Find the step where people give up, read what they asked Rendel on that screen, and open the fix in one tap. Rendel is the AI assistant inside your mobile app and website, and the SDK that runs it also records screens and taps, so funnels, retention and the month’s results sit in the same console.
Halden’s console, an invented store with Rendel running. Yours for two hours, no sign-up.
Analytics
Rendel suggests
What to fix first, found in your app's data overnight. Each one is a tap away.
Where people stop
64 people stopped at Checkout, and some asked Rendel about it there first.
People who got an answer there went on at 74%; those who did not, at 31%.
A comparison, not proof.
Questions to answerHow much does shipping cost to my address?
Your analytics tool shows that people leave at checkout. Your chat log shows them asking what shipping costs. Nobody sees both at once, so the fix waits until someone exports two files and lines them up.
The drop and the reason sit in different tools.
Checkout −36%
40%
Among US online shoppers who abandoned a cart, leaving out those who were just browsing, 40% said extra costs such as shipping, taxes or fees were too high.
Product analytics counts the drop at a step. It never heard what people asked there, so the team guesses at the fix.
A separate analytics tool means its own SDK, a tracking plan to write and keep up, another vendor to review and another bill.
When leadership asks what the assistant did for the business, a chat dashboard can count conversations and resolutions inside the chat, but not the tasks done in your app, the goals that followed or what that was worth.
Nothing to export and nothing to join. The assistant and the analytics come from the same SDK and land in the same console.
The SDK you add for the assistant records screen views and taps on buttons, links and tabs, with no tracking plan. On the web it starts with the SDK; a mobile app adds one line for screens.
navigatorObservers: [Rendel.navigatorObserver]Opened Product
Tapped Add to bag on Product
Call Rendel.track when someone signs up or buys, mark the name as a goal, and it becomes a funnel step and a figure on Value.
Rendel.track("purchase", { value: 118, currency: "USD" })PurchaseRendel.track("signup")GoalEach conversation is counted, sorted by how it ended, and tied to the screen it was asked on.
How much is shipping?
Overnight, Rendel checks your app’s data by fixed rules, with no model involved, so the morning starts with a short list of what to fix.
Each finding opens its fix: Improve for a question with no answer, Engage for a follow-up, Users for the people. Every figure opens the conversations behind it.
Start the week knowing what to fix first. Each finding comes with its evidence and a button that opens the fix, and it is the first thing your team sees when they open Analytics.
Every night Rendel reads the last 30 days and ranks up to five findings by the people affected and the size of the gap. Set one aside for 30 days or dismiss it, and the next one moves up.
Rendel suggests
What to fix first, found in your app's data overnight. Each one is a tap away.
If their question had no answer, the card opens Improve on it. If they got an answer and still left, it offers a follow-up from Engage instead.
Where people stop
64 people stopped at Checkout, and some asked Rendel about it there first.
Where people stop
22 people stopped at Bag, and some asked Rendel about it there first.
When questions per 100 visits on a screen at least double from one week to the next, it shows here, with a tap to read what was asked.
Questions rising on a screen
Questions on Product doubled: 6 per 100 visits, up from 3 the week before.
41 questions asked there this week.
Ten or more repeated taps on one control in a week, with how many of those people asked Rendel within two minutes after.
Repeated taps
People tapped Apply code on Checkout again and again: 18 repeated taps this week.
From 6 people. Questions asked within 2 minutes after: 4.
Week-by-week return of people who talked to Rendel in their first week, next to those who didn’t, labelled a comparison, not proof.
Rendel and coming back
People who talked to Rendel in their first week came back in week 4 at 22%, against 11% for those who did not.
Of the people who arrived from 1 Sept to 28 Sept, 140 talked to Rendel in their first week and 268 did not.
A comparison, not proof.
Your most tapped controls with no caption, listed so your developer can name them in code with RendelName or data-rendel-name, and every report reads them in your words.
Buttons without a name
Give 3 of your most tapped buttons a name, and every report reads them in your words.
The top one, IconButton, was tapped 189 times.
RendelName('wishlist', child: IconButton(…))Webdata-rendel-name="wishlist"Rendel suggests the funnels that matter: the path to a purchase, one for each goal you mark, and one through a conversation with Rendel. Each shows who started, who finished and where most were lost.
Suggested funnels
Built from the paths people take most. Open one to see where each step loses people.
Active people, how often they come back, visits per person and week-1 retention over the last 30 complete days. Each has an info button that says exactly how it is counted.
Key numbers
The last 30 complete days. The info button on each says how it is counted.
Active people
612
96 a day on average, 241 in the last 7 days
How often they come back
16%
Daily people as a share of monthly
Visits per person
2.4
In the last 30 days
Week-1 retention
27%
New people who came back the week after
How “Active people” is counted
People who used your app in the last 30 complete days, each counted once. A person is the signed-in user, or the device when nobody is signed in. The day and week figures are their own distinct counts, never added up.
Home shows this month’s app users, tasks done and resolved conversations, counted the same way as on Value, each with what it counts beside it and one tap from Value.
Home
November so farSee value
An action is anything Rendel ran for a user, a look-up included. A task is an action that changed something for them, like placing an order.
See why a step loses people, not only that it does: any funnel, retention curve or path can set apart the people who talked to Rendel and show what they asked on that screen. It is product analytics for your mobile app and website, built from the screens and taps the SDK records and the events your app sends.
Pick dates up to 180 days back, the live or test key, a segment, up to five filters and one breakdown. The report is kept in the address, so a link opens the same view for a teammate.
Your question
Showing Last 30 days, Live, Everyone. Change any of it and the report below follows.Showing Last 30 days, Live, Everyone, Platform is ios, by Asked about. Change any of it and the report below follows.
Two to eight steps from screens, taps, events, goals, purchases or a conversation with Rendel, in order or any order, with up to 30 days to finish. Each bar shows people, both percentages and the median time to the next step, and an older SDK is never counted as a zero.
Purchase path
4 steps in order, with 7 days to finish. Percentages are of step 1 and of the step before.
Conversion over time
Taps measured on 94% of active devices. Devices on older SDKs are left out of tap steps, not counted as zero.
On any funnel, see how many finished among people who talked to Rendel after starting and among everyone else, once each side has at least 30 people. The questions asked on each step’s screen sit underneath.
Rendel here
How many finished among people who talked to Rendel after starting, and among everyone else.
A comparison, not proof: people who choose to talk to Rendel may differ from those who do not.
Questions asked here · 51 questions on these dates
What people asked Rendel on Checkout, newest first. Each one is a chance to answer before they give up.
Group people by their first visit or a step they took, choose what counts as coming back, and read it by day, week or month for up to 12 periods, with one curve for people who talked to Rendel and one for those who didn’t.
Who comes back
Each group, period by period
Start from a screen, event, tap or conversation, and see the ten most common routes that follow it or lead to it within one visit, and how the ones through a conversation end.
After asking Rendel
See whether people get back to what they came for once Rendel answers: buying, the next screen, or leaving.
Top paths after a conversation with Rendel
Every event, screen, tap, goal and purchase from the last 30 days, with people and times. Open one for its trend and the questions asked there, and mark a tracked event as a goal from the same row.
Everything your app sends
Events, screens, taps, goals and purchases from the last 30 days. Open one for its trend and the questions asked there.
add_to_cartEventGoal4201,180signupGoal158158add_to_cart · Over time
This is a goal: Value counts it as something people achieved.
The last 100 things people did, refreshed every 5 seconds, with a switch for the test key. Each row names the person only by a short key that changes daily and reads back to no one.
Live
The last 100 things people did, newest first, refreshed every 5 seconds while you watch.
Refreshing every 5 seconds
Save the bar’s filters as a segment, such as people who asked about shipping, count only them in any report, and message them from Engage.
Segments
Follow the same people across every report, and reach them from Send in one tap.
Ask Rendel in the console how many people who reached checkout bought last week. It answers with the numbers and a sentence, opens the same report in Explore, and saves it for the team if you want it again. It writes no SQL: your question becomes the same read-only report Explore runs.
Analytics
Funnels
2 steps in order, with 7 days to finish. Percentages are of step 1 and of the step before.
How many people who reached checkout bought last week?
Of the 46 people who reached Checkout last week, 29 bought: 63%. Most of the others stopped before paying.
ExploreFunnels · Last 7 daysAn example answer: Rendel writes its own words for each question.
Answer “what did the assistant do for us?” with counts you can defend and an estimate that names its sources. What happened is counted from every conversation; what it is worth is estimated from your own numbers or a cited average, and always marked so.
Resolved conversations, tasks done and the estimated support cost avoided, in one line for whoever you report to. The note under it says whose numbers the estimate used.
October in one sentence
In October, Rendel resolved 486 conversations for your users. It did 212 tasks for them in your app. Estimated: about $1,020 in support costs avoided, 1.3× what Rendel cost.
Support costs are estimated with your own numbers.
See the month's value
Resolved
51%
up from 44% last monthTasks done
212
up from 164 last monthSupport cost avoided
$1,020
EstimatedA conversation counts as resolved when it closed after an answer from your knowledge or a task done, with no handoff, no unmet request, no thumbs-down and no failed action left unfixed. Answers from the model alone count as unverified, and small talk is left out.
Resolved
51% up from 44% last month
of conversations ended with an answer from your knowledge or a task done, and no handoff to your team
Counted when a conversation closes, a day after its last message
Where the rest went To your team in the app 38 · To your channels 23
How many tasks Rendel did that changed something for a user, like an order placed, how often an answer took someone straight to the right screen, and what share of your app users asked it something, each against last month.
For your users
Tasks done
212
actions that changed something for a user, like an order placed
up from 164 last month
Screens opened
158
times an answer took someone straight to the right screen
up from 121 last month
Asked RendelLive key only
31%
of app users asked Rendel something: 1,102 of 3,560
Sign-ups, bookings and purchases your app reports, with how many came within 7 days after a conversation with Rendel. Goals your server sends signed show as verified, and purchase value after a conversation is counted, never added to the estimate.
What people went on to do
Goals reached
268
goals and purchases your app reported this month
41 goals and 227 purchases; 88 came within 7 days after a conversation with Rendel
12 of 41 goals came signed from your server. The rest were reported by the app.
Purchases within 7 days of a conversation with Rendel
34%
34% of the $18,420 in purchases your app reported this month, in USD. Counted, not estimated, and not necessarily caused by Rendel, so it is never added to the figures above.
Whether people who talked to Rendel bought or reached a goal more often than those who didn’t. It waits until each side has at least 30 app users, and says plainly that it is a comparison, not proof.
Compared with app users who didn’t ask
See whether people who talk to Rendel convert more
This turns on once at least 30 app users on each side have used your app this month. So far: 12 who talked to Rendel, 28 who didn’t. It counts:
App users who talked to Rendel bought or reached a goal 1.8× as often
14% of the 1,102 app users who talked to Rendel this month bought or reached a goal within 7 days after a conversation, against 7.8% of the 2,458 who didn’t talk to it. Counted on app users whose app reports screens.
A comparison, not proof: people who ask for help may already be keener to buy.
Support cost avoided, the staff hours behind it, what the app cost and the ratio of the two, worked out from four inputs. Each input says whether it is your number, worked out from two others, or a cited average, and one with no confirmed average waits for yours.
For your business
Support costs are estimated with your own numbers.
Support cost avoided
$777$1,020
from about 97 contacts that never reached your support teamabout 20 staff hours, from 170 contacts that never reached your support team
EstimatedWhat this app cost
$799
Value for cost
1.0×1.3×
estimated support costs avoided, set against what Rendel cost
EstimatedHow these are worked out
Support contacts avoided = resolved conversations × the share that would have reached support. Support cost avoided = contacts avoided × the cost of one contact. Staff hours = contacts avoided × minutes per contact ÷ 60, shown beside the cost and never added to it. Value for cost = support cost avoided ÷ what this app cost.
What each answer led to, whether an action, a screen, your knowledge or nothing of yours, counted and never read, with the share answered from your knowledge or an action. Rows with no answer link straight to Questions to answer.
What people came for
What each answer led to: an action, a screen, your knowledge, or nothing of yours. Counted, never read.
Answered from your knowledge or an action
82%
Rated helpful
88%
With answer ratings onRead any of the last 12 months. A month still running is compared with the same days of the month before, and a dated note shows when a change on Rendel’s side, not yours, moved a figure.
Resolved
Our changes in this period
19 Oct: taps and form answers started getting answers from your knowledge too.
24 Oct: a question answered with one of your actions or screens no longer counts as a gap.
An owner freezes a month into a private link with the counts, the estimates and the inputs they came from, and no conversation text or names. It expires in 90 days, can be revoked and prints to PDF. Share links stay off until an owner turns them on.
What happened
What it's worth
Estimated. From the app's own numbers.
Each owner of a live app gets last month’s conversations, tasks done, screens opened and Resolved by email, with a link to Value and a one-click unsubscribe. Owners and admins also get a Monday summary of the week.
Halden in October 2026
On the live key: 1,048 conversations, 212 tasks done for your users, and 158 screens opened from an answer.
486 conversations were resolved, 51% of those counted.
On 19 October, taps and form answers started getting answers from your knowledge too. From that day more conversations can count as resolved, with nothing changed in your app.
Every figure is counted from your app’s own traffic. What it is worth, and the numbers that is worked out from, are on the Value tab.
Open AnalyticsYou get this because you own this app in Rendel. One click below turns it off.
Find where the assistant itself needs you, and open the page that fixes it: how conversations turn into tasks done, which actions fail, what it answered without your knowledge and what users said, over 7, 30 or 90 days. With Support on, it also shows your team’s handoffs; see Human support.
Conversations, then proposed, shown and completed actions, with four headline numbers above. The biggest fall after an action was proposed is highlighted, and you can narrow the funnel to one assistant version.
Conversations
1,012
last 30 days
Completed when proposed
59%
of conversations that proposed an action
Users who asked
864
Median answer time
3.4 s
server time, all versions
Conversation funnel
All versionsThe highlighted step is the biggest fall after an action was proposed.
A failing action opens its failed runs, a confirmation people often decline opens its rules, a subject with nothing of yours behind it opens Improve, and a funnel drop opens the people behind it.
What to look at
What stands out in these days, each with the page that fixes it.
Each action by the name you gave it, with calls, completed, failed, declined and median time, failing ones first, and every number opens the runs behind it. Daily lines show conversations, the action failure rate and the share answered from your content.
Action performance
Conversations per day
Action failure rate
Answered from your content
Views, people and questions asked on each of your most viewed screens, so you can see where people need help in your app.
Screens
Your most viewed screens in the last 30 days, and the questions asked on each.
Questions Rendel answered with nothing of yours behind them, grouped by subject, with whether each is rising and example questions that open their conversations. One tap takes a subject to Questions to answer.
Answered without your knowledge
48 answers had none of your knowledge or actions behind them. Questions are grouped by a word they share.
Each version of your assistant’s persona with its answers, the share of them with nothing of yours behind them, and its helpful rate, plus every thumbs up and down with reasons, linked to the conversations and the people who rated down.
Assistant versions
What users said
412 helpful · 56 not helpful
See who rated an answer not helpfulGo from any number to the people behind it: everyone who talked to Rendel, on your app and your website, in one list, and for each person their conversations, their visits to your app and every answer as they saw it.
Whoever talked last is at the top. Search by name, email, id or something they said, and filter by what happened: an action that failed, an answer rated not helpful, an answer your content didn’t cover. Then message the people you found, in one step from Send, without exporting a list.
All users who asked
1,284
+186 new in the last 7 days
Users who asked
402
Users helped by an action
233
Each visit in the last 30 days, newest first: the screens they opened, what they tapped, the goals they reached, and the moment they talked to Rendel. Their conversations, memory, requests and follow-ups are a tab away.
Emma Hughes
Mobile and webLast seen 2h agoSessions, 30 days 4Reached signup ×1
What they did in your app
Today · iOS · 12 min · 5 screens
Yesterday · Web · 4 min · 2 screens
A transcript puts each question above its answer, drawn the way it looked in your app, in your colours and at phone width, and under it what Rendel did: the action, the rating and what it answered from. Developer details stay folded away.
Transcript
Does the Court Low run small?
Best sellerCourt Low$118Most reviewers say it runs small, so take one size up.
Add it in a 42, please.
Added to the bag
Example questions, failed runs, funnel drops and rated-down answers open the conversations or the people behind them, so a figure is never a dead end. Roles that don’t read conversations see counts and group names instead.
Can I pay in installments?
I can't answer that one yet: Halden hasn't told me about it. I can find you something in the store, check an order or start a return.Get screens, taps and the events that matter without a second SDK or a tracking plan: the Mobile and Web SDK that runs Rendel records screen views and taps, and your app adds the rest with one call. Text typed into fields is never captured.
Each page or screen is recorded by its own name, with the ids in a path replaced, so /orders/8812 reads as /orders/:id and every order counts as one screen.
navigatorObservers: [Rendel.navigatorObserver]Each tap keeps its screen, the kind of control, a locator and whether it was a repeated tap, and on buttons, links and tabs a caption of up to 50 characters. Captions and properties that look like an email, a phone number, a card number, an IBAN, an 11-digit national ID or a run of four or more digits are masked on the device and again on the server.
What Rendel records
Rendel.track counts the steps that matter, with up to 20 flat properties. Mark a name as a goal and it becomes a funnel step and a Value figure; a purchase carries its value and currency; a goal your server sends signed counts as verified.
Rendel.track("add_to_cart", { product_id: "court-low", price: 118 });Rendel.track("signup");Rendel.track("purchase", { value: 118, currency: "USD" });
Everything your app sends
add_to_cartEvent4201,180signupGoal158158purchasePurchase116131Every event belongs to a visit, which ends after 30 minutes idle or in the background, or when someone signs in or out. A signed-in person is one person on every device, and what they did just before signing in on that device moves onto them.
Country: NLIP address: never stored
Halden is an invented store for sneakers, clothing and accessories, with Rendel running in its app for five weeks. This is what its console shows when you open Analytics, step by step. The numbers move each day; the story stays.
Analytics
October in one sentence
In October, Rendel resolved 486 conversations for your users. It did 212 tasks for them in your app. Estimated: about $1,020 in support costs avoided, 1.3× what Rendel cost.
Support costs are estimated with your own numbers.
See the month's value
October in one sentenceIn October, Rendel resolved 486 conversations for your users.
Rendel suggests
Where people stop
64 people stopped at Checkout, and some asked Rendel about it there first.
People who got an answer there went on at 74%; those who did not, at 31%.
A comparison, not proof.
Purchase path
4 steps in order, with 7 days to finish. Percentages are of step 1 and of the step before.
Purchase path
Rendel here
How many finished among people who talked to Rendel after starting, and among everyone else.
A comparison, not proof: people who choose to talk to Rendel may differ from those who do not.
Questions asked here · 51 questions on these dates
What people asked Rendel on Checkout, newest first. Each one is a chance to answer before they give up.
After asking Rendel
Top paths after a conversation with Rendel
Purchase path
For your users
Tasks done
212
up from 164 last month
Screens opened
158
up from 121 last month
Asked Rendel
31%
For your business
Support cost avoided
$1,020
EstimatedWhat this app cost
$799
Value for cost
1.3×
EstimatedWhere people stop
64 people stopped at Checkout, and some asked Rendel about it there first.
People who got an answer there went on at 74%; those who did not, at 31%.
A comparison, not proof.
Purchase path
4 steps in order, with 7 days to finish. Percentages are of step 1 and of the step before.
For your users
Tasks done
212
actions that changed something for a user, like an order placed
up from 164 last month
Screens opened
158
times an answer took someone straight to the right screen
up from 121 last month
Asked RendelLive key only
31%
of app users asked Rendel something: 1,102 of 3,560
Halden’s console opens on Insights. Follow the checkout drop to the funnel, the questions and the paths, then open Value. Yours for two hours, no sign-up.
Both can count a funnel. Only one of them was in the conversation, so only one can put the reason next to the step.
| Question | Rendel | A separate analytics tool |
|---|---|---|
| Why did people leave at this step? | The step, what people asked on that screen, and whether they got an answer | The step and how many left |
| What does it take to start? | The SDK you add for the assistant records screens and taps, with one track call per goal: no second SDK and no tracking plan | Its own SDK, a tracking plan, and another vendor to review |
| Who is behind this number? | The person’s visits with their conversations in place, and each answer as they saw it | A profile with an event stream |
| What do we do about it? | Answer the missing question in Improve or send a follow-up from Engage, one tap from the finding | Export the group to a messaging or support tool |
| What did the assistant do for the business? | Resolved by a written rule, tasks done, goals after a conversation, and an estimated value with every source named | Only what you send it about the assistant |
| Did talking to it make a difference? | Talked to Rendel or not, side by side in funnels, retention and Value, once each side has 30, labelled a comparison, not proof | A join across two datasets, built by hand |
| How does leadership see it? | A frozen month as a private link that prints to PDF, and an email to owners on the 3rd | A dashboard someone builds and shares |
| What leaves the device? | Screen names, taps, the events you send and visit details; never typed text, list-row text, tap position or a picture of the screen | Depends on the tool; replay tools record the screen and mask parts of it |
Rendel has no session replay, heatmaps, A/B tests, feature flags, ad attribution, warehouse export or public analytics API. If you rely on those, or need analytics where Rendel doesn’t run, keep that tool beside it.
Capture is set per app, every person can be left out, and words from conversations reach only the roles that read them.
Owners and admins turn taps on, keep them without labels, or turn them off, keep the taps of all devices, half or a tenth, and list up to 50 screens or CSS selectors where taps keep no label. Every change is in the audit log, and devices pick it up the next time they start.
What Rendel records in your app
Tap capture
OnWithout labelsOffFunnels still show where people tap: each tap is counted by its screen and kind of control, with no caption at all.
Devices whose taps are kept
All devicesHalfA tenthChosen by device, so each sampled person's path stays whole. Screen views, events, goals and purchases are always counted in full.
Places where taps keep no label
None yet. Taps keep their caption everywhere, with emails, phone numbers and card numbers already masked.
#account-balanceSaved. Each device picks it up the next time it starts.
Every role except customer support reads Insights and Explore. An analyst sees counts and group names, and no figure under 5 people; roles that read conversations also see the questions themselves.
3 groups of questions behind this, 51 questions in all, 1 without an answer.
Asked about Pick up in a studiofewer than 5
Asked about Pick up in a studio4
Text typed into fields, a dropdown’s choice, the text of list rows, where on the screen a tap landed, keystrokes, the clipboard, scrolling, any picture of the screen, and anything inside Rendel’s own panel.
When someone declines analytics, your app calls Rendel.setAnalyticsEnabled(false), and their device stops sending screen views, taps, events and sessions while Rendel keeps answering. Your app owns the consent banner; Rendel shows none.
Each person’s screen views, taps and events are kept with no end date unless an owner picks 365, 180, 90 or 30 days. Daily counts stay either way, so no figure already on Analytics drops.
An owner can erase one user: their conversations, notes, goals and devices go, with each device’s screen views, taps and events.
Raw screen views, taps, events and sessions sit in an analytics database in Frankfurt (AWS eu-central-1), run by ClickHouse as a subprocessor; names, daily counts and settings are in Rendel’s main database, also in Frankfurt. None of it is sold or used for advertising, the SDK reads no advertising identifier, and the IP address is never stored.
A shared month has no conversation text and no names, expires in 90 days and can be revoked. The monthly email has no question text and a one-click unsubscribe. Making a link and switching links on or off are in the audit log.
What the SDK records, in fullStorage and retention
| What | Owner | Admin | Developer | Conversation reviewer | Analyst | Customer support |
|---|---|---|---|---|---|---|
| Open Analytics: Insights, Explore, Value, Operations | Yes | Yes | Yes | Yes | Yes | No |
| Read question text and example conversations in reports | Yes | Yes | Yes | Yes | Counts and group names | No |
| Users, a person’s page and transcripts | Yes | Yes | Yes | Yes | No | No |
| Save a funnel or segment; set an insight aside | Yes | Yes | Yes | No | No | No |
| Set the value inputs and goals; choose what is recorded | Yes | Yes | No | No | No | No |
| Choose how long analytics is kept; erase a user; export conversations | Yes | No | No | No | No | No |
| Turn share links on and make one | Yes | No | No | No | No | No |
Open Analytics: Insights, Explore, Value, Operations
Read question text and example conversations in reports
Users, a person’s page and transcripts
Save a funnel or segment; set an insight aside
Set the value inputs and goals; choose what is recorded
Choose how long analytics is kept; erase a user; export conversations
Turn share links on and make one
Value, Operations and Users count conversations on any version of the SDK. Funnels, retention, paths and Insights need the SDK version that records screens and taps, and one call for each goal.
Value, Operations, Users and transcripts count every conversation on any SDK version.
On the web, screen views, taps and visits are then recorded from the moment the SDK starts. A mobile app adds one line for screens: navigatorObservers: [Rendel.navigatorObserver].
Rendel.track("purchase", { value, currency }) for purchases, and Rendel.track("signup") or any event with up to 20 properties for the rest.
In Settings › Value, mark which events count as goals and set your industry, currency and four inputs. Until you do, cited averages are used where one could be confirmed, and the page says so.
Send goals signed from your server for a verified count, turn on answer ratings in Appearance › Message box, let an owner switch on share links, choose what is recorded in Settings › Data & privacy, and wire setAnalyticsEnabled to your consent banner.
The first insights come after the first nightly pass that has enough data, usually within a few days of people using the app.
| What it counts | Mobile | Web |
|---|---|---|
| Value, Operations, Users, transcripts | Any | Any |
| Purchases with track("purchase") | 0.59.0 | 0.38.0 |
| Screens counted, goals with track | 0.62.0 | 0.41.0 |
| Taps, events with properties, visits, the consent switch | 0.63.0 | 0.42.0 |
| Answer ratings | 0.49 | 0.26 |
Current releases: Mobile SDK 0.82.0, Web SDK 0.57.0.
Analytics has no separate price and no add-on. Recording screens, taps and events uses no messages, and each plan includes a monthly allowance of analytics events.
For the screens, taps and events your app sends, it has funnels, retention, paths, breakdowns, segments, a live feed and nightly insights, with what people asked beside them. It has no session replay, heatmaps, A/B tests, feature flags, ad attribution, warehouse export or public analytics API, so keep a tool for those if you rely on them.
Screen views, with ids replaced, from the moment the web SDK starts; a mobile app adds one line for screens. Taps on buttons, links, tabs and other controls: the screen, the kind of control, a locator and whether it was a repeated tap, plus a caption of up to 50 characters on buttons, links and tabs. Goals and purchases take one track call.
No. Typed text, a dropdown’s choice, list-row text, where a tap landed, keystrokes, the clipboard and any picture of the screen are never recorded. In captions and properties, anything that looks like an email, phone number, card number, IBAN, 11-digit national ID or four or more digits in a row is masked on the device and again on the server.
Yes. Rendel.setAnalyticsEnabled(false) stops a person’s screen views, taps, events and visits while Rendel keeps answering. Per app, taps can be on, kept without labels or off, and up to 50 screens can keep no labels. In code, RendelMask and data-rendel-mask drop a caption; RendelIgnore and data-rendel-ignore drop the tap.
In Frankfurt (AWS eu-central-1): raw taps, events and visits in an analytics database run by ClickHouse as a subprocessor, the rest in Rendel’s main database there. It is kept with no end date unless an owner chooses 365, 180, 90 or 30 days, and erasing one user removes their screen views, taps and events.
A conversation that closed after Rendel answered from your knowledge or completed a task, with no handoff to a person, no request it couldn’t meet, no thumbs-down and no failed action left unfixed. Answers from the model alone count as unverified, small talk is left out, and a conversation is counted a day after its last message.
Resolved conversations times the share that would otherwise have reached your support team gives contacts avoided; times the cost of one contact gives the support cost avoided. It is always marked estimated. Each input is your number or a cited average, and one with no confirmed average waits for yours.
No. It shows people who talked to Rendel next to those who didn’t once each side has at least 30, and labels it a comparison, not proof: people who ask for help may already be keener to buy. Purchases within 7 days of a conversation are counted and never added to the estimate.
Yes. Example questions, failed runs, funnel drops and rated-down answers open the conversations or the people behind them, and a funnel step shows the questions asked on its screen. Roles that don’t read conversations, such as an analyst, see counts and group names, and no figure under 5.
Yes, with Ask Rendel in the console. It reads your events, funnels, retention, paths and insights, answers with the numbers and a sentence, and opens the same report in Explore to save. It writes no SQL, it is on by default except for regulated organizations, and it allows 30 analytics reads per organization every 10 minutes.
An owner can turn on share links and freeze a month into a private link with counts, estimates and their sources, and no conversation text or names. It expires in 90 days and prints to PDF. On the 3rd, each owner of a live app also gets last month by email.
No separate price and no add-on: Analytics is on every plan, Pilot included. Recording screens, taps and events uses no messages, and each plan includes a monthly allowance of analytics events, shown in Organization settings › Plan & usage.
Start free on your own, or tell us what your customers do most and we’ll show you that exact workflow, working, in a tailored demo.
Prefer email?
sales@rendel.aiTell us about your app and we’ll show the right workflow in a tailored demo.