==👉 Find the 14 problems set below==

Binary Search on Answer — Interview Study Order

🟢 Group 1 — Core / Must Master

These establish the basic candidate → feasible() → first True pattern.

#	Problem	LC	Pattern	Company tags*	Links
1	Koko Eating Bananas	875	Minimum rate → count time	Amazon, Google	koko-eating-bananas
2	Capacity to Ship Packages Within D Days	1011	Minimum capacity → greedy grouping	Amazon, Meta	capacity-to-ship-packages
3	Find the Smallest Divisor Given a Threshold	1283	Minimum divisor → counting	Amazon, Apple	find-the-smallest-divisor-given-a-threshold
4	Minimum Speed to Arrive on Time	1870	Minimum speed → time calculation	Google, Amazon	koko-eating-bananas

What you should learn from this group:

X = candidate answer
        ↓
Can X satisfy the constraint?
        ↓
FFFFTTTT
        ↓
find first True

After these, you should be able to write the basic template without thinking.

---

🟡 Group 2 — Minimize the Maximum

This is probably the most important family after Group 1.

The common transformation is:

“Minimize the maximum ___”

becomes:

“Assume the maximum is X. Can I make the entire problem work?”

#	Problem	LC	Validator	Company tags*	Links
5	Split Array Largest Sum	410	Greedy partition	Google, Meta, Amazon	split-array-largest-sum
6	Minimized Maximum of Products Distributed to Any Store	2064	Greedy allocation	Amazon, Microsoft	minimized-maximum-of-products-distributed-to-any-store
7	Minimum Limit of Balls in a Bag	1760	Count required splits	Google, Amazon	minimum-limit-of-balls-in-a-bag
8	Book Allocation	—	Greedy partition	Amazon, Microsoft	split-array-largest-sum
9	Painter’s Partition	—	Greedy partition	Amazon, Microsoft	split-array-largest-sum

Important:

You don’t need to separately “learn” Book Allocation and Painter’s Partition after LC 410.

They are essentially the same family:

candidate maximum load
        ↓
greedily create groups
        ↓
groups <= K ?
        ↓
first feasible

So I’d study Split Array Largest Sum deeply, then use the others as reinforcement.

---

🔵 Group 3 — Maximize the Minimum

This is the other major pattern you absolutely need.

The wording usually looks like:

Maximize the minimum distance/value.

You instead ask:

Can I achieve a minimum of at least X?

Now the predicate is:

TTTTFFFF

and you find the last True.

#	Problem	LC	Validator	Company tags*	Links
10	Magnetic Force Between Two Balls	1552	Greedy placement	Amazon, Meta, Google	magnetic-force-between-two-balls-or-aggressive-cows
11	Aggressive Cows	—	Same greedy placement	Amazon, Google	magnetic-force-between-two-balls-or-aggressive-cows
12	Divide Chocolate	1231	Greedy partition	Google	divide-chocolate_lc-1231

Again, 1552 is the one I’d learn properly.

The reusable pattern:

candidate minimum distance = X
             ↓
greedily place objects
             ↓
can place >= K?
             ↓
TTTTFFFF
             ↓
last True

---

🟣 Group 4 — Binary Search + Counting

This is where the pattern becomes more interesting.

Instead of a straightforward greedy validator, you count how many things satisfy a property for candidate X.

#	Problem	LC	Validator	Company tags*	Links
13	K-th Smallest Pair Distance	719	Two pointers + counting	Google, Amazon	kth-smallest-pair-distance
14	K-th Smallest Element in a Sorted Matrix	378	Count <= X	Amazon, Google	kth-smallest-element-in-a-sorted-matrix
15	K-th Smallest Number in Multiplication Table	668	Mathematical counting	Google	kth-smallest-number-in-multiplication-table
16	Maximum Candies Allocated to K Children	2226	Count pieces	Google, Amazon	maximum-candies-allocated-to-k-children-(lc-2226)

This teaches a very useful abstraction:

candidate X
    ↓
count(X)
    ↓
is count(X) >= K ?
    ↓
predicate
    ↓
binary search

Priority

For your interview, I’d do:

719 → 2226 → 378 → 668

You don’t necessarily need all four if time becomes tight.

---

🔴 Group 5 — Advanced / Different Validator

Do these only after the previous groups feel natural.

Problem	LC	Why it’s different	Company tags*	Links
Minimize Max Distance to Gas Station	774	Continuous / floating-point BS	Google	minimize-max-distance-to-gas-station
Swim in Rising Water	778	Binary search + graph feasibility	Google, Meta
Ugly Number III	1201	Binary search + inclusion-exclusion	Google
K-th Smallest Prime Fraction	786	More specialized predicate	Google

Swim in Rising Water

I would not prioritize this as a Binary Search-on-Answer problem.

It is valuable, but the important lesson is really:

Binary Search
+
BFS/DFS feasibility

and the problem has other standard solutions, particularly Dijkstra.

So don’t let it take time away from the core families.

---

🎯 Your Actual Study Roadmap

If your interview is coming fast, I’d reduce everything to this:

Phase 1 — Basic predicate

1. Koko Eating Bananas
2. Capacity to Ship Packages
3. Smallest Divisor Given Threshold

↓

Phase 2 — Minimize Maximum

4. Split Array Largest Sum / Book allocation/ painter's partition
5. Minimized Maximum of Products
6. Minimum Limit of Balls in a Bag
 

↓

Phase 3 — Maximize Minimum

7. Magnetic Force Between Two Balls / aggressive cows
8. Divide Chocolate

↓

Phase 4 — Counting Predicate

9. K-th Smallest Pair Distance
10. Maximum Candies Allocated to K Children
11. K-th Smallest in Sorted Matrix

↓

Phase 5 — Advanced

12. Minimize Max Distance to Gas Station
13. Ugly Number III

That’s the 13-problem core set I’d use.

---

The patterns you should be able to recognize after these

Pattern	Representative problem
Minimum rate	Koko
Minimum capacity	Ship Packages
Minimum threshold/divisor	Smallest Divisor
Minimize maximum partition	Split Array
Minimize maximum allocation	Minimized Maximum
Minimize maximum after splitting	Balls in a Bag
Maximize minimum distance	Magnetic Force
Maximize minimum value	Divide Chocolate
Binary search + pair counting	K-th Pair Distance
Binary search + value counting	K-th Smallest Matrix
Binary search + mathematical counting	Multiplication Table
Binary search + graph feasibility	Swim in Rising Water
Continuous answer search	Gas Station

One thing I’d change from the Gemini list

For your limited time, learn the pattern, then solve 1–2 variations rather than collecting dozens of nearly identical problems.

*Company tags are approximate historical interview/problem-bank tags, not guarantees of what a company will ask.