sample-reviews/*-review.md, with the search string printed in Sources.Chatbot funnel· Conversation design
Chatbot Funnel — What It Is, What Counts as a Step in a Chat, and How Each Builder Counts the Drop-Off
Quick answer: A chatbot funnel is a conversation with a destination and a count. The destination is one outcome the business wants (an email captured, a call booked, an order placed, a broadcast opt-in) and the count is how many contacts got past each message on the way there. The Manychat review describes the product as built for "moving prospects through messenger funnels (welcome → qualification → offer → checkout)," and that four-word chain is the designed sense of the term: an ordered sequence of asks. The measured sense sits on top of it. Landbot's help article, read 14 September 2026, draws the count directly on the flow: "Red arrows show you the drop-off metric, AKA the percentage of visitors leaving the flow at that specific point." Manychat's article says its automation view shows "how users interact with each step, where they drop off, and which parts of your flow need improvement," and Tidio's shows "the amount of visitors who have reached each step of the flow (in percentages)." Same idea, three different counting rules, which is most of what this entry has to explain.
This entry owns the term: its two senses, what a step and a drop-off are when the medium is a chat, how the funnel learns which entry point a contact came through, and how three builders count a step. How to design a funnel that survives a pause between sessions is the multi-step funnel guide.
Funnel, flow, and drip
Three words get used for a sequence of chatbot messages, and they are not the same thing.
| Term | What it is | What it is measured by | Owned by |
|---|---|---|---|
| Conversation flow | The whole graph of what the bot can say and do: every branch, fallback, menu and hand-off | Whether each path works | Chatbot conversation flow |
| Chatbot funnel | One path through that graph, chosen because it ends at an outcome the business wants, with a count of who reached each step | The share reaching each step, and the share reaching the end | this entry (the term) and the multi-step funnel guide (the design) |
| Drip campaign | A sequence of business-initiated messages sent on a schedule after a trigger, with the contact not necessarily replying between them | Delivery, reads, replies, and the stop rule | Drip campaign, drip campaign guide |
The practical difference between the last two is who has to act. A funnel step waits for the contact: it asks a question, offers buttons, requests an email, and nothing further happens until the contact answers. A drip step does not wait; the next message goes out on the clock whether or not the last one was read. A funnel can contain a drip (the follow-up that goes out if the contact stalls at the offer) and a drip can feed a funnel (the reminder whose button starts the qualification), but the count you draw over each is different, because in a funnel every step has a denominator that is the previous step, and in a drip the denominator is the send.
What counts as a step in a chat
On a website a funnel step is a page, and a drop-off is a visitor who left without loading the next one. In a chat there are no pages, so the step has to be defined by what it asks for. Our definition, and the one the three builders' analytics implicitly use: a step is any message whose next message depends on the contact doing something. A menu with three quick replies is a step. A question that expects a typed answer is a step. An email ask is a step. A message that just says something and moves on is not, which is why Landbot's help article notes that "for blocks that send messages (Send a Message block) the drop-off will always be 0%, as these messages are always delivered unless the user leaves before that block." There is nothing to drop off from if nothing was asked.
Three things follow that a page funnel never has to deal with.
A drop-off is silence, and silence has no timestamp of its own. A contact who does not answer might come back in an hour, tomorrow, or never, and the builder decides when to stop counting them as in progress. Tidio's flow analysis shows "the last 7 days' data or the data from the last flow's update (if the date is less than a week ago)," and its page says nothing about how a late answer is attributed; our reading is that a contact who answers on day eight is not in the same count as the question. Landbot's Flow Analytics "counts how many times a user passes through the block, without considering if it was in the same session or not (second visit)," so a contact who restarts is counted twice at the first step. Neither is wrong; they are different questions.
The step can be a hand-off. Tidio's Handoff analysis reports "the points in your flow where visitors abandoned the interaction (e.g. sent a chat message instead of using the flow further)." In a chat the contact can always ignore the buttons and type, and the builder has to classify that as a drop-off, a hand-off, or an answer to be read. The buttons vs free text guide is the design side of that choice; the counting side is that a typed reply at a button step is a different event from no reply, and only Tidio's article says which bucket it lands in.
The funnel is not the conversation. A contact can take three sessions to get through five steps, and the builder's session context (a tag, a custom field) is what lets step four be offered on the third visit instead of step one. Manychat's article gives the mechanism from the other direction: "To ensure your automation only triggers once, you can set conditions based on user actions or use tags that prevent the automation from running again for the same user." The tag that stops a restart is the record of where the contact is.
How the funnel knows where a contact came from
A page funnel gets its source from the referrer. A chat funnel has to be told, and each channel has a way to tell it.
On Messenger the instrument is the ref parameter on an m.me link. Meta's page, updated 23 March 2026 and read 14 September 2026, says the parameter "provides your business with more context about the conversation such as a link on your website versus a link in a store," that such links "can be embedded into QR Codes," and that the value arrives in a messaging_postbacks webhook when the person taps Get Started, or in a messaging_referrals webhook when they already have a thread, in which case the click also "will reset the 24-hour standard messaging window." One caveat sits on the same page: "Referrals on m.me links might not work for some Messenger for Android customers or Pages with georestriction. We do not make guarantees that referrals will always work, and they may be restricted in some instances." A source label that arrives most of the time is still a source label, but the entry-point count is a floor, not a census. The Messenger tutorial lists the other entry points and which of them open the window.
On WhatsApp the equivalent is the referral block that a Click-to-WhatsApp ad attaches to the contact's first message, naming the ad and carrying a per-click identifier; that entry also has Meta's rule for the free 72-hour window a reply within 24 hours earns.
Inside a builder the source becomes a trigger. Manychat's article says an automation's analytics can be filtered "by specific triggers to analyze the performance of associated steps," which is the per-source funnel in one click, and our Manychat review captured a Telegram Ref URL trigger on the Telegram tab of the trigger picker (Messenger was not connected in that session) and lists site widgets, comment-to-DM triggers, story replies and QR codes among its growth tools. Manychat's note that "triggers and message nodes are channel-specific" is the constraint: a Messenger trigger never starts an Instagram automation, so a funnel that spans channels is two funnels with the same steps, and the counts have to be added by hand.
How three builders count a step
The rules below are what each vendor's help center said on 14 September 2026. They are the reason two funnels with the same steps and the same traffic will show different numbers in different tools.
| Builder | Where the count lives | What a step count means | What resets or breaks it |
|---|---|---|---|
| Manychat | The automation view in the Flow Builder | "how users interact with each step, where they drop off"; a Smart Delay block reports Overall, Waiting and Passed, with "Total and Unique counts" on click; a Randomizer splits traffic into "up to six variations" for a test | Analytics can be filtered by trigger; an automation pauses after "30 blocks without a pause for each subscribed contact," where "any block without a button or Smart Delay is considered a block without a pause" |
| Landbot | Analyze > Flow Analytics (the page calls it "still on BETA"), plus Analyze > Metrics for Chats, Finished and Goals | Blue arrows carry "the percentage of people that, when reached that block, followed each path"; red arrows carry the drop-off "at that specific point"; a Goal block counts "the number of users that chose the specific path where the selected Goal block is located," with a rate "calculated according to the total number of users that have interacted with your bot" | Second visits are counted as new passes; editing a published flow produces "N/A" for the changed part until the date filter is moved past the edit; a Jump to within the same bot and the back button also produce N/A; deleting a Goal block deletes its metrics |
| Tidio | Analyze mode inside the flow editor, plus the Flows list | Flow shows "the number of times each trigger started the flow, and the amount of visitors who have reached each step of the flow (in percentages)"; Handoff shows where visitors "sent a chat message instead of using the flow further"; the list shows Executed, Engagement and Handoff for "the last 7 days" | Data covers the last seven days or the period since the flow was last updated, whichever is shorter; a flow with no interactive node has no Handoff metric; "if you don't require any action from the visitor, it will not increase your engagement metric" |
Two things in that table matter more than the rest. Landbot's Finished metric counts "users that reached a block that wasn't connected to any other," which means a flow that ends in a dead-end message counts as finished whether or not the contact got what they came for; the goal is the honest end of the funnel and the last block is not. And Manychat's thirty-block pause rule is a design constraint dressed as a technical one: a funnel that fires more than thirty messages at a contact without once waiting for a button is not a funnel, and the builder pauses it. The tier gates on these views (Typebot's and Chatbase's on Pro, Tidio's fixed reports) are in the reviews section below.
The window is a step
On a website nothing stops step four from being offered a week after step three. On WhatsApp, Messenger and Instagram something does. Manychat's Smart Delay article, updated 3 December 2025 and read 14 September 2026, puts it in one line: "A Smart Delay does not reopen the 24-hour window." It continues: "if a contact interacts with your automation and then enters a Smart Delay lasting longer than 24 hours, a regular message will not be delivered. However, some contacts may still receive the message if they reopen the 24-hour window later, for example, by sending a message in DMs." For WhatsApp the instruction is to "use a Message Template after the Smart Delay," and for Telegram "no additional setup is required, as this channel does not have a 24-hour window limitation."
So a funnel with a pause in it has a step that no builder draws: the contact either re-opens the window themselves, or the next message is a template on WhatsApp, a Utility Template or a Marketing Message on Messenger (the message tags an SMB used for this stopped working in 2026; the Messenger tutorial has the dates and the replacement path), or nothing. The win-back entry sets out what each channel allows when the business speaks first, including Meta's per-user cap on marketing templates and its rule that marketing templates are not delivered to US phone numbers (Meta's page does not address in-window delivery). For the funnel count this means the drop-off after a long delay is partly the contact's decision and partly the channel's, and a report that shows 60 percent lost between the offer and the follow-up is not telling you which. The multi-step funnel guide is built around keeping the asks inside one window and treating anything that has to cross one as a separate, permissioned message.
What the numbers cannot tell you
A funnel report answers one question: of the contacts who reached this step, how many reached the next. It does not say why (a 40 percent drop at the email ask could be the ask, its position, or the audience the entry point brought in; the email capture guide works through the first two). It does not say whether the outcome was worth anything; chatbot conversion rate is the end-to-end figure and the lead qualification playbook is where the denominator gets chosen so it cannot be gamed. And it does not say what a silent contact did next: an abandoned funnel and a contact who bought on the website after reading the offer look identical in the flow view. Our reading is that the per-step count is the most useful number a builder gives you and the least self-explanatory, and that the honest way to read it is against a change: the same step, before and after one edit, with the version rules above respected so the two periods are comparable.
What our fifteen reviews record about funnels
Searched 14 September 2026, case-insensitively, across the fifteen files matched by sample-reviews/*-review.md: funnel appears in 14 (every review except SendPulse). Most uses are the designed sense: the Manychat review on "messenger funnels (welcome → qualification → offer → checkout)" and, among its cons, that "larger funnels with many branches and conditional logic become difficult to maintain and optimize"; the Wati review recording the same complaint from G2 and Capterra reviewers; the AiSensy review on Click-to-WhatsApp as "entry points for the WhatsApp funnel."
The measured sense is Scenario F of our protocol ("custom funnel exploration"), and the results split. Manychat: "custom funnel builder available." Typebot: "Pro tier ($89/mo) unlocks in-depth analytics including funnel analysis, drop-off identification," with a screenshot of per-block drop-off. Chatbase: conversion funnels among the advanced analytics gated to Pro. BotPenguin: "custom funnel builder is available with limited customization options." AiSensy: "not available," the surface "is broadcast-level only." Tidio: "no custom funnel builder," which matches the help article, since Tidio draws the count over the flow you built rather than letting you define a funnel across flows. Wati: "deeper funnel views and cohort analysis require Business tier." Landbot: "custom funnel/cohort depth unverified," in a review whose hands-on session ran on a 14-day Pro trial, although its Metrics screenshot shows a Funnel icon and the help article above documents the feature. The Blip review captured a five-stage funnel (Enviadas, Recebidas, Lidas, Respondidas, Falharam) that is a template-delivery funnel, not a step funnel: it counts what happened to a send, not what the contact did next.
drop-off or drop off appears in only 3 (AiSensy, Tars, Typebot). None of the fifteen protocols ran the same funnel through two builders and compared the counts; that is a gap in the protocol, recorded here, not a finding about the platforms. Corrections to editorial@chatbotscape.com.
Related terms
- Chatbot conversion rate — the end-to-end figure the funnel decomposes, with its benchmarks and attribution window.
- Chatbot abandonment rate — the share of conversations quit before resolution, a support-side cousin of drop-off.
- Drip campaign — the scheduled sequence that does not wait for a reply, and its permission mechanics.
- Chatbot analytics — where the funnel view sits among the other dashboards, with the tier gates by platform.
- Click-to-WhatsApp ads — the entry point that labels the contact and can earn a 72-hour window.
- Lead generation chatbot — the most common funnel an SMB builds, and what it captures.
FAQ
What is a chatbot funnel?
The ordered steps a contact passes through in a conversation on the way to one outcome (a lead, a booking, an order, an opt-in), and the count of how many contacts reached each step. The word covers both the sequence a business designs and the per-step report a builder draws over it.
What counts as a step in a chatbot funnel?
Any message whose next message depends on the contact doing something: a button menu, a question, an email ask. A message that only says something is not a step, which is why Landbot's analytics show a fixed 0 percent drop-off on its Send a Message block. A drop-off is the contact not answering, and each builder decides for itself how long to wait before counting the silence.
Is a chatbot funnel the same as a drip campaign?
No. A funnel step waits for the contact to act; a drip step goes out on a schedule whether or not the last message was read. A funnel can contain a drip as its follow-up and a drip can start a funnel, but the counts are different: in a funnel each step's denominator is the previous step, in a drip it is the send. The drip campaign entry owns the scheduled kind.
How does the funnel know which ad or link a contact came from?
On Messenger, from the ref parameter on an m.me link, delivered in a webhook and usable inside a QR code; Meta says the referral "might not work" for some Android users or geo-restricted Pages, so treat the count as a floor. On WhatsApp, from the referral block a Click-to-WhatsApp ad attaches to the first message. Inside Manychat the source becomes a trigger, and the automation's analytics can be filtered by trigger.
Why does my funnel lose most contacts after a delay of a day or more?
Because on WhatsApp, Messenger and Instagram a delay does not reopen the 24-hour window. Manychat's help states that a regular message after a Smart Delay longer than 24 hours "will not be delivered" unless the contact reopens the window themselves; on WhatsApp the next message has to be an approved template. Part of that drop is the channel refusing the send, not the contact leaving. Telegram has no such window.
Sources
- Meta for Developers, Messenger Platform, m.me Links (developers.facebook.com/documentation/business-messaging/messenger-platform/discovery/m-me-links, stamped updated 23 March 2026), read 14 September 2026: the
refparameter "provides your business with more context about the conversation such as a link on your website versus a link in a store"; "m.me links with ref parameters can be embedded into QR Codes"; delivery inmessaging_postbackson Get Started and inmessaging_referralsfor an existing conversation, which "will reset the 24-hour standard messaging window"; "Referrals on m.me links might not work for some Messenger for Android customers or Pages with georestriction. We do not make guarantees that referrals will always work"; ref length "can not exceeed 2,083 characters." - Manychat Help, How to build a Manychat automation (help.manychat.com/hc/en-us/articles/14281166306332-How-to-build-a-Manychat-automation, stamped updated 3 December 2025), read 14 September 2026: "The automation view offers a clear overview of your stats, showing you how users interact with each step, where they drop off, and which parts of your flow need improvement. You can also filter analytics by specific triggers"; "A single automation can include up to 30 blocks without a pause for each subscribed contact ... Any block without a button or Smart Delay is considered a block without a pause"; the Randomizer's "up to six variations"; "triggers and message nodes are channel-specific"; the run-once rule via conditions or tags.
- Manychat Help, Smart Delay (help.manychat.com/hc/en-us/articles/14281197046812-Smart-Delay, stamped updated 3 December 2025), read 14 September 2026: "A Smart Delay does not reopen the 24-hour window"; the non-delivery rule for delays longer than 24 hours and the exception when the contact reopens the window; "Use a Message Template after the Smart Delay" for WhatsApp; Telegram "does not have a 24-hour window limitation"; the Overall, Waiting and Passed statistics with "Total and Unique counts."
- Landbot Help, Bot's Analyze Section (help.landbot.io/article/j346fejchf-analyze-section, stamped "updated 6 months ago" on 14 September 2026), read 14 September 2026: Chats as "the number of users that started a conversation with the bot"; Finished as "the number of users that reached a block that wasn't connected to any other"; Flow Analytics "still on BETA"; "Block-by-block conversion: That number you see over every blue arrow shows the percentage of people that, when reached that block, followed each path"; "Block drop-off: Red arrows show you the drop-off metric, AKA the percentage of visitors leaving the flow at that specific point"; the Send a Message 0 percent rule; the See drop-offs threshold; "counts how many times a user passes through the block, without considering if it was in the same session or not (second visit)"; the version-conflict, same-bot Jump to and back-button N/A cases; "For WhatsApp: This section behaves exactly the same way as in the Landbot (web) version."
- Landbot Help, Custom Goals (help.landbot.io/article/ftjfbf4ef5-custom-goals, stamped "updated 3 months ago" on 14 September 2026), read 14 September 2026: Goal completed as "the number of users that chose the specific path where the selected Goal block is located in your flow"; Rate "calculated according to the total number of users that have interacted with your bot"; goals reusable across bots; "if you delete a Goal block from the bot flow, the metrics of this goal will also be deleted."
- Tidio Help Center, Flow Performance Analytics (help.tidio.com/hc/en-us/articles/5474508902556-Flow-Performance-Analytics, dated 26 August 2024), read 14 September 2026: Executed, Engagement and Handoff "in the last 7 days"; "if you don't require any action from the visitor, it will not increase your engagement metric"; "Don't create different flows with the same trigger."
- Tidio Help Center, Flow and Handoff Analytics (help.tidio.com/hc/en-us/articles/5465124266396-Flow-and-Handoff-Analytics, dated 29 March 2024), read 14 September 2026: the Edit and Analyze modes; Flow shows "the number of times each trigger started the flow, and the amount of visitors who have reached each step of the flow (in percentages)"; "We present the last 7 days' data or the data from the last flow's update (if the date is less than a week ago)"; Handoff shows "the points in your flow where visitors abandoned the interaction (e.g. sent a chat message instead of using the flow further)"; "not all flows will include the Handoff metric."
- Chatbotscape review corpus (the fifteen platform reviews listed at /reviews), searched 14 September 2026 from the repository root. Denominator:
ls sample-reviews/*-review.md | wc -lreturns 15.grep -liE 'funnel' sample-reviews/*-review.mdreturns 14 (all except sendpulse);grep -liE 'drop-?off|drop off' sample-reviews/*-review.mdreturns 3 (aisensy, tars, typebot). Passages cited:manychat-review.mdlines 165, 253 to 255, 507, 573 to 574 and 762;typebot-review.mdlines 466 and 618;chatbase-review.mdline 193;tidio-review.mdlines 649 and 709;blip-review.mdlines 648 to 658;wati-review.mdlines 404 to 405 and 458;botpenguin-review.mdline 619;aisensy-review.mdlines 839 and 927;landbot-review.mdlines 383 and 744. - Ahrefs Keywords Explorer, US overview, queried 14 September 2026 — the demand, difficulty, CPC and parent-topic figures in this entry's keyword note.
- Chatbotscape evaluation methodology. /methodology (continuously updated).