Analytics · Flows

See exactly where your flow leaks contacts

Flows fail silently. A node times out, a branch is mis-wired, an AI prompt returns garbage — and 30 percent of your funnel walks away without anyone noticing. Flow Analytics overlays per-node success, drop-off, revenue and SLA right on the canvas you built, so the bottleneck is visible the moment you open the flow.

  • Per-node success, drop-off and elapsed time
  • Revenue attributed to specific flow paths
  • SLA breach heatmap on long-running flows
  • Compare flow versions side by side
The problem

Flows fail silently and steal conversions

A WhatsApp flow looks fine until it does not. Contacts enter at the top, conversions trickle out the bottom, and somewhere in the middle 35 percent of the audience vanishes — into a branch that fires the wrong template, a wait node that holds them too long, an AI step that returns an empty answer, or a webhook that 500s and never retries. The funnel metric tells you the conversion rate is 12 percent. It does not tell you that node 7 is responsible for half the loss.

The traditional debugging path is brutal. You re-read the canvas, manually trace a few contacts through it, scroll through logs, build a one-off query against the event store, and three hours later you have a hunch about where the leak is. By the time you fix it, you have run a campaign at 35 percent below its real potential, and the next flow will leak somewhere else.

The fix is to make the flow itself a measurement surface. Every node should publish how many contacts entered, how many succeeded, how many dropped, how long they spent inside, what revenue they touched and which SLA they breached. Then the canvas is a heatmap as well as an editor. You open the flow, see where the red is and act on it — without leaving the page.

What it is

Flow Analytics, in depth.

Flow Analytics turns every node in the Flow Builder into a live measurement point. As contacts flow through, each node publishes counters for entered, succeeded, dropped, timed-out, errored and elapsed time. The canvas overlays these counters directly on the node tile, with a heatmap of drop-off so the leakiest step is visible at first glance. There is no separate analytics view to build — the editor and the analytics are the same surface.

Drop-off is computed properly. If a contact enters node 5 and never reaches node 6 within the flow's timeout, they are counted as dropped at node 5 with a reason: timeout, branch mismatch, errored webhook, agent intervention, opt-out. This decomposition is the difference between "5 percent of contacts dropped at the AI step" and "the AI step times out 5 percent of the time on Hindi prompts longer than 200 characters" — actionable, specific, fixable.

Revenue overlay shows which paths through the flow generate revenue and which do not. A two-branch flow where branch A leads to a Razorpay payment and branch B leads to an agent handoff can be compared on revenue per entrant, with the dollar number rendered on each path. Decisions to promote a branch, prune a step or rewrite a template become grounded in money rather than hand-wave conversion rates.

Flow versions are first-class. When you publish a new version of a flow, Flow Analytics keeps the previous version visible so you can compare entrants, success rate and revenue side by side over time. This is how you prove a flow change worked, instead of guessing. Combined with A/B traffic splits at any node, it gives you a clean iteration loop on automation — not just on campaigns.

Capabilities

Everything you get with Flow Analytics.

7 capabilities
01

On-canvas overlay

Every node shows entered, succeeded, dropped, errored and elapsed time inline on the canvas. A heatmap colour-codes drop-off so the leakiest step is visible without opening any side panel.

02

Drop-off decomposition

Drops are categorised by reason — timeout, branch mismatch, errored integration, agent takeover, contact opted out. Each reason is actionable and links to the underlying conversations for investigation.

03

Revenue per path

Attach a revenue event to a flow (a Razorpay payment, a Shopify order, a custom webhook). The canvas overlays revenue per entrant on each path, so dollar-weighted decisions replace conversion-rate guesses.

04

SLA monitoring

Define an SLA on a flow (must complete within 2 hours) and the analytics surface breach rate per node. SLA-breaching flows can trigger an alert, route to a manager queue or fail safely with a fallback.

05

Version comparison

Compare flow versions side by side on entrants, success rate, drop-off and revenue. Make a change, publish v2, and watch real numbers tell you whether the change worked rather than reading agent anecdotes.

06

AI step diagnostics

AI nodes publish prompt-level metrics — average token count, response time, success rate, refusal rate, fallback rate. Spot a degraded model or a bad prompt template before customers complain.

07

Per-segment breakdown

Compare flow behaviour across segments — paid vs free, new vs returning, India vs UAE. A flow that converts at 18 percent overall might be 28 percent for paid and 6 percent for free, which changes the fix.

Use cases

Built for the way teams actually work.

SaaS
Case 01

Onboarding flow drop-off

A B2B SaaS sees its trial-to-paid conversion plateau. Flow Analytics shows the integration-setup step times out for 22 percent of contacts. A 60-second wait extension and a retry branch lift conversion by 8 percent the following week.

E-commerce
Case 02

Cart-recovery branch optimisation

An e-commerce brand runs a 4-step cart-recovery flow. Revenue overlay reveals branch A (discount) generates twice the per-entrant revenue of branch B (free shipping) for first-time buyers. They reroute the segment and watch revenue lift in real time.

Financial Services
Case 03

Loan-application flow SLA

An NBFC requires loan applications to complete within 24 hours. Flow Analytics surfaces a 14 percent SLA breach at the document-upload step. They add an SMS fallback after 4 hours and breach rate falls to 3 percent.

EdTech
Case 04

AI agent prompt diagnostics

An edtech runs an AI agent flow for course recommendation. Flow Analytics shows the AI step has a 9 percent refusal rate on Hindi prompts. They retune the system prompt for multilingual handling and refusal rate drops below 1 percent.

Healthcare
Case 05

Appointment-booking funnel

A clinic chain runs an appointment-booking flow. Drop-off is concentrated at the time-slot selection step on mobile. They redesign the list response to use a button-based picker and bookings rise 23 percent across the chain.

How it works

From signup to first send in minutes.

Flow Analytics is included on every SabNode workspace. No separate billing, no extra setup — flip it on from your workspace settings.

  1. 01

    Run the flow

    Flow Analytics begins collecting data the moment a flow goes live. No instrumentation needed — every node automatically publishes counters and timing.

  2. 02

    Open the canvas

    Open the flow in the editor. Counters and a heatmap render on each node by default. Switch to revenue or SLA overlay for different lenses.

  3. 03

    Drill into a node

    Click any node to see drop-off breakdown, errored events, elapsed-time distribution and the conversations of contacts who passed through.

  4. 04

    Compare versions

    Open the version compare view to see the metrics of the previous flow version alongside the current one over the same date range.

  5. 05

    Iterate and republish

    Make a targeted fix to the leakiest node, publish a new version and watch real numbers update. Roll back from history if the new version underperforms.

Plays well with

Works with the tools you already ship on.

RazorpayStripeShopifyMetaMixpanelAmplitudeBigQuerySlack
Frequently asked

Questions about Flow Analytics.

Can't find what you're looking for? Talk to our team.

Do I need to instrument my flow for analytics to work?
No. Every node in the Flow Builder publishes counters and timing automatically. The only optional setup is attaching a revenue event (a Razorpay or Shopify webhook) to a flow so the revenue overlay has data to render. Everything else, including drop-off decomposition, works out of the box from the moment a flow goes live.
How is "dropped" defined?
A contact is counted as dropped at a node if they enter the node and do not progress to a downstream node within the flow's timeout window. We further categorise the drop by reason — timeout, branch mismatch, errored integration, agent takeover, opt-out — so the number tells you not just where to look but what to fix.
Can I see analytics for a specific cohort?
Yes. The per-segment breakdown lets you compare a flow's behaviour across any contact segment — paid vs free, new vs returning, by country, by source, by language. A flow that converts at 18 percent overall might be 28 percent for paid and 6 percent for free, which usually changes the fix you would make.
How does revenue attribution work for flows?
Attach a revenue event to the flow — typically a Razorpay payment, a Stripe charge, a Shopify order or a custom webhook from your backend. The platform attributes revenue to the path the contact took through the flow within an attribution window you set (default 7 days). Per-path revenue per entrant appears on the canvas.
Can I A/B test a flow node?
Yes. Any node can be replaced by an A/B split node that routes contacts randomly between two or more downstream paths. Flow Analytics shows per-arm metrics with a significance signal, exactly like multi-step campaigns. The winner can be promoted by editing the split to 100 percent on the winning arm.
How do I monitor SLA breaches in real time?
Set an SLA on the flow (must complete within X minutes or hours). Flow Analytics tracks the percentage of contacts breaching the SLA per node and surfaces it on the canvas. You can also route breaches to an alert channel (Slack, email) or trigger a fallback flow that recovers the conversation automatically.
Does this cover AI Studio steps?
Yes. AI nodes publish additional diagnostics — average token count, response time, success rate, refusal rate, fallback rate to a human. You can see whether a recent model swap or prompt change degraded performance, and you can compare AI step behaviour across languages or segments to catch issues that hide in averages.
Analytics · Flows

Ship flow analytics into production this week.

No credit card. No sales call required. Spin up a workspace, plug in a number, and your team is live in under an hour.