CartSense -Trust-First AI Reordering
REORDERING GROCERIES SHOULD TAKE SECONDS, NOT SCROLLING. CARTSENSE SUGGESTS THE CART. YOU APPROVE IT.
CartSense -Trust-First AI Reordering
REORDERING GROCERIES SHOULD TAKE SECONDS, NOT SCROLLING. CARTSENSE SUGGESTS THE CART. YOU APPROVE IT.
Section 1 · The problem
An AI cart that charges on its own can break trust
Every week, people rebuild a cart they already know. Automation takes that work away. One surprise charge gives it back, and then some, because now the person has to watch the app.
Who it affects: anyone who orders groceries from an app on a routine and pays from the same account each time.
My question: how might we take the weekly rebuild off people without making them feel the app took control of their money? This is my starting guess. Real people have not checked it.
Why I see it this way: I worked as a Team Leader in a BPO. Customers there rarely complained about a feature. They complained about a surprise, and about the edge case nobody planned for. That is why every decision below starts from an edge case.
Where trust breaks
Three questions a user asks the moment an app spends for them.
Fear 1
“Why is this in my cart?”
Every AI pick has a chip and a reason.
Fear 2
“Did the price change on me?”
An orange chip flags every price change.
Fear 3
“Can I make it stop?”
Turn predictions off, reset what the app learned, or block an item.
Section 1 · The problem
An AI cart that charges on its own can break trust
Every week, people rebuild a cart they already know. Automation takes that work away. One surprise charge gives it back, and then some, because now the person has to watch the app.
Who it affects: anyone who orders groceries from an app on a routine and pays from the same account each time.
My question: how might we take the weekly rebuild off people without making them feel the app took control of their money? This is my starting guess. Real people have not checked it.
Why I see it this way: I worked as a Team Leader in a BPO. Customers there rarely complained about a feature. They complained about a surprise, and about the edge case nobody planned for. That is why every decision below starts from an edge case.
Where trust breaks
Three questions a user asks the moment an app spends for them.
Fear 1
“Why is this in my cart?”
Every AI pick has a chip and a reason.
Fear 2
“Did the price change on me?”
An orange chip flags every price change.
Fear 3
“Can I make it stop?”
Turn predictions off, reset what the app learned, or block an item.
Section 2 · Research
What reviews and docs already say
Desk research from docs and reviews. I did not test these apps hands-on, so “not shown” means not shown in the sources I read.
One Misfits Market reviewer, via Onclarity, calls the app “a maze” for turning off auto-shopping. Reviews at 1 to 3 stars across five US grocery apps list substitutions nobody asked for among the main complaints (Unstar). Blinkit review summaries on Google Play list wrong items and out-of-stock products (Kimola).
Instacart
Buy it again list
You build the cart yourself
Pay
Shows what you bought before, not
what changed.
My answer: an orange chip on every
price change.
Blinkit
Reorder
Wrong on missing item
Review summaries list wrong items
and out-of-stock products. I did not
test the app.
My answer: review comes before the
charge.
Misfits Market
Auto - Shopping on
Charged
Hard to turn off
Reviewers say it is hard to turn off, and there is no review step.
My answer: review comes before the
charge.
CartSense
Predicted cart
→
You review
→
You approve
→
Charged
The Review before any change
SECTION · MY ANSWER TO THE GAPS
GAP · HARD TO TURN OFF
Controls in one place.
Turn predictions off, reset what the app learned, block an item. Set a ₹600 spend cap. A price rise after approval always asks again.
GAP · WRONG ITEMS, SILENT PRICE CHANGES
Review before the charge.
Every pick carries a chip and a reason. A price change gets an orange chip. Nothing is charged until you approve.
Section 3 · The insight
Trust breaks when the app acts and the user finds out later.
A wrong pick you can see costs one tap. A wrong pick you find at the door costs the customer.
Section 4 · The solution
A cart that stops and asks
CartSense builds the weekly cart, then stops at a review screen. Nothing is charged until you approve. Each fear from Section 1 gets one answer.
Fear 1
Why is this in my cart?
+ AI Suggestion
Remove
Based on your Tuesday morning routine.
Quaker Oats
Qty: 1 kg ₹72
− 1 +
Every pick shows its reason
A green chip marks what the system added. One line says why.
Fear 2
Did the price change on me?
₹82 → ₹89 since your last order
Epigamia Greek Yogurt
Qty: 1 kg ₹89
− 1 +
Price changes get flagged
An orange chip shows the old price, only when it changed. A rise after you approve needs a new yes.
Fear 3
Can I make it stop?
Turn predictions off
◯
Reset what the app learned
›
Block an item
›
Three controls, one place
All in AI Preferences. You also set a ₹600 spend cap there.
See the screens →
Section 5 · How I got here
How I got here
Option 1
Dropped
Option 3
Kept
Option 2
Dropped
Droppped
Full Automation
A background checkout with no review step. A wrong item reaches checkout unseen, so the feature fails at its one job.
Kept
Review, then approve
The cart stops at a review screen. Nothing is charged until you approve. Only then you see this confirmation, with each AI pick marked and a link to AI preferences.
Droppped
Auto-charge, 30-min cancel
Looks safer, but the money is already committed. You must catch the mistake in time.
Rubber-stamp risk
Two options I designed and dropped, and the one I kept.
The review adds a tap. Fast approval can turn into tapping without reading. I kept the tap, because one wrong charge costs more trust than one extra tap.
The risk
Tap without reading
Approve & pay ₹329
☝ ☝ ☝
Same button every week. The thumb learns the spot.
My defence
Untested
2 changes since last week
Approve
What changed sits next to the pay button, at the moment of paying. Task 3 in the test plan checks it.
Section 7 · The screens
The screens, in the order a week unfolds
1
The screen opens before any charge. Decision 1.
2
Green chip marks what the system added. Decision 2.
3
One line says why. Remove and a stepper sit on each card. Decision 3.
4
Orange chip shows the old price, only when it changed. See When it breaks.
Interface sketch. Item names and prices are illustrative.
The screens, in the order a week unfolds. Item names and prices are illustrative.
1. Home
Home shows Popular near you, with an Add button. Before 2 orders the app is a normal store. No AI marks yet.400
2 Empty State
0 of 2 orders. Nothing predicted yet. You stay in manual control. Predictions start after 2 orders.
3. Notification
The weekly cart arrives as a lock-screen notification. It says nothing is charged yet and opens the Review screen.
4. Review your cart
Every AI pick is marked and gives a reason. Not now removes a pick without a fight. The subtitle says nothing is charged until you approve.
5. Price jump
An orange chip flags a price change before you pay. The changed item moves to the top of the list.
6. AI Preferences
Three controls: turn predictions off, reset what the app learned, block an item. Spend cap is set here, at ₹600. Price rises is a fixed rule you cannot switch off: any rise after you approve asks you again.
7. Cap stops a payment
If a cart goes over the ₹600 cap you set, payment holds. Nothing is charged. You remove items or raise the cap.
Section 6 · The decisions
Four decisions and what each one cost
01
Review before paying
The weekly cart opens as a review screen . you always see it first .
02
AI picks get a chip .
Cost : more noise, more to read
03
“ Based on your Tuesday morning routine “
Every picks gives a reason .
04
Hold payment when something changes.
Price rise
Over cap
Out of stock
Cost : friction. Some will read it as the app being in the way.
Section 8 · When something breaks
What happens when a price rises after you approve
Any price rise after you approve needs a new yes. Step through the flow.
1. You approve
You said yes to ₹329. The cart has 4 items, and you approved the whole subtotal.
2. The price rises
Quaker Oats goes from ₹72 to ₹101. That is up 40%. Any rise after you approve stops the payment.
3. Payment holds
Payment is on hold. Nothing was charged. A changed price is a new decision, so you decide again.
Interface sketch. Item names and prices are illustrative.
Also: out of stock
Fresh Banana is out of stock. The app suggests a close match, Robusta Banana at ₹76, which is ₹4 more. It is never added on its own. Approve & pay stays locked until you choose “Use this instead” or “Remove bananas”.
Section 9 · Proof
What I can and can’t claim
I have no live data. The tap counts at the top come from my own screens: 2 taps with no edits (open the Cart tab, then Approve & pay) and 3 with one removal.
That count is not evidence. It shows what I traded away, not whether people want this. The test plan is at the end of this section.
The path to a charge
v1 scope and constraints
Ship in v1
Review screen, AI chip with a reason, price-change chip, predictions off switch. These four answer the three fears.
Cut from v1
Spend cap, substitute suggestions, the “Since last week” box and the notification. They wait until tests show people read the review screen.
Engineering needs
Order history per user, a stored last price per item, and a payment hold until approval. I have not checked these with engineers.
Metric I would watch
Carts approved after the user read or edited something, next to the charge-complaint rate. A high no-edit rate alone could mean rubber-stamping, not trust.
Section 9 · Proof
What I can and can’t claim
I have no live data. The tap counts at the top come from my own screens: 2 taps with no edits (open the Cart tab, then Approve & pay) and 3 with one removal.
That count is not evidence. It shows what I traded away, not whether people want this. The test plan is at the end of this section.
The path to a charge
Four rules the cart keeps
- Rule 1
No charge before approval
The cart is build, then it stops, Open the cart , tab , review tap approve & pay . only that tap charges
- Rule 2
A price rise after approval needs a new yes
Quaker Oats goes from ₹72 to ₹101 after you approved payment holds. You choose again pay 358 , Or remove the oats and pay 257 .
- Rule 3
Substitutes are never added silently
Fresh banana is out , Robusta Banana is suggested at 76 , Which is 4 more . Approve & pay stay locked until you choose .
- Rule 4
Every AI pick has a reason
Each AI chips says why the item is there . " Based on Your Tuesday order . "
1. Home
Home shows Popular near you, with an Add button. Before 2 orders the app is a normal store. No AI marks yet.400
1. Home
Home shows Popular near you, with an Add button. Before 2 orders the app is a normal store. No AI marks yet.400
2 Empty State
0 of 2 orders. Nothing predicted yet. You stay in manual control. Predictions start after 2 orders.
2 Empty State
0 of 2 orders. Nothing predicted yet. You stay in manual control. Predictions start after 2 orders.
7. Cap stops a payment
If a cart goes over the ₹600 cap you set, payment holds. Nothing is charged. You remove items or raise the cap.
7. Cap stops a payment
If a cart goes over the ₹600 cap you set, payment holds. Nothing is charged. You remove items or raise the cap.
4. Review your cart
Every AI pick is marked and gives a reason. Not now removes a pick without a fight. The subtitle says nothing is charged until you approve.
4. Review your cart
Every AI pick is marked and gives a reason. Not now removes a pick without a fight. The subtitle says nothing is charged until you approve.
3. Notification
The weekly cart arrives as a lock-screen notification. It says nothing is charged yet and opens the Review screen.
3. Notification
The weekly cart arrives as a lock-screen notification. It says nothing is charged yet and opens the Review screen.
5. Price jump
An orange chip flags a price change before you pay. The changed item moves to the top of the list.
5. Price jump
An orange chip flags a price change before you pay. The changed item moves to the top of the list.
Section 3 · The insight
Trust breaks when the app acts and the user finds out later.
A wrong pick you can see costs one tap. A wrong pick you find at the door costs the customer.
Section 4 · The solution
A cart that stops and asks
CartSense builds the weekly cart, then stops at a review screen. Nothing is charged until you approve. Each fear from Section 1 gets one answer.
Fear 1
Why is this in my cart?
+ AI Suggestion
Remove
Based on your Tuesday morning routine.
Quaker Oats
Qty: 1 kg ₹72
− 1 +
Every pick shows its reason
A green chip marks what the system added. One line says why.
Fear 2
Did the price change on me?
₹82 → ₹89 since your last order
Epigamia Greek Yogurt
Qty: 1 kg ₹89
− 1 +
Price changes get flagged
An orange chip shows the old price, only when it changed. A rise after you approve needs a new yes.
Fear 3
Can I make it stop?
Turn predictions off
◯
Reset what the app learned
›
Block an item
›
Three controls, one place
All in AI Preferences. You also set a ₹600 spend cap there.
See the screens →
Section 6 · The decisions
Four decisions and what each one cost
Section 7 · The screens
The screens, in the order a week unfolds
1
The screen opens before any charge. Decision 1.
2
Green chip marks what the system added. Decision 2.
3
One line says why. Remove and a stepper sit on each card. Decision 3.
4
Orange chip shows the old price, only when it changed. See When it breaks.
Interface sketch. Item names and prices are illustrative.
The screens, in the order a week unfolds. Item names and prices are illustrative.
02
AI picks get a chip .
Cost : more noise, more to read
03
“ Based on your Tuesday morning routine “
Every picks gives a reason .
04
Hold payment when something changes.
Price rise
Over cap
Out of stock
Cost : friction. Some will read it as the app being in the way.
01
Review before paying
The weekly cart opens as a review screen . you always see it first .
02
AI picks get a chip .
Cost : more noise, more to read
03
“ Based on your Tuesday morning routine “
Every picks gives a reason .
04
Hold payment when something changes.
Price rise
Over cap
Out of stock
Cost : friction. Some will read it as the app being in the way.
Section 2 · Research
What reviews and docs already say
Desk research from docs and reviews. I did not test these apps hands-on, so “not shown” means not shown in the sources I read.
One Misfits Market reviewer, via Onclarity, calls the app “a maze” for turning off auto-shopping. Reviews at 1 to 3 stars across five US grocery apps list substitutions nobody asked for among the main complaints (Unstar). Blinkit review summaries on Google Play list wrong items and out-of-stock products (Kimola).
■ Gap or safety net too late
■ Review before charge
In the 5 sources I read, none shows why an item is in the cart or what changed since the last order. CartSense targets that gap.
Section 3 · The insight
Trust breaks when the app acts and the user finds out later.
A wrong pick you can see costs one tap. A wrong pick you find at the door costs the customer.
Instacart
Buy it again list
→
You build the cart yourself
→
Pay
Shows what you bought before, not what changed.
My answer: an orange chip on every price change.
Blinkit
Reorder
→
Wrong or missing items
Review summaries list wrong items and out-of-stock products. I did not test the app.
My answer: review comes before the charge.
Misfits Market
Auto-shopping on
→
Charged
→
Hard to turn off
Reviewers say it is hard to turn off, and there is no review step.
My answer: three controls in one place.
Where the safety net sits
CartSense
Predicted cart
→
Your review
→
→
You approve
→
Charged
Predicted cart
You approve
The review comes before any charge.
Section 4 · The solution
A cart that stops and asks
CartSense builds the weekly cart, then stops at a review screen. Nothing is charged until you approve. Each fear from Section 1 gets one answer.
Fear 1
Why is this in my cart?
+ AI Suggestion
Remove
Based on your Tuesday morning routine.
Quaker Oats
Qty: 1 kg ₹72
− 1 +
Every pick shows its reason
A green chip marks what the system added. One line says why.
Fear 2
Did the price change on me?
₹82 → ₹89 since your last order
Epigamia Greek Yogurt
Qty: 1 kg ₹89
− 1 +
Price changes get flagged
An orange chip shows the old price, only when it changed. A rise after you approve needs a new yes.
Fear 3
Can I make it stop?
Turn predictions off
◯
Reset what the app learned
›
Block an item
›
Three controls, one place
All in AI Preferences. You also set a ₹600 spend cap there.
See the screens →
Section 5 · How I got here
How I got here
Two options I designed and dropped, and the one I kept.
Option 1
Dropped
Full automation
A background checkout with no review step. A wrong item reaches checkout unseen, so the feature fails at its one job.
Option 2
Dropped
Auto-charge, 30-minute cancel window
Looks safer, but the money is already committed. You must catch the mistake in time.
Option 3
Kept
Review, then approve
The cart stops at a review screen. Nothing is charged until you approve. Only then you see this confirmation, with each AI pick marked and a link to AI preferences.
Rubber-stamp risk
The review adds a tap. Fast approval can turn into tapping without reading. I kept the tap, because one wrong charge costs more trust than one extra tap.
The risk
Tap without reading
Approve & pay ₹329
☝ ☝ ☝
Same button every week. The thumb learns the spot.
My defence
Untested
2 changes since last week
Approve
What changed sits next to the pay button, at the moment of paying. Task 3 in the test plan checks it.
1. Home
Home shows Popular near you, with an Add button. Before 2 orders the app is a normal store. No AI marks yet.400
3. Notification
The weekly cart arrives as a lock-screen notification. It says nothing is charged yet and opens the Review screen.
2 Empty State
0 of 2 orders. Nothing predicted yet. You stay in manual control. Predictions start after 2 orders.
4. Review your cart
Every AI pick is marked and gives a reason. Not now removes a pick without a fight. The subtitle says nothing is charged until you approve.
5. Price jump
An orange chip flags a price change before you pay. The changed item moves to the top of the list.
6. AI Preferences
Three controls: turn predictions off, reset what the app learned, block an item. Spend cap is set here, at ₹600. Price rises is a fixed rule you cannot switch off: any rise after you approve asks you again.
7. Cap stops a payment
If a cart goes over the ₹600 cap you set, payment holds. Nothing is charged. You remove items or raise the cap.
01
Review before paying
The weekly cart opens as a review screen. You always see it first.
02
AI picks get a chip.
Cost: More noise, more to read.
03
“Based on your Tuesday morning routine.”
Every pick gives a reason.
04
Pause payment when something changes.
Trade-off: Extra friction may make the app feel like it’s getting in the way.
Price rise
Out of stock
Over cap
04
Pause payment when something changes.
Trade-off: Extra friction may make the app feel like it’s getting in the way.
Price rise
Out of stock
Over cap
Section 8 · When something breaks
What happens when a price rises after you approve
Any price rise after you approve needs a new yes. Step through the flow.
- Rule 1
No charge before approval
The cart is build, then it stops, Open the cart , tab , review tap approve & pay . only that tap charges
- Rule 2
A price rise after approval needs a new yes
Quaker Oats goes from ₹72 to ₹101 after you approved payment holds. You choose again pay 358 , Or remove the oats and pay 257 .
- Rule 3
Substitutes are never added silently
Fresh banana is out , Robusta Banana is suggested at 76 , Which is 4 more . Approve & pay stay locked until you choose .
- Rule 4
Every AI pick has a reason
Each AI chips says why the item is there . " Based on Your Tuesday order . "
Interface sketch. Item names and prices are illustrative.
Also: out of stock
Fresh Banana is out of stock. The app suggests a close match, Robusta Banana at ₹76, which is ₹4 more. It is never added on its own. Approve & pay stays locked until you choose “Use this instead” or “Remove bananas”.
Section 9 · Proof
What I can and can’t claim
I have no live data. The tap counts at the top come from my own screens: 2 taps with no edits (open the Cart tab, then Approve & pay) and 3 with one removal.
That count is not evidence. It shows what I traded away, not whether people want this. The test plan is at the end of this section.
The path to a charge
v1 scope and constraints
Ship in v1
Review screen, AI chip with a reason, price-change chip, predictions off switch. These four answer the three fears.
Cut from v1
Spend cap, substitute suggestions, the “Since last week” box and the notification. They wait until tests show people read the review screen.
Engineering needs
Order history per user, a stored last price per item, and a payment hold until approval. I have not checked these with engineers.
Metric I would watch
Carts approved after the user read or edited something, next to the charge-complaint rate. A high no-edit rate alone could mean rubber-stamping, not trust.
CartSense -Trust-First AI Reordering
REORDERING GROCERIES SHOULD TAKE SECONDS, NOT SCROLLING. CARTSENSE SUGGESTS THE CART. YOU APPROVE IT.
Section 5 · How I got here
How I got here
Option 1
Dropped
Option 3
Kept
Option 2
Dropped
Option 1
Dropped
Option 3
Kept
Option 2
Dropped
Droppped
Full Automation
A background checkout with no review step. A wrong item reaches checkout unseen, so the feature fails at its one job.
Kept
Review, then approve
The cart stops at a review screen. Nothing is charged until you approve. Only then you see this confirmation, with each AI pick marked and a link to AI preferences.
Droppped
Auto-charge, 30-min cancel
Looks safer, but the money is already committed. You must catch the mistake in time.
Rubber-stamp risk
Two options I designed and dropped, and the one I kept.
The review adds a tap. Fast approval can turn into tapping without reading. I kept the tap, because one wrong charge costs more trust than one extra tap.
The risk
Tap without reading
Approve & pay ₹329
☝ ☝ ☝
Same button every week. The thumb learns the spot.
My defence
Untested
2 changes since last week
Approve
What changed sits next to the pay button, at the moment of paying. Task 3 in the test plan checks it.
Section 8 · When something breaks
What happens when a price rises after you approve
Any price rise after you approve needs a new yes. Step through the flow.
- Rule 1
No charge before approval
The cart is build, then it stops, Open the cart , tab , review tap approve & pay . only that tap charges
- Rule 2
A price rise after approval needs a new yes
Quaker Oats goes from ₹72 to ₹101 after you approved payment holds. You choose again pay 358 , Or remove the oats and pay 257 .
- Rule 3
Substitutes are never added silently
Fresh banana is out , Robusta Banana is suggested at 76 , Which is 4 more . Approve & pay stay locked until you choose .
- Rule 4
Every AI pick has a reason
Each AI chips says why the item is there . " Based on Your Tuesday order . "
Interface sketch. Item names and prices are illustrative.
Also: out of stock
Fresh Banana is out of stock. The app suggests a close match, Robusta Banana at ₹76, which is ₹4 more. It is never added on its own. Approve & pay stays locked until you choose “Use this instead” or “Remove bananas”.
6. AI Preferences
Three controls: turn predictions off, reset what the app learned, block an item. Spend cap is set here, at ₹600. Price rises is a fixed rule you cannot switch off: any rise after you approve asks you again.
6. AI Preferences
Three controls: turn predictions off, reset what the app learned, block an item. Spend cap is set here, at ₹600. Price rises is a fixed rule you cannot switch off: any rise after you approve asks you again.
SECTION · MY ANSWER TO THE GAPS
GAP · HARD TO TURN OFF
Controls in one place.
Turn predictions off, reset what the app learned, block an item. Set a ₹600 spend cap. A price rise after approval always asks again.
GAP · WRONG ITEMS, SILENT PRICE CHANGES
Review before the charge.
Every pick carries a chip and a reason. A price change gets an orange chip. Nothing is charged until you approve.
SECTION · MY ANSWER TO THE GAPS
GAP · HARD TO TURN OFF
Controls in one place.
Turn predictions off, reset what the app learned, block an item. Set a ₹600 spend cap. A price rise after approval always asks again.
GAP · WRONG ITEMS, SILENT PRICE CHANGES
Review before the charge.
Every pick carries a chip and a reason. A price change gets an orange chip. Nothing is charged until you approve.
Instacart
Buy it again list
You build the cart yourself
Pay
Shows what you bought before, not
what changed.
My answer: an orange chip on every
price change.
Blinkit
Reorder
Wrong on missing item
Review summaries list wrong items
and out-of-stock products. I did not
test the app.
My answer: review comes before the
charge.
Misfits Market
Auto - Shopping on
Charged
Hard to turn off
Reviewers say it is hard to turn off, and there is no review step.
My answer: review comes before the
charge.
Section 2 · Research
What reviews and docs already say
Desk research from docs and reviews. I did not test these apps hands-on, so “not shown” means not shown in the sources I read.
One Misfits Market reviewer, via Onclarity, calls the app “a maze” for turning off auto-shopping. Reviews at 1 to 3 stars across five US grocery apps list substitutions nobody asked for among the main complaints (Unstar). Blinkit review summaries on Google Play list wrong items and out-of-stock products (Kimola).
Where the safety net sits
■ Gap or safety net too late
■ Review before charge
In the 5 sources I read, none shows why an item is in the cart or what changed since the last order. CartSense targets that gap.
Section 7 · The screens
The screens, in the order a week unfolds
1
The screen opens before any charge. Decision 1.
2
Green chip marks what the system added. Decision 2.
3
One line says why. Remove and a stepper sit on each card. Decision 3.
4
Orange chip shows the old price, only when it changed. See When it breaks.
Interface sketch. Item names and prices are illustrative.
The screens, in the order a week unfolds. Item names and prices are illustrative.
Section 6 · The decisions
Four decisions and what each one cost
Section 1 · The problem
An AI cart that charges on its own can break trust
Every week, people rebuild a cart they already know. Automation takes that work away. One surprise charge gives it back, and then some, because now the person has to watch the app.
Who it affects: anyone who orders groceries from an app on a routine and pays from the same account each time.
My question: how might we take the weekly rebuild off people without making them feel the app took control of their money? This is my starting guess. Real people have not checked it.
Why I see it this way: I worked as a Team Leader in a BPO. Customers there rarely complained about a feature. They complained about a surprise, and about the edge case nobody planned for. That is why every decision below starts from an edge case.
Where trust breaks
Three questions a user asks the moment an app spends for them.
Fear 1
“Why is this in my cart?”
Every AI pick has a chip and a reason.
Fear 2
“Did the price change on me?”
An orange chip flags every price change.
Fear 3
“Can I make it stop?”
Turn predictions off, reset what the app learned, or block an item.