Post-call analysis is the work of turning a finished phone call into data a team can act on: what was said, how the call went, and what to do next. On Bland, a finished call produces a transcript and an AI summary, and Enterprise teams can add structured answers to questions they write once.
A team taking calls at volume wants to know whether a caller booked, which plan they asked about, and whether the call was handled well, without playing every recording. That is the job post-call analysis does for a conversational AI platform that answers inbound calls and places outbound calls.
What can a team pull from a single phone call?
A team can pull four kinds of data from one call on Bland: the record of what happened, variables the agent collected, tags that mark where the call went, and, on Enterprise, structured answers to questions the team writes.
Bland's dashboard lists every call, message thread, and web chat under Conversations, where a team can filter and read them, and export calls and message threads. Message threads come from Bland's messaging, which is available on Enterprise plans and the Agent Phone Plan, and the Agent Phone Plan includes SMS only.
Open a call and its call logs show a transcript with speaker labels and timestamps, an audio recording when one exists, and an AI summary of the conversation.
Variables are values the agent extracted during the call, such as party_size: 8 or appointment_time: 9:30 AM. Pathway tags are labels a team places on nodes in Bland's pathway editor, and when the agent reaches a tagged node, the tag appears as a marker on the audio waveform and as a chip in the call's Pathway Tags section.
How do structured answers from a call work on Bland?
Bland's structured answers are called dispositions, and a disposition is a standing question a team asks of every call an agent takes, such as whether the caller booked or which plan they asked about. A team writes the question once, publishes it, and from then on every call the agent handles produces an answer.
Dispositions are an Enterprise feature, and Bland switches them on for enterprise organizations one at a time. Each value in a disposition names where its answer comes from:
- A judge from Evaluations, for questions that need judgement, such as whether a call was handled well.
- A pathway variable the agent already extracts, which Bland re-extracts against the corrected post-call transcript, so the answer can differ from what the agent saw live.
- An extractor, for structured data the agent wasn't already collecting.
- Custom code, for a calculation, or for an answer that has to reach a team's own systems.
A result can carry evidence: the transcript quotes its values cited, each with the speaker and a start and end time in milliseconds.
A team can test a disposition before it goes live. A test run uses real calls the agent has already taken and shows each call, what each value produced, and the evidence behind it. A test run bills exactly like a production run, and it sends no webhook, so nothing downstream sees the results.
Publishing then freezes the draft as an immutable version, and finished runs keep the exact definition they ran under, so editing a disposition never rewrites history.
Delivery timing is a choice. Immediately on conversation end delivers on a 30-minute deadline and suits a downstream system that acts while the call is fresh, and Queued delivers on a 24-hour deadline and suits reporting, QA, and analytics, where nothing is waiting on the result.
Dispositions replace Bland's earlier citations and outcomes. An outcome was deterministic code, so it is now a custom code value, and a citation was an LLM answering a question, so it is now a judge value. Read Bland's dispositions documentation for the full setup.
Where does call data go after the call ends?
Call data goes to a webhook when a team sets one up, and each call is also listed in Conversations in Bland's dashboard. Post-call webhooks are HTTP notifications that Bland sends to a team's server after a call completes, so a team's own systems can react without polling Bland's API. A payload carries complete transcripts, extracted variables, and call duration, among other fields.
Inbound calls report to the webhook set on the phone number they arrive on. For outbound calls, a team passes a webhook parameter, which must start with https://, when it sends the call with an agent. SMS conversations, available on Enterprise plans and the Agent Phone Plan, fire webhooks with the same structure as voice calls, so one endpoint can serve both channels.
Bland's post-call webhooks tutorial walks through a test with a temporary endpoint before a team builds its own.
Two delivery details are worth knowing before a team builds on this:
- Bland doesn't retry a post-call webhook when a team's endpoint returns an error or can't be reached, though a webhook held for a disposition can be retried until its hold deadline. A team resends it with Bland's Resend Post Call Webhook endpoint.
- Disposition follow-up events are retried. Bland waits up to 60 seconds for a 2xx response, so the same event can arrive more than once, and a team deduplicates on the
X-Webhook-Idempotency-Keyheader.
A disposition can also ride in the call's own webhook. Hold original webhook waits for the disposition and sends one combined payload, which can delay that webhook's delivery by up to 30 minutes, and it requires Immediately on conversation end. The delay is in delivering data after a call ends, not in the call itself.
In Bland's dashboard, turning on the Dispositions column in Conversations adds one column per key a team's dispositions produce, so a team can audit a batch of calls without opening each one.
Calls can be exported over a date range of up to 180 days, and Bland emails a link to the file when the export finishes. Bland's integrations connect Bland to other systems. Salesforce, HubSpot, Cal.com, Calendly, Notion, and Slack connect from Bland's dashboard. Anything else connects through Bland's REST API and webhooks, which Bland's forward-deployed engineers can help with on Enterprise.
How do alerts catch a problem across many calls?
Alerts watch call traffic continuously and notify a team when a metric starts trending the wrong way, so nobody has to sample transcripts to find out. Each alert combines a condition, one or more severity levels, and a delivery target, and Bland's alerts tutorial covers setup.
Call length and API errors are available on every plan. Enterprise plans unlock more metrics, including latency, transcription score, silence count, and sentiment score. A team can also write a custom condition in plain language, such as caller asked for a manager, and point a condition at structured signals it already produces, including pathway tags.
An alert has up to three severity levels, INFO, WARNING, and CRITICAL, and each level has its own threshold, a percentage of calls that must violate it, and a lookback window of 10 minutes, 30 minutes, 1 hour, 3 hours, or 24 hours.
The percentage is what keeps alerts quiet: a single long call won't fire a CRITICAL alert, because Bland waits until enough calls in the window are violating. A team can also set a minimum call count so low-volume periods don't produce misleading percentages.
When a level fires, Bland can email one or more addresses, place an outbound voice call to a list of numbers, optionally with a custom spoken message, or invoke a custom tool, which a team can use to route the alert to Slack, PagerDuty, or a webhook.
How does call data feed back into fixing an agent?
Call data feeds a fix in four moves: flag a bad call, find the decision that went wrong, re-run that turn, and let Norm apply and verify a repair.
- Flag the call. Every voice row in Conversations has a Flag button that opens a new issue, with a name, a severity, a description, and a note, or adds the call to an existing issue.
- Find the decision. For a call that ran a pathway, the Detail view shows what the agent decided at each step: which node it chose, whether a node's loop condition was met, which route it took and which routes it left, and what variables it extracted.
- Re-run the turn. Testbed takes one turn from a real conversation, lets a team change what the agent was told, and runs it again to see whether the answer changes. Testbed doesn't replay the call, so a fix that works there still has to be tried against a whole conversation in Evaluations. Running a turn several times shows whether an answer is stable, which is a different question from whether it is correct.
- Let Norm repair it. A team flags a call as a Triage issue and hits Norm Fix, and Norm pulls in the failing call and its agent, reproduces the failure, applies a fix, and verifies it against the original call. Read more about Norm.
Norm, Bland's AI assistant, works from plain English, so the person fixing an agent doesn't need to be an engineer.
How do enterprises govern call data and the changes it drives?
Enterprises govern the agent changes that call data drives with environments and evaluations. Every agent has development, staging, and production environments, and live calls run the pinned production version. When a team edits a pathway it is working on a draft, and live calls are not affected. Evaluations score an agent's calls, and a passing run can be required before a version is promoted.
A published disposition runs in every environment, on every call the agent handles except three kinds that Bland skips without leaving an error: calls made while an organization's billing is blocked or its balance has run out, calls marked for redaction, and calls that have aged past the organization's retention window.
On security and compliance, Bland is SOC 2 Type II, HIPAA compliant with a BAA, GDPR compliant with a DPA, and PCI DSS v4.0.1 self-attested. Bland's Trust Portal holds its SOC 2 Type II report, a penetration test summary, its Data Processing Agreement, and its Business Associate Agreement, and security teams can file a security review request or ask for customer references. Start at Bland's trust and security page.
What does post-call analysis cost on Bland?
Bland's pricing is $0.14 a minute on the self-serve Start plan. Enterprise pricing is custom, set with Bland's sales team. Phone is billed separately, on a customer's own carrier or Bland's at pass-through cost.
Dispositions carry their own cost, and when a team publishes one, the publish drawer estimates the cost per call before the team commits.
A customer result:
Idaho Housing and Finance Association saves $750,000 a year by replacing its IVR with Bland, which handles 4,000 inbound calls a day and cut average call time by 20%.
"We save thousands of dollars a day."
Greg Blake, Chief Information Officer, Idaho Housing and Finance Association
Frequently asked questions
What is post-call analysis?
Post-call analysis is the work of turning a finished phone call into data a team can act on: a transcript, a summary, the values the agent collected, and structured answers to questions about how the call went. On Bland, a team reads that data in Bland's dashboard or receives it on a post-call webhook.
Is post-call analysis available on every Bland plan?
Some of it is. Call length and API errors are available as alert metrics on every plan, and Enterprise plans unlock more. Dispositions are an Enterprise feature.
How soon does post-call data arrive?
A standard post-call webhook is sent right after a call completes. A disposition arrives on a 30-minute deadline when a team picks Immediately on conversation end, and on a 24-hour deadline when it picks Queued.
Can a team test a structured question before it runs on live calls?
Yes, on Enterprise. A team runs a disposition against real calls an agent has already taken and reads what each value produced. A test run bills like a production run and sends no webhook.
Can post-call data reach a team's own systems?
Yes. Bland sends it to a webhook URL the team provides. Inbound calls use the webhook set on the phone number, and outbound calls take a webhook parameter that starts with https://.
How was this page put together?
Each feature here is described from Bland's own pages on dispositions, webhooks, alerts, Conversations, and Testbed, as captured on 20 November 2026, and from Bland's fact file. No calls were tested or timed for this page.
