telemarketing-sales-rule; FCC guidance on telemarketing and the Telephone Consumer Protection Act, https://www.fcc.gov/general/telemarketing-and-telephone-consumer-protection-act; FTC Q&A for the Do Not Call Registry, https://www.ftc.gov/business-guidance/resources/q-and-do-not-call-registry; NIST Privacy Framework, https://www.nist.gov/privacy-framework; U.S. Small Business Administration guidance on managing a business, https://www.sba.gov/business-guide/manage-your-business. Each source was checked on November 9, 2026 and is used for general operating context, not as a local performance benchmark. These references frame calling rules, privacy, and record-keeping expectations but do not validate any single queue's voicemail codes. Record the definition and sample size beside the result so the review can be repeated.