Abandoned calls
Calls where the caller hung up while waiting, before a person answered, whether they were ringing or sitting in a queue.
An abandoned call is someone who wanted to talk to you and gave up. Unlike voicemail, they left nothing behind.
How to count it: Abandoned calls divided by all inbound calls, after you exclude a short-abandon band of hang-ups in the first few seconds (usually wrong numbers or instant regrets). State the seconds you exclude: CallRail treats every hang-up under 15 seconds as abandoned outright, Dialpad excludes hang-ups under 10 seconds by default for its contact-center product, and Five9 uses a 2-second rule. Pick one number and apply it everywhere you compare.
Watch out: Letting the phone system count a call as answered when it only reached a menu, so a caller who hung up in the menu never shows as abandoned. Two vendors rarely mean the same thing by “abandoned.”
- CallRail’s abandoned status already is its short-abandon bucket (any hang-up under 15 seconds), so a caller who waits longer and then hangs up shows as missed instead.
- CallTrackingMetrics drops abandons under its service-level threshold from the service-level number only, not from the abandonment rate.
- Read your own system’s definition before comparing it with anyone else’s.
Outside data
-
US contact centers held call abandonment in a 5% to 7% band from 2012 through 2023, ending at 7.1% in 2023. ContactBabel, undated page, checked Sep 25, 2026 https://assets.ringcentral.com/us/report/us-dmg-2024.pdf
-
COPC, a call-center standards and certification firm, says a 3% to 5% abandonment target is typical, and recommends choosing that target first, then deriving your service level from it, rather than the other way around. COPC Inc., Sep 12, 2024 https://www.copc.com/service-level-are-you-measuring-it-the-wrong-way/
How to pull it
CallRailCompute it from the export
- Open the call log (Analytics, then Activity, then Calls, then Call Log) and read the status column: CallRail marks any hang-up under 15 seconds Abandoned, and anything that rings out with no voicemail Missed.
- Export the log (Edit Columns, then Export, choosing Excel) and count Abandoned rows against all inbound rows.
From CallRail: Call answer status, checked Sep 25, 2026.
CallTrackingMetrics
- If your plan has Queues (Sales Engage, Enterprise, Growth or Connect), open Reports, then Queue Report, for Total Abandoned and its documented service-level formula.
- Without Queues, open Activities, then Calls, and use the Missed Calls filter or the dial_status field in an export; CTM counts a hang-up as missed with no separate abandoned status outside queues.
From CallTrackingMetrics: Queue Report, checked Sep 25, 2026.
RingCentral
- Open the Analytics Portal, then Business Analytics, for the # Abandoned metric: inbound calls where the caller hung up before reaching a live agent, in a queue or on hold.
- Calls ending at your IVR or voicemail are not counted here, unlike the older legacy Performance Reports KPI of the same name, which does include voicemail.
From RingCentral: Business Analytics Data Dictionary, checked Sep 25, 2026.
Dialpad
- Open Analytics, then the Calls tab, for % of Abandoned. The Unanswered tab shows # of Short-abandoned.
- The default short-abandoned threshold is 10 seconds; only Contact Center accounts can change it.
From Dialpad: Analytics Glossary, checked Sep 25, 2026.
Zoom Phone
- With the Customer Engagement Pack add-on, open Reports, then Zoom Phone, then Call Queue analytics, for the Abandoned KPI and the Realtime tab’s short and long abandon thresholds.
- Without that add-on, use the call log’s Call Result filter for Abandoned instead.
From Zoom: Call queue analytics KPIs glossary, checked Sep 25, 2026.
Nextiva
- With Call Center Pro, Enterprise or Power Suite, open the Supervisor Dashboard, then Reporting, then the Abandoned Call Report, for Calls Abandoned and Calls Abandoned in X secs.
- Set your own threshold under Performance Parameters, then Service Level Options, including Abandoned Call Interval.
From Nextiva: Abandoned Call Report, checked Sep 25, 2026.
8x8
- Open Analytics for 8x8 Work, then Company Summary, for Total Abandoned Calls and Average Abandoned Time.
- With Supervisor access, the Call Queues dashboard adds Abandoned %, computed as Abandoned in Waiting divided by Queue Entries.
From 8x8: Glossary of Terms for Analytics for 8x8 Work, checked Sep 25, 2026.
Any system: the universal method
- From the call log, count calls that ended with no person answering and no voicemail left, after the first few seconds.
- Divide by all inbound calls.
Service level
The share of calls a person answers within a target number of seconds you choose.
It turns speed to answer into a promise you can check every hour, not just once a month.
How to count it: Calls answered by a person within your target seconds, divided by all calls offered, after you exclude your short-abandon band. Name both the target seconds and the formula: a widely cited handout runs one set of call data through four common formulas and gets service levels from 82% to 94%, so the formula alone can move your number by more than ten points.
Watch out: Reporting a daily average that hides the two hours a day when almost nobody gets through. Measure by half-hour interval, the way call centers do, not as one number for the whole day. And “80% of calls answered within 20 seconds” is a common convention, not a standard every desk must hit; choose your own target seconds on purpose, and say what it is whenever you report the number.
How to pull it
CallTrackingMetrics
- If you have Queues, open Reports, then Queue Report, for Activities Below Service Level, computed as answered_under_sla divided by the sum of answered_under_sla, answered_over_sla and abandoned_over_sla.
From CallTrackingMetrics: Queue Report, checked Sep 25, 2026.
RingCentral
- With Live Reports licenses, add the Service Level widget, or check the legacy Analytics Portal’s % SLA KPI.
- Where the threshold seconds are set is not documented publicly; ask your RingCentral admin or reseller.
From RingCentral: Business Analytics Data Dictionary, checked Sep 25, 2026.
Zoom Phone
- With the Customer Engagement Pack add-on, open Call Queue analytics and set your Service Level Threshold under the Overview tab, or under Inbound Monitor for a real-time view.
From Zoom: Managing Call Queue analytics, checked Sep 25, 2026.
Nextiva
- With Call Center Pro, Enterprise or Power Suite, open the Call Center Report, set your own Service Level seconds, and read Calls Answered in XX secs and % Calls Answered in XX secs.
From Nextiva: Call Center Report, checked Sep 25, 2026.
Any system: the universal method
- From the call log, count calls answered by a person within your target seconds, after excluding your short-abandon band.
- Divide by all inbound calls, and group by half-hour interval, not just by day.
Handle time
How long one call takes from hello to done: talk time, plus any hold, plus the notes and follow-up after the call.
Handle time decides how many calls one person can take in an hour. Read it separately for calls that signed and calls that did not: long calls that sign are the job, long calls that do not are the leak.
How to count it: Talk time, plus hold time, plus after-call work or wrap-up time, averaged per answered call. Report signed and not-signed calls separately, because a 20-minute sign-up conversation and a 2-minute unqualified call are different jobs, and blending them erases both.
Watch out: Using talk time alone. The minutes spent on notes after the call are real minutes no one can answer the phone.
- Check your units before you compare systems: Dialpad exports duration in minutes, not seconds.
- RingCentral’s handle time includes hold but not park.
- Customer-service handle-time norms of six to seven minutes also do not transfer to a sign-up call; treat it as its own kind of call.
Outside data
-
Across 190,702 entries into a public Erlang-C staffing calculator, the average handle time entered was 6 minutes and 3 seconds. Vendor dataCall Centre Helper, Jan 23, 2026 https://www.callcentrehelper.com/industry-standards-metrics-125584.htm
-
US contact centers’ average service-call duration rose from 306 seconds in 2012 to 442 seconds in 2023. ContactBabel, undated page, checked Sep 25, 2026 https://assets.ringcentral.com/us/report/us-dmg-2024.pdf
-
Personal-injury firm Morgan & Morgan states its sign-up calls average 20.5 minutes, first hello to signed retainer, run as one continuous conversation rather than a callback. SecondhandScaling Law Firms, Jul 23, 2026 https://scalinglawfirms.com/personal-injury-law-firm-kpis/
-
Bench standard · a target, not an average
Under 10 minutes to qualify a car-crash case; the sign-up itself can run longer.
How to pull it
CallRailCompute it from the export
- Open the Calls by Agent report for Total Duration and Average Duration. CallRail has no hold or after-call-work field, so this is talk plus any hold, without wrap-up.
From CallRail: Calls by Agent report, checked Sep 25, 2026.
CallTrackingMetrics
- With Connect, Sales Engage or Enterprise, open Reports, then Agent Activity, for Talk Time, Hold Time, Wrapup Time and Handle Time per agent.
From CallTrackingMetrics: Agent Activity, checked Sep 25, 2026.
Dialpad
- Open Analytics, then the Duration tab, for Average hold time, Average wrap up time and Average handle time; the glossary defines AHT as average call duration plus wrap-up time, divided by calls.
- Exported duration fields are in minutes, not seconds.
From Dialpad: Analytics Glossary, checked Sep 25, 2026.
Zoom Phone
- With the Customer Engagement Pack add-on, open Call Queue analytics for Handle Time (Avg.): the sum of talk, hold and wrap-up time, divided by calls.
From Zoom: Call queue analytics KPIs glossary, checked Sep 25, 2026.
Nextiva
- With Call Center Pro, Enterprise or Power Suite, open the Agent Summary Report for Average Talk Time, Average Hold Time and Average Handle Time.
From Nextiva: Agent Summary Report, checked Sep 25, 2026.
8x8
- Open Analytics for 8x8 Work, then Company Summary, for Average Talk Time, which already includes hold. The Call Queues dashboard also shows wrap-up time, capped by default around 75 seconds.
From 8x8: Glossary of Terms for Analytics for 8x8 Work, checked Sep 25, 2026.
Any system: the universal method
- From the call log, take the duration of each answered new-caller call.
- Join to your intake list to mark which calls became signed cases.
- Average each group.
Occupancy (often called utilization)
The share of a person’s logged-in time spent handling calls, rather than waiting for the next one. Contact centers call this “occupancy”; a related number, “utilization,” instead divides logged-in time by the whole scheduled shift, breaks and meetings included, and the two words get used loosely on vendor blogs.
Too low and you are paying for idle time. Too high and there is no one free when the next call rings, because calls arrive in bunches, not evenly.
How to count it: Occupancy: time spent handling calls (talk, hold and after-call work) divided by time logged in and available to take the next one. Say which of occupancy and utilization you are reporting, since they answer different questions.
Watch out: Planning the desk around the average. Calls bunch up, so a desk that looks fine on average still leaves callers ringing in the busy minutes.
- A small 2 to 4 person intake desk also runs a mathematically lower occupancy than a large call center at the very same speed of answer; call-center trainers give examples around 85% to 90% for large agent groups against 55% to 70% for small ones, so comparing your desk to a big center’s number will make a healthy desk look idle.
- Long after-call work, from a slow CRM or typed-out notes, inflates this number without anyone actually being less busy.
Outside data
-
A contact-center management study guide notes that agents tend to start burning out around 90% occupancy once it holds for several half-hours in a row, though it does not name the underlying studies. Vendor dataBrad Cleveland & Debbie Harne, Mar 1, 2025 https://www.bradcleveland.com/wp-content/uploads/2025/03/INSIGHTS-Study-Guide-for-CC-Mgmt.pdf
-
A widely used call-center training program’s own example occupancy ranges run 85% to 90% for large agent groups against 55% to 70% for small groups, the opposite of one number every desk should hit. ICMI, Jul 15, 2011 https://www.icmi.com/files/StudentResourcePage/CCMetrics/Slides_VC654.pdf
-
In a 2022 panel of call-center consultants, 75% to 80% occupancy was cited most often as optimal, several panelists called 90% unsustainable, and panelists suggested 70% to 75% for centers under 100 agents, closer to the size of a PI intake desk. Call Centre Helper, May 17, 2022 https://www.callcentrehelper.com/what-is-the-right-figure-for-contact-centre-occupancy-206495.htm
How to pull it
CallTrackingMetrics
- With Connect, Sales Engage or Enterprise, open Reports, then Agent Activity, for Utilization (busy versus logged-in time) and Occupancy (busy versus Ready time).
From CallTrackingMetrics: Agent Activity, checked Sep 25, 2026.
RingCentralCompute it from the export
- Base RingEX has no occupancy percentage; Live Reports’ Agent Details widget shows only a status-duration column.
- RingCX, RingCentral’s separate contact-center product, does report Voice Agent Occupancy Rate and Voice Agent Utilization natively.
From RingCentral: RingCX Analytics data dictionary, Metrics, checked Sep 25, 2026.
Dialpad
- With Contact Center, open Analytics, then Reports, then Agent productivity, for the Agent Productivity Report’s Occupancy rate and Avg handle time.
From Dialpad: Agent Productivity Report, checked Sep 25, 2026.
NextivaCompute it from the export
- With Call Center Pro, Enterprise or Power Suite, open the Agent Summary Report and compute it yourself from the Staffed, Talk, Hold and Unavailable Time columns. Nextiva does not publish a ready-made occupancy percentage.
From Nextiva: Agent Summary Report, checked Sep 25, 2026.
Any system: the universal method
- Add up each person’s handle time for the week.
- Divide by the hours they were logged in to take calls.
After-hours coverage
The share of your calls and leads that arrive when your desk is closed, and what answers them.
Accidents do not keep office hours. If a real share of your leads arrive at night and on weekends, that is where your answer rate and speed to lead are decided.
How to count it: Calls and leads that arrived outside staffed hours, divided by all calls and leads. Then the live-answer rate for those calls on its own, split human versus AI or answering service if you use one.
Watch out: Counting an answering service or AI agent that only takes a message as coverage. Coverage means the caller can be qualified, not just recorded.
Outside data
-
In a CallRail survey of 100 legal professionals in October 2025, 68% of respondents said attorneys or paralegals handle intake alongside their caseloads, and firms with dedicated intake staff were less likely to report business lost to missed calls, 34%, than the overall sample, 52%. Vendor dataCallRail, undated page, checked Sep 25, 2026 https://cdn.mediavalet.com/usva/callrail/xS7CApAuhkCQWm0WiBuM9w/KyoWZCfwC0ynj6--yGRAcw/Original/CallRail_2026%20Legal%20Marketing%20Outlook%20Report.pdf
-
Google’s own help pages recommend enabling message or booking leads so people can reach a firm at nights and weekends, when the phone itself will not be answered. Google Local Services Help, undated page, checked Sep 25, 2026 https://support.google.com/localservices/answer/7527305
How to pull it
Any system: the universal method
- Mark each call and lead by whether it arrived inside or outside staffed hours.
- Count each group and compare their answer rates.
Missed-call callback time
How many minutes pass between a missed call and the firm calling that person back.
A missed call is not lost yet. How fast you call back decides whether it becomes one.
How to count it: For each missed call from a new caller: the time of the firm’s first outbound call to that same number, minus the time the call was missed. If the same number missed more than once in a short window, measure from the first miss. Report the median and the slowest 10%, and split business hours from after hours.
Watch out: No phone system in this guide measures this for you. Every “callback” feature these systems advertise, including Dialpad, RingCentral, Zoom, Nextiva and Five9, means a caller in a queue asking to be called back, not the firm returning a missed call, so this stays something you compute yourself.
- A callback placed from someone’s personal cell phone will not show in the office phone system’s log.
- A callback placed through your intake CRM’s own dialer often shows a different caller ID and will not match either.
- If the caller rings back first and reaches you before you call out, that is a lead you did not save, not a callback, so label it separately.
How to pull it
Any system: the universal method
- Export inbound and outbound calls with phone numbers and times.
- For each missed inbound call, find the next outbound call to the same number.
- Minutes between them is the callback time.
The share of new leads the firm actually speaks with, a live two-way conversation, within a window you set (seven days is common).
Every lead you never reach is a lead you paid for and never got to sign. It is usually the biggest gap between the marketing report and the case count.
How to count it: Leads with at least one live two-way conversation within your window, divided by all new leads in the period, with spam and duplicates removed. Report phone-in leads separately from form and message leads: a caller is reached by definition, and blending the two inflates the number.
Watch out: Counting a voicemail left, an automated text, or an email sent as contact. Contact means a conversation. Per-attempt connect rate is also not the same thing as per-lead contact rate; a lead that took five attempts to reach still counts once, as reached.
Outside data
No credible public contact-rate benchmark exists for law firms, or for any industry, that we could verify. Set your own baseline: pull this month’s number, then judge next month against it.
How to pull it
Lead Docket (Filevine)Compute it from the export
- Lead Docket’s default status Chase means “contacted and not heard back from.” A lead that moves past Chase to any status other than Lost has been reached.
- Export the Leads Status History report for each lead’s status changes to compute this.
From Lead Docket: Lead Statuses, checked Sep 25, 2026.
LawmaticsCompute it from the export
- Add a Contacted stage to your Intake Pipeline and turn on Track Time in Stage for it, under Settings, then Pipelines. Tracking only starts from when you turn it on.
- Build a Custom Report with the Stage and Time in Stage columns to see who reached that stage and when.
From Lawmatics: Track Matters’ Time in Pipeline Stages, checked Sep 25, 2026.
Any system: the universal method
- Export last month’s leads with phone number and created time, and your call log with phone number, time and duration.
- Count a lead as reached if your intake system shows a conversation, or if the call log has a call of 2 minutes or more with that number after the lead was created. Say which rule you used.
- Contact rate = reached leads, divided by all leads.
Sign rate
The share of leads that become signed cases.
It is the number your marketing is finally judged on. Read it two ways: out of all leads, and out of qualified leads, because they answer different questions.
How to count it: Signed cases divided by all leads, and signed cases divided by qualified leads. Label each chart with its denominator; that label is usually the single biggest reason two firms’ “sign rates” disagree. Track the leads you turn down at intake, with a reason, as their own line. Count leads by the month they arrived, not the month they happened to sign, and give recent months time to mature before you judge them.
Watch out: Comparing a sign rate of qualified leads with someone else’s sign rate of all leads, or of leads already marked ready to sign. A stage-specific conversion number like that looks nothing like a lead-to-sign rate, even from the same desk.
How to pull it
Lead Docket (Filevine)
- Open the Intake Dashboard for Signups today, this month and last month, alongside Leads and Wanted Leads.
- For a source or period breakdown, use Leads by Marketing Source or Leads Over Time, filtered to All Leads, Wanted Leads or Signed Leads.
From Lead Docket: The Intake Dashboard, checked Sep 25, 2026.
Clio (Grow and Manage)
- Open Reports, then Lead source revenue report, for Hired conversion rate by source.
- Or open Reports, then Matters added report, for Conversion rate, hired divided by all matters added, by source or matter type.
From Clio: Clio Grow Reports, checked Sep 25, 2026.
SmartAdvocate
- Open the Intake Performance Dashboard for Sign Up (wanted, divided by inquiries), Conversion (retained, divided by wanted) and Overall Retained (retained, divided by inquiries).
- Quote the dashboard’s own definitions, not just its column labels: “Sign Up” here is a wanted percentage, not a signed one.
From SmartAdvocate: Intake Performance Dashboard, checked Sep 25, 2026.
Lawmatics
- Open Reports, then Analytics, then Stats by Source, for New Leads, New Clients and Conversion Rate by marketing source.
From Lawmatics: Stats by Marketing Source, checked Sep 25, 2026.
MyCase
- Open Leads, then Leads Insights, for leads converted this month against all time, and leads added this month against all time.
From MyCase: Leads Overview, checked Sep 25, 2026.
Any system: the universal method
- Export last month’s leads with status and signed date.
- Signed out of all = signed, divided by all leads. Signed out of qualified = signed, divided by leads not turned down.
Drop rate
The share of signed cases the firm later drops, read about 90 days after signing.
There is no right drop rate, only a right drop rate for the strategy you chose. Strict sign criteria mean few drops, but cases left on the table. Broad criteria (sign now, evaluate later) mean more cases signed, but more drops and more staff time on cases you will release.
Read it next to sign rate. Broad criteria push both up together.
How to count it: Cases signed in a month that were dropped within about 90 days, divided by all cases signed that month. Read it once the 90 days have passed.
Watch out: Reading last month’s drop rate this month. Most drops have not happened yet.
Outside data
No reliable public number exists for law firms. Set your own baseline from cases signed at least 90 days ago, then judge each month against it.
-
Our read · not an industry average
Choose your sign criteria on purpose, write them down, and check that your drop rate matches the strategy you chose. We see roughly 9 to 11% as a healthy drop rate for a balanced desk. Plenty of operators call that range healthy too. It is our read from the desks we have run, not an industry average.
When it is bad, look here first
How to pull it
Any system: the universal method
- Export cases signed three months ago with their current status.
- Drop rate = cases now dropped or released, divided by all cases signed that month.
Time to sign
How long it takes from a lead arriving to the retainer being signed.
The longer a case waits to sign, the more chances it has to sign somewhere else.
How to count it: Retainer signed time minus lead created time, for each signed case. Report the median, in hours or days, split by channel and case type where you can; a catastrophic case and a routine one do not sign at the same speed.
Watch out: Averaging in a handful of cases that took months. Use the median. Also, only signed cases have a time to sign: show how old your still-open leads are alongside it, or the number will look better every month simply because the slow ones have not signed yet.
How to pull it
Lead Docket (Filevine)Compute it from the export
- Export the Leads report for each lead’s Created Date, and the Leads Status History report for the Changed Date it moved to Signed Up.
- Subtract the two for each lead.
From Lead Docket: Leads, checked Sep 25, 2026.
Clio (Grow and Manage)Compute it from the export
- Open Reports, then Matter revenue report, for Created at and Status date.
- Status date holds only the last status change, so this works only for matters that have not changed status again since Hired; check a sample first.
From Clio: Clio Grow Reports, checked Sep 25, 2026.
SmartAdvocateCompute it from the export
- Run the Case Status Change report for the timestamp each case moved into its signed or retained status.
- Compare it to the case’s open date and time from the Daily Cases Opened report.
From SmartAdvocate: Standard Reports, checked Sep 25, 2026.
Lawmatics
- Use Convert to Hired: it logs the conversion date and tracks Days to Close, from creation to conversion, automatically.
- For a full distribution rather than one summary number, build a Custom Report with Created At and Converted At columns instead.
From Lawmatics: Matter Statuses, checked Sep 25, 2026.
MyCaseCompute it from the export
- Open Leads, then the Converted tab, for created at and converted at on every converted lead.
- Click Export this data to a CSV file and subtract the two columns.
From MyCase: Leads Overview, checked Sep 25, 2026.
Any system: the universal method
- Export signed cases with lead created time and retainer signed time.
- Time to sign = signed minus created. Take the median.
Text response time
How many minutes pass between a potential client texting the firm and a person texting back.
Many people would rather text than call, especially at night. A text that waits until morning is a lead that waited until morning.
How to count it: For each inbound text thread from a new contact: first reply by a person minus first inbound text. Report the median.
Watch out: Counting an automatic reply as the response.
How to pull it
Any system: the universal method
- Export text threads with message times and who sent each one.
- For each new contact, minutes from their first text to your first human reply.
AI-answered share and handoff rate
The share of calls an AI agent answers, and how often it passes the caller to a person.
AI answering can cover nights and overflow. These two numbers show whether it is covering the gaps or quietly handling calls a person should have taken.
How to count it: Calls answered by the AI agent, divided by all inbound calls. Then calls the AI passed to a person, divided by calls the AI answered.
Watch out: Judging the AI on calls answered. Judge it on the cases those calls became.
How to pull it
Any system: the universal method
- From the AI vendor’s call log, count calls it answered and calls it transferred.
- Divide by all inbound calls, and by calls it answered.
E-sign completion
The share of retainers sent for e-signature that get signed, and the share signed within 24 hours.
A retainer that is sent but not signed is a case you have almost won. This number shows how many get stuck at the last step.
How to count it: Retainers marked Completed or Signed, not merely Sent or Viewed, divided by retainers sent. Then retainers signed within 24 hours of sending, divided by retainers sent.
Watch out: Counting a retainer as sent when it went to a wrong or old email address, or counting one a client only opened, a Viewed or Partially Viewed status, as if it were signed.
How to pull it
Lead Docket (Filevine)
- If you send retainers through Vinesign, open Vinesign’s own Reports tab for a bar graph of documents by status (Sent, In Progress, Signed) in a date range.
- Whether a signed-in-Vinesign timestamp is a selectable column in Lead Docket’s own reports is unconfirmed; use the Reports tab counts directly for now.
From Filevine: Vinesign Admin Tools, checked Sep 25, 2026.
Lawmatics
- Open Docs & Forms, then E-Signatures, for each request’s status, Sent, Processing, Viewed, Partially Viewed, Partially Submitted, Completed, Expired or Canceled, plus the date sent and the last update.
- Whether these are selectable custom-report columns is unconfirmed; count directly from the list for now.
From Lawmatics: Managing E-Signature Requests & Deadlines, checked Sep 25, 2026.
Any system: the universal method
- Export e-sign envelopes with sent time and completed time.
- Completion = completed, divided by sent. Within 24 hours = completed within 24 hours of sent, divided by sent.
Graded-call share
The share of intake calls scored for quality, by AI or by a person, against a written rubric.
You cannot coach calls nobody listened to. AI transcription makes it realistic to score every call instead of a handful a week.
How to count it: Calls with a quality score, divided by all answered new-caller calls. Score against a rubric with a few auto-fail items (for example, never asking whether the caller is already represented by another attorney), not just a single average number; an average alone can hide a real compliance miss.
Watch out: Scoring only the calls someone flagged, which are never typical. One widely cited survey of call-center calculator users found 41% of centers monitor fewer than four calls per agent a month, too small a sample to mean much. Also remember California and a few other states carry two-party consent rules for recording and AI scoring; check before you switch it on.
Outside data
-
A widely cited survey of call-center calculator users states a norm of random scoring of 4 calls per advisor a month, with quality scores ordinarily 75% to 90%, and found that 41% of centers monitor fewer than four calls a month. Vendor dataCall Centre Helper, Jan 23, 2026 https://www.callcentrehelper.com/industry-standards-metrics-125584.htm
-
A quality-assurance industry average score of 85% is commonly cited, with 90% to 99% considered good. SQM Group, Dec 7, 2022 https://www.sqmgroup.com/resources/library/blog/7-essential-customer-service-metrics-and-how-you-measure-them
-
In a CallRail survey of 100 legal professionals in October 2025, 22% of respondents said their firm uses call tracking or recording. Vendor dataCallRail, undated page, checked Sep 25, 2026 https://cdn.mediavalet.com/usva/callrail/xS7CApAuhkCQWm0WiBuM9w/KyoWZCfwC0ynj6--yGRAcw/Original/CallRail_2026%20Legal%20Marketing%20Outlook%20Report.pdf
How to pull it
CallRail
- On Premium Conversation Intelligence, open Settings, then Workflow, then Manage, then AI features, for call summaries, sentiment, keyword spotting and Lead Score (Good, Fair or Poor, out of 100).
- Export the call log as “Excel, with transcriptions” (Premium Conversation Intelligence only) to see scores alongside every call.
From CallRail: AI features, checked Sep 25, 2026.
Dialpad
- With an AI Scorecards add-on (Contact Center or Coaching Teams), grade calls by AI, by suggestion, or turn AI off and score manually.
From Dialpad: AI Scorecards, checked Sep 25, 2026.
8x8
- With the Quality Management and Speech Analytics (QM/SA) add-on, run speech and text analytics and quality evaluations on calls answered directly.
- QM/SA cannot transcribe calls answered through a queue or ring group, which is how most intake teams answer, so check whether your setup qualifies before relying on it.
From 8x8: Quality Management and Speech Analytics Support for 8x8 Work, checked Sep 25, 2026.
Any system: the universal method
- Count answered new-caller calls with a recorded score in your call tracking or QA tool.
- Divide by all answered new-caller calls.
