interview-reps 0.5.0 → 0.6.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +5 -0
- package/content/CLAUDE.md +51 -7
- package/content/behavioral/REVIEW.md +51 -0
- package/content/design/REVIEW.md +48 -0
- package/content/product/P21-support-reply-drafts/README.md +29 -0
- package/content/product/P22-receipt-categories/README.md +29 -0
- package/content/product/P23-pharmacy-symptom-helper/README.md +31 -0
- package/content/product/P24-listing-descriptions/README.md +29 -0
- package/content/product/P25-group-class-bookings/README.md +28 -0
- package/content/product/P26-home-repair-press-release/README.md +28 -0
- package/content/product/P27-merchant-cash-advances/README.md +30 -0
- package/content/product/P28-retire-the-desktop-app/README.md +28 -0
- package/content/product/P29-live-courier-location/README.md +28 -0
- package/content/product/P30-carts-across-devices/README.md +28 -0
- package/content/product/P31-ticket-on-sale/README.md +29 -0
- package/content/product/P32-transfer-status/README.md +28 -0
- package/content/product/P33-new-revenue-by-plan/README.md +47 -0
- package/content/product/P34-weekly-app-opens/README.md +46 -0
- package/content/product/P35-second-orders/README.md +49 -0
- package/content/product/P36-on-time-by-city/README.md +46 -0
- package/content/product/P37-trials-that-never-pay/README.md +52 -0
- package/content/product/P38-smaller-grocery-baskets/README.md +51 -0
- package/content/product/P39-clinic-bookings-abroad/README.md +52 -0
- package/content/product/REVIEW.md +136 -1
- package/dist/cli.mjs +1911 -573
- package/package.json +1 -1
|
@@ -0,0 +1,47 @@
|
|
|
1
|
+
# P33 · New monthly revenue by plan
|
|
2
|
+
|
|
3
|
+
Category: SQL · Level: core · Time: 25 min
|
|
4
|
+
|
|
5
|
+
You are the product manager of a subscription meal-kit service. Your lead asks which plan brought in the most new monthly revenue from people who signed up in October 2026. Write a query against the two tables below and explain what it shows.
|
|
6
|
+
|
|
7
|
+
`customers`
|
|
8
|
+
|
|
9
|
+
| id | signed_up_on | country | is_test |
|
|
10
|
+
|---|---|---|---|
|
|
11
|
+
| 1 | 2026-10-03 | US | 0 |
|
|
12
|
+
| 2 | 2026-10-11 | CA | 0 |
|
|
13
|
+
| 3 | 2026-09-28 | US | 0 |
|
|
14
|
+
| 4 | 2026-10-20 | US | 1 |
|
|
15
|
+
|
|
16
|
+
`subscriptions`
|
|
17
|
+
|
|
18
|
+
| id | customer_id | plan | started_on | monthly_price | ended_on |
|
|
19
|
+
|---|---|---|---|---|---|
|
|
20
|
+
| 10 | 1 | family | 2026-10-03 | 80 | |
|
|
21
|
+
| 11 | 2 | couple | 2026-10-11 | 50 | 2026-10-25 |
|
|
22
|
+
| 12 | 2 | family | 2026-10-25 | 80 | |
|
|
23
|
+
| 13 | 3 | couple | 2026-10-01 | 50 | |
|
|
24
|
+
| 14 | 4 | family | 2026-10-20 | 80 | |
|
|
25
|
+
|
|
26
|
+
Before you answer, find out what "new monthly revenue" means here, which customers should not count, and what to do with someone who changed plans.
|
|
27
|
+
|
|
28
|
+
## If you ask
|
|
29
|
+
|
|
30
|
+
- "New monthly revenue" is the monthly price of the plan a customer started on; prices are in dollars.
|
|
31
|
+
- Test accounts are marked with `is_test = 1` and never count.
|
|
32
|
+
- A customer who switches plans gets a new subscription row. For this question, count only the plan they first started on.
|
|
33
|
+
- Count a subscription even if it has ended since. `ended_on` is empty while it is active.
|
|
34
|
+
- Any common SQL dialect is fine; say which date functions you assume. Out of scope: refunds and tax.
|
|
35
|
+
|
|
36
|
+
## What a strong answer covers
|
|
37
|
+
|
|
38
|
+
- Restates the question exactly, including which signups count, before writing any SQL.
|
|
39
|
+
- Leaves out every row the question excludes, and says how the query does it.
|
|
40
|
+
- Counts each customer the way the question asks, even when they have more than one subscription.
|
|
41
|
+
- Returns one row per plan with its total, in an order that answers "which".
|
|
42
|
+
- Explains what the result shows and names one check they would run on it.
|
|
43
|
+
|
|
44
|
+
## Follow-ups
|
|
45
|
+
|
|
46
|
+
- Your lead now wants the same answer for each week of October. What changes in the query?
|
|
47
|
+
- The finance team counts revenue from the day a plan is paid, not the day it starts. Does your answer change?
|
|
@@ -0,0 +1,46 @@
|
|
|
1
|
+
# P34 · Weekly app opens by platform
|
|
2
|
+
|
|
3
|
+
Category: SQL · Level: core · Time: 25 min
|
|
4
|
+
|
|
5
|
+
You are the product manager of a language-learning app. Before a planning meeting you need the number of different people who opened the app each week of November 2026, split by platform. Write a query against the two tables below and explain what it shows.
|
|
6
|
+
|
|
7
|
+
`users`
|
|
8
|
+
|
|
9
|
+
| id | created_at | country |
|
|
10
|
+
|---|---|---|
|
|
11
|
+
| 101 | 2026-08-14 09:12:00 | DE |
|
|
12
|
+
| 102 | 2026-11-02 18:40:00 | BR |
|
|
13
|
+
| 103 | 2026-10-30 07:05:00 | US |
|
|
14
|
+
|
|
15
|
+
`events`
|
|
16
|
+
|
|
17
|
+
| user_id | event_name | occurred_at | platform |
|
|
18
|
+
|---|---|---|---|
|
|
19
|
+
| 101 | app_open | 2026-11-02 08:00:01 | ios |
|
|
20
|
+
| 101 | app_open | 2026-11-02 08:00:01 | ios |
|
|
21
|
+
| 101 | lesson_done | 2026-11-02 08:09:30 | ios |
|
|
22
|
+
| 102 | app_open | 2026-11-03 19:15:00 | android |
|
|
23
|
+
| 103 | app_open | 2026-11-04 12:00:00 | web |
|
|
24
|
+
|
|
25
|
+
Before you answer, find out which event counts as opening the app, when a week starts, how time zones work, and whether the events can be trusted as they are.
|
|
26
|
+
|
|
27
|
+
## If you ask
|
|
28
|
+
|
|
29
|
+
- An open is the `app_open` event. Times are in UTC.
|
|
30
|
+
- Weeks start on Monday. A week that runs into December still counts as a November week if it starts in November.
|
|
31
|
+
- A person who opens the app on two platforms counts once on each platform.
|
|
32
|
+
- A known bug sometimes logs the same event twice, with the same user and the same second.
|
|
33
|
+
- Any common SQL dialect is fine; say which date functions you assume. Out of scope: time spent in the app.
|
|
34
|
+
|
|
35
|
+
## What a strong answer covers
|
|
36
|
+
|
|
37
|
+
- States what is being counted and over which weeks before writing any SQL.
|
|
38
|
+
- Counts people, not rows, and says how the query avoids counting anyone twice in a week.
|
|
39
|
+
- Groups by platform and week the way the question asks, with the week boundaries handled.
|
|
40
|
+
- Keeps only the events that answer the question.
|
|
41
|
+
- Explains how to read the result and names a check that would catch a wrong count.
|
|
42
|
+
|
|
43
|
+
## Follow-ups
|
|
44
|
+
|
|
45
|
+
- The planning meeting also wants the share of each week's people who were new that week. What would you add?
|
|
46
|
+
- One platform's count doubles in the last week. What do you check before you present it?
|
|
@@ -0,0 +1,49 @@
|
|
|
1
|
+
# P35 · Second orders within 30 days
|
|
2
|
+
|
|
3
|
+
Category: SQL · Level: advanced · Time: 25 min
|
|
4
|
+
|
|
5
|
+
You are the product manager for growth at a pet-supplies shop that sells online. You want to know how many new customers come back soon: of the customers whose first completed order was in September 2026, what share placed another completed order within 30 days of it, split by the channel they came from? Write a query against the two tables below and explain what it shows.
|
|
6
|
+
|
|
7
|
+
`customers`
|
|
8
|
+
|
|
9
|
+
| id | first_seen_at | channel |
|
|
10
|
+
|---|---|---|
|
|
11
|
+
| 1 | 2026-08-20 | search |
|
|
12
|
+
| 2 | 2026-09-01 | social |
|
|
13
|
+
| 3 | 2026-09-05 | search |
|
|
14
|
+
| 4 | 2026-09-12 | referral |
|
|
15
|
+
|
|
16
|
+
`orders`
|
|
17
|
+
|
|
18
|
+
| id | customer_id | ordered_at | total | status |
|
|
19
|
+
|---|---|---|---|---|
|
|
20
|
+
| 501 | 1 | 2026-08-21 10:00:00 | 42.00 | completed |
|
|
21
|
+
| 502 | 1 | 2026-09-03 09:30:00 | 18.50 | completed |
|
|
22
|
+
| 503 | 2 | 2026-09-02 14:00:00 | 25.00 | cancelled |
|
|
23
|
+
| 504 | 2 | 2026-09-04 08:15:00 | 31.00 | completed |
|
|
24
|
+
| 505 | 2 | 2026-09-29 20:45:00 | 12.00 | refunded |
|
|
25
|
+
| 506 | 3 | 2026-09-06 11:00:00 | 60.00 | completed |
|
|
26
|
+
|
|
27
|
+
Before you answer, find out what counts as a customer's first order, which order statuses count, and how the 30 days are measured.
|
|
28
|
+
|
|
29
|
+
## If you ask
|
|
30
|
+
|
|
31
|
+
- The first order is the first completed order a customer ever placed, not only the first one in September.
|
|
32
|
+
- A cancelled order never counts. A refunded order still counts as an order placed.
|
|
33
|
+
- The 30 days run from the first order's time; an order exactly 30 days later counts.
|
|
34
|
+
- The channel is the one on the customer's row; it never changes.
|
|
35
|
+
- Any common SQL dialect is fine; say which date and window functions you assume. Out of scope: order value.
|
|
36
|
+
|
|
37
|
+
## What a strong answer covers
|
|
38
|
+
|
|
39
|
+
- Pins down who is in the group being measured before measuring anything about them.
|
|
40
|
+
- Applies the status rules the same way to the first order and to the second.
|
|
41
|
+
- Measures the 30 days from the right moment and handles the boundary as asked.
|
|
42
|
+
- Returns a share per channel with the counts behind it, so a small channel is not over-read.
|
|
43
|
+
- Explains the result, names a caveat in the data and how to check it.
|
|
44
|
+
|
|
45
|
+
## Follow-ups
|
|
46
|
+
|
|
47
|
+
- The marketing lead wants the same share for every month this year in one result. What changes?
|
|
48
|
+
- One channel has 40 customers and the others have thousands. How do you present the result?
|
|
49
|
+
- Refunds turn out to be mostly fraud. Does the definition change, and what happens to the numbers?
|
|
@@ -0,0 +1,46 @@
|
|
|
1
|
+
# P36 · On-time deliveries by city
|
|
2
|
+
|
|
3
|
+
Category: SQL · Level: advanced · Time: 25 min
|
|
4
|
+
|
|
5
|
+
You are the product manager for delivery quality at a grocery delivery service. Operations says on-time deliveries got worse in October 2026. Find the on-time share for each city in October, and the three cities where it fell most compared with September. Write a query against the two tables below and explain what it shows.
|
|
6
|
+
|
|
7
|
+
`cities`
|
|
8
|
+
|
|
9
|
+
| id | name | launched_on |
|
|
10
|
+
|---|---|---|
|
|
11
|
+
| 1 | Riverton | 2024-03-01 |
|
|
12
|
+
| 2 | Lakeside | 2025-06-15 |
|
|
13
|
+
| 3 | Hillcrest | 2026-09-20 |
|
|
14
|
+
|
|
15
|
+
`deliveries`
|
|
16
|
+
|
|
17
|
+
| id | city_id | promised_by | delivered_at | status |
|
|
18
|
+
|---|---|---|---|---|
|
|
19
|
+
| 9001 | 1 | 2026-09-10 18:00:00 | 2026-09-10 17:52:00 | delivered |
|
|
20
|
+
| 9002 | 1 | 2026-10-02 12:00:00 | 2026-10-02 12:20:00 | delivered |
|
|
21
|
+
| 9003 | 2 | 2026-10-05 19:30:00 | | cancelled |
|
|
22
|
+
| 9004 | 2 | 2026-10-07 09:00:00 | | delivered |
|
|
23
|
+
| 9005 | 3 | 2026-10-12 20:00:00 | 2026-10-12 19:45:00 | delivered |
|
|
24
|
+
|
|
25
|
+
Before you answer, find out what on time means, which deliveries count, which month a delivery belongs to, and what to do with rows that look wrong.
|
|
26
|
+
|
|
27
|
+
## If you ask
|
|
28
|
+
|
|
29
|
+
- On time means delivered at or before `promised_by`. A delivery belongs to the month of `promised_by`.
|
|
30
|
+
- Cancelled deliveries do not count in either the on-time share or the total.
|
|
31
|
+
- About 1% of delivered rows have no `delivered_at` because of a logging bug fixed in November.
|
|
32
|
+
- Hillcrest launched in late September, so it had very few deliveries that month.
|
|
33
|
+
- Any common SQL dialect is fine. Out of scope: why a delivery was late.
|
|
34
|
+
|
|
35
|
+
## What a strong answer covers
|
|
36
|
+
|
|
37
|
+
- Defines the share and its denominator before writing SQL, and says which rows are left out.
|
|
38
|
+
- Decides what to do with rows missing a delivery time, says why, and says how it moves the result.
|
|
39
|
+
- Compares the two months city by city in one result, in points of share, with the counts behind it.
|
|
40
|
+
- Deals with a city that has too little data in one month instead of letting it top the list.
|
|
41
|
+
- Explains what the result shows and what they would look at next.
|
|
42
|
+
|
|
43
|
+
## Follow-ups
|
|
44
|
+
|
|
45
|
+
- Operations asks whether the drop is the cities or the mix of deliveries (more evening slots, say). How would you tell?
|
|
46
|
+
- The logging bug turns out to hit late deliveries more often. What does that do to your answer?
|
|
@@ -0,0 +1,52 @@
|
|
|
1
|
+
# P37 · Trial teams that never pay
|
|
2
|
+
|
|
3
|
+
Category: Case presentation · Level: core · Time: 75 min
|
|
4
|
+
|
|
5
|
+
Panel case. Tailspin makes an online whiteboard for teams. New teams get a free 14-day trial. Over the last quarter, the share of trials that become paying teams fell from 15% to 12%, and the head of growth has asked you to present what is going on and what to do. The finance lead will be on the panel and is worried about the revenue target.
|
|
6
|
+
|
|
7
|
+
Ten weeks ago the team launched a template gallery: anyone can open a template from a search result and start a trial in one click. Trials have grown 40% since.
|
|
8
|
+
|
|
9
|
+
Table 1. Trials and trial to paid, by where the team signed up
|
|
10
|
+
|
|
11
|
+
| Signed up from | Share of trials before | Share now | Trial to paid before | Trial to paid now |
|
|
12
|
+
|---|---|---|---|---|
|
|
13
|
+
| Search ads | 45% | 30% | 15% | 15% |
|
|
14
|
+
| Invite from a paying team | 30% | 22% | 20% | 19% |
|
|
15
|
+
| Template gallery | 5% | 35% | 6% | 5% |
|
|
16
|
+
| Partner sites | 20% | 13% | 10% | 10% |
|
|
17
|
+
| All | 100% | 100% | 15% | 12% |
|
|
18
|
+
|
|
19
|
+
Table 2. Trials and trial to paid, by people active in the first week
|
|
20
|
+
|
|
21
|
+
| People active in week 1 | Share of trials before | Share now | Trial to paid before | Trial to paid now |
|
|
22
|
+
|---|---|---|---|---|
|
|
23
|
+
| 1 | 40% | 58% | 6% | 5% |
|
|
24
|
+
| 2 to 4 | 35% | 27% | 18% | 18% |
|
|
25
|
+
| 5 or more | 25% | 15% | 25% | 26% |
|
|
26
|
+
|
|
27
|
+
You have five slides for the panel: what is happening, why, what Tailspin could do, what you would do, and how you would roll it out. The panel will ask follow-up questions after.
|
|
28
|
+
|
|
29
|
+
Before you answer, find out what the business cares about most (the rate, the number of paying teams, or revenue), what the template gallery was meant to do, and what the team can change in a quarter.
|
|
30
|
+
|
|
31
|
+
## If you ask
|
|
32
|
+
|
|
33
|
+
- New paying teams per month went from 900 to 950, because there are more trials.
|
|
34
|
+
- The paid plan is priced per person; teams of five or more bring most of the revenue.
|
|
35
|
+
- The template gallery was meant to bring in people who would later invite their team.
|
|
36
|
+
- One team of six people can work on this next quarter.
|
|
37
|
+
- Out of scope: changing prices or the length of the trial.
|
|
38
|
+
|
|
39
|
+
## What a strong answer covers
|
|
40
|
+
|
|
41
|
+
- Splits the change using the tables before explaining it, and checks whether the falling rate is the real problem.
|
|
42
|
+
- Puts a number from the brief in the insight headline, so the point is clear without the speaker.
|
|
43
|
+
- Lays out real options and recommends one, saying why it beats the others.
|
|
44
|
+
- Gives a plan with first steps, an owner, and a result that would make them stop or change course.
|
|
45
|
+
- Answers panel questions directly, with a number or a condition, rather than repeating the slides.
|
|
46
|
+
|
|
47
|
+
## Follow-ups
|
|
48
|
+
|
|
49
|
+
- Your recommendation changes the template gallery. What would the team that built it say, and what would you tell them?
|
|
50
|
+
- How would you know in four weeks whether your plan is working?
|
|
51
|
+
- If you could run only one experiment, which one, and why that one?
|
|
52
|
+
- Leadership wants trial to paid back at 15% this quarter. Is that the right target?
|
|
@@ -0,0 +1,51 @@
|
|
|
1
|
+
# P38 · Smaller grocery baskets
|
|
2
|
+
|
|
3
|
+
Category: Case presentation · Level: core · Time: 75 min
|
|
4
|
+
|
|
5
|
+
Panel case. Adatum delivers groceries in four cities. Over the last eight weeks, the average order value fell from $68 to $62, while orders per week rose 9%. The chief operating officer has asked you to present why baskets are smaller and what Adatum should do. The head of the warehouse team will be on the panel.
|
|
6
|
+
|
|
7
|
+
Eight weeks ago a free-delivery offer on a customer's first order started. Six weeks ago a new warehouse system went live in the largest city, which has 55% of all orders.
|
|
8
|
+
|
|
9
|
+
Table 1. Orders and average order value, by how long the customer has ordered
|
|
10
|
+
|
|
11
|
+
| Customer has ordered for | Share of orders before | Share now | Average order before | Average order now |
|
|
12
|
+
|---|---|---|---|---|
|
|
13
|
+
| Under 3 months | 20% | 34% | $52 | $50 |
|
|
14
|
+
| 3 to 12 months | 35% | 30% | $68 | $67 |
|
|
15
|
+
| Over a year | 45% | 36% | $76 | $68 |
|
|
16
|
+
|
|
17
|
+
Table 2. What happened to items in the largest city, per 100 orders
|
|
18
|
+
|
|
19
|
+
| Measure | Before | Now |
|
|
20
|
+
|---|---|---|
|
|
21
|
+
| Orders with an item out of stock at packing | 4 | 9 |
|
|
22
|
+
| Orders with a substitution | 18 | 31 |
|
|
23
|
+
| Orders with an item refunded | 6 | 14 |
|
|
24
|
+
| Complaints about missing items | 1.5 | 3.2 |
|
|
25
|
+
|
|
26
|
+
You have five slides for the panel: what is happening, why, what Adatum could do, what you would do, and how you would roll it out. The panel will ask follow-up questions after.
|
|
27
|
+
|
|
28
|
+
Before you answer, find out whether smaller baskets are a problem in themselves, what each change was meant to do, and what the other three cities look like.
|
|
29
|
+
|
|
30
|
+
## If you ask
|
|
31
|
+
|
|
32
|
+
- Delivery costs $5, free on orders over $50. The first-order offer brought 40% more first orders.
|
|
33
|
+
- In the other three cities, out-of-stock and refund rates have not changed.
|
|
34
|
+
- The warehouse team says the new system is still being tuned and asks for six more weeks.
|
|
35
|
+
- Profit per order depends mostly on order value; small orders lose money after delivery.
|
|
36
|
+
- Out of scope: the prices of goods.
|
|
37
|
+
|
|
38
|
+
## What a strong answer covers
|
|
39
|
+
|
|
40
|
+
- Separates the effects in the tables (who is ordering, and what happens to their orders) before naming causes.
|
|
41
|
+
- Puts a number from the brief in the insight headline.
|
|
42
|
+
- Weighs options that touch different teams, and recommends one with a reason.
|
|
43
|
+
- Gives a plan with an owner, a first week and a line that would make them stop.
|
|
44
|
+
- Handles a challenge from a panel member who owns part of the problem without blaming them.
|
|
45
|
+
|
|
46
|
+
## Follow-ups
|
|
47
|
+
|
|
48
|
+
- Finance asks you to end the free-delivery offer to fix basket size. What do you say?
|
|
49
|
+
- Your plan touches the warehouse team's work. How do you get them to agree?
|
|
50
|
+
- What would you check first on Monday morning, and why that?
|
|
51
|
+
- Baskets recover but orders fall back to where they were. Was your plan a success?
|
|
@@ -0,0 +1,52 @@
|
|
|
1
|
+
# P39 · A second country for clinic bookings
|
|
2
|
+
|
|
3
|
+
Category: Case presentation · Level: advanced · Time: 75 min
|
|
4
|
+
|
|
5
|
+
Panel case. Litware runs an app for booking appointments at private clinics in its home country, where it handles half of all online bookings. Three months ago it started a pilot with 40 clinics in the capital of a neighbouring country. The chief executive must tell the board whether to launch across that country next year, keep piloting, or stop, and has asked you to present a recommendation. The same money could instead add dental clinics at home.
|
|
6
|
+
|
|
7
|
+
Table 1. The two markets
|
|
8
|
+
|
|
9
|
+
| Measure | Home country | Neighbouring country |
|
|
10
|
+
|---|---|---|
|
|
11
|
+
| Private clinics | 12,000 | 9,000 |
|
|
12
|
+
| Clinics that take bookings online | 55% | 20% |
|
|
13
|
+
| Litware's share of online bookings | 50% | under 1% |
|
|
14
|
+
| Largest rival's share of online bookings | 30% | 60% |
|
|
15
|
+
| Fee Litware charges per booking | $2.10 | $2.10 in the pilot |
|
|
16
|
+
|
|
17
|
+
Table 2. The pilot against the home average
|
|
18
|
+
|
|
19
|
+
| Measure | Home average | Pilot |
|
|
20
|
+
|---|---|---|
|
|
21
|
+
| Bookings per clinic per week | 110 | 45 |
|
|
22
|
+
| Clinics still using it after 3 months | 92% | 70% |
|
|
23
|
+
| Bookings made by phone instead of the app | 30% | 65% |
|
|
24
|
+
| Cost to sign up one clinic | $400 | $1,300 |
|
|
25
|
+
|
|
26
|
+
You have five slides for the panel: what is happening, why, what Litware could do, what you would do, and how you would roll it out. The panel will ask follow-up questions after.
|
|
27
|
+
|
|
28
|
+
Before you answer, find out what the board most wants (growth, profit or learning), how long the company can wait for a return, what the pilot was meant to prove, and what the rival is likely to do.
|
|
29
|
+
|
|
30
|
+
## If you ask
|
|
31
|
+
|
|
32
|
+
- The board wants a path to 30% more bookings in three years, and will fund one large bet.
|
|
33
|
+
- A full launch costs about the same as adding dental clinics at home, which a survey says 40% of home clinics would use.
|
|
34
|
+
- The pilot clinics that stayed are mostly in two districts near universities.
|
|
35
|
+
- The rival charges $1.50 per booking and has a contract with the largest clinic group.
|
|
36
|
+
- Out of scope: raising money or buying a company.
|
|
37
|
+
|
|
38
|
+
## What a strong answer covers
|
|
39
|
+
|
|
40
|
+
- Reads the pilot against what it was meant to prove, and says what is still unknown.
|
|
41
|
+
- Puts a number from the brief in the insight headline.
|
|
42
|
+
- Compares the launch with the alternative use of the money, on criteria they state.
|
|
43
|
+
- Makes one recommendation, including "not yet" if that is where the data points, and says what would change it.
|
|
44
|
+
- Gives a plan with milestones and a line that would make the company stop.
|
|
45
|
+
- Keeps answers to panel challenges short and specific, with a number or a condition.
|
|
46
|
+
|
|
47
|
+
## Follow-ups
|
|
48
|
+
|
|
49
|
+
- The pilot clinics that stayed are near universities. Does that change your recommendation?
|
|
50
|
+
- The rival cuts its fee to zero the month you launch. What do you do?
|
|
51
|
+
- What is the cheapest way to learn the thing you are least sure of?
|
|
52
|
+
- A board member says the home market is nearly full. How do you answer?
|
|
@@ -10,7 +10,8 @@ weak area, name the KIND of gap and the question they should ask themselves next
|
|
|
10
10
|
|
|
11
11
|
1. Read the prompt: `product/<Pnn-slug>/README.md`, including its `## If you ask` answers and its
|
|
12
12
|
`## What a strong answer covers` criteria. Its Category line says which kind of question it is:
|
|
13
|
-
Product sense, Execution, Strategy, Estimation or
|
|
13
|
+
Product sense, Execution, Strategy, Estimation, Analytical, AI product, Written doc, Technical, SQL or
|
|
14
|
+
Case presentation. The last five have their own note sections; read the notes under those sections.
|
|
14
15
|
2. Read the person's notes and timing (the app passes them in the review request).
|
|
15
16
|
3. Read earlier reviews of the same prompt in `.reps/product/reviews/<Pnn>-*.md`, if any, and note
|
|
16
17
|
their gaps and "Next time" lines.
|
|
@@ -55,6 +56,91 @@ line under each says how to read it for the kinds where it looks different.
|
|
|
55
56
|
recommendation or answer? Did the run fit the target time, and do the later sections show the time
|
|
56
57
|
reached them? Only the total time is recorded, so never guess how long each section took.
|
|
57
58
|
|
|
59
|
+
### Reading the dimensions for the AI product, written doc, technical, SQL and case presentation kinds
|
|
60
|
+
|
|
61
|
+
These five kinds keep the same eight dimensions and the same header, so every score and tag counts with
|
|
62
|
+
the rest. Read each dimension as below instead of the line above. Each one still needs a 3 for a solid
|
|
63
|
+
run, so a dimension with nothing to judge in the notes scores 1 or 2.
|
|
64
|
+
|
|
65
|
+
**AI product.** The question behind every line: does this problem need a model, and would they know
|
|
66
|
+
before launch whether it works?
|
|
67
|
+
- goal: the goal, the constraints and the scope are pinned down before the design.
|
|
68
|
+
- users: who is served and who is affected when the output is wrong, and which group they picked.
|
|
69
|
+
- structure: they decided between AI, rules and plain software from the problem, said why, and the
|
|
70
|
+
design follows that choice. Choosing a model without asking whether one is needed scores 2 at most.
|
|
71
|
+
- analysis: the design is concrete: what the person sees, what the model takes in and gives back, and
|
|
72
|
+
where a person stays in control.
|
|
73
|
+
- prioritization: they chose what goes in the first version and set a launch bar to clear.
|
|
74
|
+
- metrics: tests before launch (offline evals with the cases they would test) and online measures after,
|
|
75
|
+
with a guardrail. A single accuracy number with no test set scores 2 at most.
|
|
76
|
+
- trade-offs: how mistakes show up and whom they hurt, the fallback when the model is wrong or slow, and
|
|
77
|
+
the cost and latency of each request against the value.
|
|
78
|
+
- communication: as above.
|
|
79
|
+
|
|
80
|
+
**Written doc** (a requirements doc, a press release with FAQ, or a strategy memo). Judge it as a document
|
|
81
|
+
someone else must decide from, not as interview notes.
|
|
82
|
+
- goal: the problem is stated with who has it, how they know, and why now.
|
|
83
|
+
- users: the reader can tell whom it is for and which group comes first.
|
|
84
|
+
- structure: it reads without the author. A person who was not there could follow it from start to end:
|
|
85
|
+
terms defined, sections in order, no shorthand only the author understands.
|
|
86
|
+
- analysis: the proposal is specific enough to build or act on, and the non-goals mark its edges.
|
|
87
|
+
- prioritization: decision clarity. The reader can say what is being decided and what the author recommends.
|
|
88
|
+
- metrics: as above, with a time frame.
|
|
89
|
+
- trade-offs: depth, not a list. The alternatives turned down, why, and what the chosen path costs.
|
|
90
|
+
- communication: length discipline. It says what it must in the space it needs; repetition, padding, or a
|
|
91
|
+
document that runs out before Metrics and Open questions scores 2 at most. Note whether it fits the
|
|
92
|
+
document type the prompt asked for.
|
|
93
|
+
|
|
94
|
+
**Technical.** A product manager's view of how it works: enough to make good product calls, not a full
|
|
95
|
+
system design.
|
|
96
|
+
- goal: the user goal, the scale and the scope come first.
|
|
97
|
+
- users: link to the user. Each technical choice is tied back to what the person notices.
|
|
98
|
+
- structure: the right components. The main objects and parts are there, and nothing important is missing.
|
|
99
|
+
- analysis: the APIs and the data flow are concrete: who calls what, with what, in what order.
|
|
100
|
+
- prioritization: not over-building. They kept to what the product needs now and said what waits.
|
|
101
|
+
- metrics: the measures that show it works for the user (speed, errors, freshness), with a target.
|
|
102
|
+
- trade-offs: latency, consistency and cost reasoned about for this product, with what each choice gives up.
|
|
103
|
+
- communication: as above, and their questions for engineering are ones a team could answer.
|
|
104
|
+
|
|
105
|
+
**SQL** (an analytical question answered with a query). Read the query; never run it, never print a
|
|
106
|
+
corrected query, and never name the join, filter or grouping that is missing.
|
|
107
|
+
- goal: they restated exactly what is asked and wrote down how they read the tables and any ambiguity.
|
|
108
|
+
- users: who is counted. The rows that belong in the answer and the ones that do not (test accounts,
|
|
109
|
+
refunds, duplicates, the wrong dates) are thought about.
|
|
110
|
+
- structure: a query plan before the SQL: which tables, in what order, to what result.
|
|
111
|
+
- analysis: the joins, filters and grouping would return what they meant. Name the KIND of problem only,
|
|
112
|
+
such as "rows can be counted twice" or "the date filter misses part of the range".
|
|
113
|
+
- prioritization: the query answers the question that was asked, not a nearby one.
|
|
114
|
+
- metrics: reading the result. They say what the output shows, in which units, and what they would do next.
|
|
115
|
+
- trade-offs: checks and caveats. Edge cases in the data and how they would check the number.
|
|
116
|
+
- communication: the query is readable (layout, names, comments where needed) and the notes end on an answer.
|
|
117
|
+
|
|
118
|
+
**Case presentation** (a brief with data tables, five slides, then four panel questions). The notes hold
|
|
119
|
+
their brief notes, the five slides (each a headline line, then two or three lines) and the panel's
|
|
120
|
+
questions with the gist of each answer. Four parts carry the most weight: insight from the data, a clear
|
|
121
|
+
recommendation, the plan, and handling of follow-up questions.
|
|
122
|
+
- goal: reading the brief. The situation slide states what is asked and what changed, accurately.
|
|
123
|
+
- users: who it affects. The slides say which customers or teams the numbers are about.
|
|
124
|
+
- structure: the slides tell one story. Each headline says its point, so the five headlines alone make the case.
|
|
125
|
+
- analysis: insight from the data. The insight rests on the brief's tables and cites a number from them;
|
|
126
|
+
an insight with no number from the brief scores 2 at most.
|
|
127
|
+
- prioritization: a clear recommendation. One option, chosen against the others, with the reason.
|
|
128
|
+
- metrics: the plan. First steps with an owner and a time, how they will know it works, and a line that
|
|
129
|
+
would make them stop or change course.
|
|
130
|
+
- trade-offs: options and trade-offs. Real options, and what the recommendation gives up.
|
|
131
|
+
- communication: handling follow-up questions. Each gist answers the question asked, directly, with a
|
|
132
|
+
number or a condition, rather than repeating a slide. An empty gist scores that question as missed.
|
|
133
|
+
|
|
134
|
+
Write the case review in the teaching style, for each of the weakest two or three parts: **What we saw**
|
|
135
|
+
(quote their slide or gist), **Why it matters** (to a panel), and **Before and after**: a short example of
|
|
136
|
+
the same kind of fix on a made-up situation, never this brief's numbers or this case's answer, so it shows
|
|
137
|
+
the shape of a better headline or answer without writing theirs. End with one short drill to practise it.
|
|
138
|
+
|
|
139
|
+
Map gaps to the tags below by dimension: for example an insight with no number is `product:analysis`, a
|
|
140
|
+
panel answer that dodges the question is `communication`, a SQL join that counts rows twice is
|
|
141
|
+
`product:analysis`, an AI design with no evals is `product:metrics`, a written doc that never says what
|
|
142
|
+
to decide is `product:recommendation`, and a doc far over length is `communication`.
|
|
143
|
+
|
|
58
144
|
The prompt's `## What a strong answer covers` list is the checklist for this prompt: use it to judge
|
|
59
145
|
the dimensions, but never quote it back as the answer.
|
|
60
146
|
|
|
@@ -105,3 +191,52 @@ Then, briefly:
|
|
|
105
191
|
5. **Did last time's feedback stick?** Compare with the earlier reviews of this prompt: say plainly
|
|
106
192
|
whether their gaps came back or were fixed. Skip if this is the first run.
|
|
107
193
|
6. **Next time**: one habit, the same as the header line, with a sentence on how to practise it.
|
|
194
|
+
|
|
195
|
+
## Teaching the review
|
|
196
|
+
|
|
197
|
+
The hub asks for a teaching review: it sends the person's save points (drafts), how far they went in
|
|
198
|
+
the Help tab, and their finish answers. Teach, don't only state findings. Use these exact `##`
|
|
199
|
+
headings, so the hub can lay the review out.
|
|
200
|
+
|
|
201
|
+
**Open** with one or two plain sentences on how it went: the approach, and what cost the time.
|
|
202
|
+
|
|
203
|
+
**`## Between your drafts`**, right after the opening, only when there are two or more save points.
|
|
204
|
+
Name the drafts where something changed (for example `### Drafts 2 to 3: a second user group
|
|
205
|
+
appeared`), quote only the changed lines (at most 12 lines in one block), and add **What it means:** in
|
|
206
|
+
one to three sentences on what the change shows about how they worked. Leave the section out with one
|
|
207
|
+
draft or none.
|
|
208
|
+
|
|
209
|
+
**Plan against what they finished.** In point 2 (Time), compare the structure they set out with what
|
|
210
|
+
the notes ended up covering. When they answered "What changed from your plan", use their words.
|
|
211
|
+
|
|
212
|
+
**`## What to work on, with examples`**, after point 4. One `###` entry per finding that has a gap tag,
|
|
213
|
+
written out in full:
|
|
214
|
+
|
|
215
|
+
- **What we saw:** their words, quoted, or what the drafts showed.
|
|
216
|
+
- **Why it matters:** one to three sentences: what goes wrong, and what an interviewer makes of it.
|
|
217
|
+
- **Before** and **After**: the same short fragment (at most 6 lines each) on a different, made-up
|
|
218
|
+
product, such as a library's audiobook app: a line or two about its user, its goal or its metric,
|
|
219
|
+
first done the weak way, then the better way.
|
|
220
|
+
- **How to practise:** a short drill in words, or another prompt from the bank by its id.
|
|
221
|
+
- **What to say:** one sentence they could say in the interview.
|
|
222
|
+
|
|
223
|
+
The fragment teaches the habit (a goal stated before ideas, users split by behaviour, one measure that
|
|
224
|
+
follows from the goal) on a product that is not this prompt. It never shows this prompt's answer, and
|
|
225
|
+
never names the user group, idea, metric or number that would fix this prompt's gap, so the rule at
|
|
226
|
+
the top still holds.
|
|
227
|
+
|
|
228
|
+
**What I learned.** If they wrote one, say in one line whether it names the main gap.
|
|
229
|
+
|
|
230
|
+
At the end, after point 6, three short sections the hub shows beside the review, in this order:
|
|
231
|
+
|
|
232
|
+
1. **`## How you solved it`**: a list with Result (*solved alone*; *solved with questions* when help
|
|
233
|
+
stopped at step 4 or earlier; *solved with help* from step 5 on), Help (the furthest step and its
|
|
234
|
+
name, or none), Moved you (only when help was used: the step that got them moving), Time and
|
|
235
|
+
Drafts. Add one line on why it counts as that result.
|
|
236
|
+
2. **`## Gaps noted, and why`**: one line per tag in the Gaps header line: why this tag, in plain
|
|
237
|
+
words, and what "fixed" looks like. Then `Not tagged:` with a tag you considered and left out, and
|
|
238
|
+
why. Leave this section out when Gaps is none.
|
|
239
|
+
3. **`## When it comes back`**: `- Redo:` a weekday and date, two or three days out (three when solved
|
|
240
|
+
with help; for solid, a look back in about a week), and how: from a blank page, no help, the target
|
|
241
|
+
time. `- Then:` a sibling prompt from the bank, by its id. `- Daily:` a two or three minute drill
|
|
242
|
+
for the main gap.
|