DataCareerHub practice companion

Original fictional teaching cases. All resources in this companion are free supplemental practice. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Count completed orders and reconcile sales

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Report the count and EUR sales for completed orders only. Cancelled orders remain in the source but do not enter this KPI. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

6 completed orders; EUR 700. The two cancelled orders contribute EUR 90 to an unfiltered sum and must be excluded.

SELECT COUNT(*) completed_orders, SUM(net_cents)/100.0 sales_eur FROM orders WHERE status='completed';

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Compare two complete months

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Produce April and May sales in chronological order. Use the same status filter in both periods. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

April EUR 300; May EUR 400. Both months are complete within this fictional extract; there is no evidence of seasonality.

SELECT period,SUM(net_cents)/100.0 sales_eur FROM orders WHERE status='completed' GROUP BY period ORDER BY period;

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Calculate channel share with a common denominator

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Find each channel’s share of completed sales. Keep the denominator across channels, not within each channel. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

Store EUR 370 = 52.86%; Web EUR 330 = 47.14%. Rounded percentages sum to 100 here; other datasets can differ by rounding.

WITH s AS (SELECT channel,SUM(net_cents) cents FROM orders WHERE status='completed' GROUP BY channel) SELECT channel,cents,ROUND(100.0*cents/(SELECT SUM(cents) FROM s),2) pct FROM s ORDER BY channel;

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Find average order value at order grain

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Compute average completed order value; explain why averaging line values would answer a different question. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

EUR 116.67, from EUR 700 divided by 6 completed orders. There are no line items in this source.

SELECT ROUND(AVG(net_cents)/100.0,2) aov_eur FROM orders WHERE status='completed';

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Rank completed orders reproducibly

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Rank orders by value and break ties by order_id. Return only the top three. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

R05 18000; R02 15000; R06 12000. ROW_NUMBER gives distinct positions; use DENSE_RANK only when ties should share a rank.

SELECT order_id,net_cents,ROW_NUMBER() OVER(ORDER BY net_cents DESC,order_id) value_rank FROM orders WHERE status='completed' ORDER BY value_rank LIMIT 3;

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Keep excluded records visible

SQL practice · Retail case · 15–20 minutes. Start with retail-setup.sql in a disposable local SQLite database.

Your task

Create an audit showing status, count and amount. This supports reconciliation, not a second definition of revenue. Write the query before opening the solution. Record the expected output grain.

Hint

Apply the status filter before aggregation. State whether the result represents orders, channels or periods. Keep currency units explicit.

Worked solution

cancelled: 2 orders / 9000 cents; completed: 6 orders / 70000 cents. All source amounts total 79000 cents.

SELECT status,COUNT(*) n,SUM(net_cents) cents FROM orders GROUP BY status ORDER BY status;

Check your reasoning

Reconcile the result to the source and describe one limitation. This tiny fictional dataset cannot establish customer behavior or predict future sales.

Use SUMIFS with an explicit status

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Put completed in G2 and calculate completed sales in EUR. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

With CSV columns A:E, enter =SUMIFS(E2:E9,D2:D9,G2)/100. Expected EUR 700. Avoid including the header or treating cents as euros.

Boundary check

Change G2 to cancelled; expected EUR 90. Restore completed afterwards.

Count completed Web orders

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Count orders with channel Web and status completed. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

=COUNTIFS(C2:C9,"Web",D2:D9,"completed") returns 3: R01, R04 and R05. COUNTIFS counts rows here because each row is a unique order.

Boundary check

Check the order IDs manually; a duplicated export row would invalidate the result.

Build a reconciled PivotTable

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Create a pivot with period in Rows, net_cents summed in Values and completed selected in status Filter. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

April 30000 cents; May 40000 cents; grand total 70000 cents. Format or separately divide by 100; renaming the field EUR does not change units.

Boundary check

Refresh after any source change, ensure the table range includes new rows, and compare the grand total to SUMIFS.

Calculate growth without confusing percentage points

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Use April EUR 300 and May EUR 400 to calculate absolute and relative growth. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

If G2=300 and H2=400: =H2-G2 gives EUR 100; =IF(G2=0,"Not defined",H2/G2-1) gives 33.33% when formatted as percent. This is relative growth, not 33 percentage points.

Boundary check

A zero starting value makes relative growth undefined. Show the absolute change rather than silently replacing the denominator with one.

Reconcile inventory counts

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Using inventory-counts.csv columns A:F, calculate expected closing and counted difference. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

In G2 enter =B2+C2-D2-E2; in H2 enter =F2-G2; fill through row 7. P02 expected 15, counted 14, difference -1; every other SKU difference 0.

Boundary check

Sum H2:H7 = -1, but also count nonzero rows; positive and negative errors can otherwise cancel out.

Calculate the overall conversion rate

Excel practice · Use the supplied CSV with headers in row 1. Formulas use English function names and commas; your locale may use translated names or semicolons.

Your task

Using campaign-performance.csv, compute conversions divided by clicks across all six rows. Write the metric definition and output unit before choosing a formula.

Hint

Use a separate calculation area so the raw source stays unchanged. Keep excluded rows and reconcile the total.

Worked solution

=SUM(E2:E7)/SUM(D2:D7) =54/900 =6%. An average of campaign percentages would weight unequal-sized campaigns equally.

Boundary check

Search is 27/270=10%; Social 9/450=2%; Email 18/180=10%. Explain why the overall result is 6%.

Read and validate an order file

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Validate row count, unique IDs and known status values before analysis. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

The file contains eight unique orders. A failed assertion means stop and inspect the input; do not drop records to make the check pass.

import csv
with open('retail-orders.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
assert len(rows)==8
assert len({r['order_id'] for r in rows})==8
assert all(r['status'] in {'completed','cancelled'} for r in rows)

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Filter and sum integer money

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Calculate sales only from completed orders, retaining the raw file. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

Output 700.0 EUR. Sum integer cents before converting for display. This prevents cumulative binary rounding of currency inputs.

import csv
with open('retail-orders.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
done=[r for r in rows if r['status']=='completed']
cents=sum(int(r['net_cents']) for r in done)
assert cents==70000
print(cents/100)

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Group sales by month in Python

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Build a dictionary of completed sales per period without a third-party package. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

April 30000 and May 40000 cents. Sorting YYYY-MM strings works for this validated format, not arbitrary date labels.

import csv
with open('retail-orders.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
from collections import defaultdict
monthly=defaultdict(int)
for r in rows:
    if r['status']=='completed': monthly[r['period']]+=int(r['net_cents'])
assert dict(monthly)=={'2026-04':30000,'2026-05':40000}
print(sorted(monthly.items()))

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Detect duplicate IDs without deleting evidence

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Use the messy file and report repeated IDs. Do not overwrite it. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

With retail-orders-messy.csv the result is R01:2. A repeated ID is a review flag; identical rows may be a repeated export, but conflicting records require a source decision.

import csv
with open('retail-orders-messy.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
from collections import Counter
counts=Counter(r['order_id'] for r in rows)
duplicates={k:v for k,v in counts.items() if v>1}
assert duplicates=={'R01':2}
print(duplicates)

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Create a reproducibility fingerprint

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Fingerprint the input so another learner can identify the exact file you used. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

The digest depends on the exact bytes, including line endings. Save it beside your command, tool version and result. A digest proves byte identity, not data accuracy.

import csv
with open('retail-orders.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
import hashlib
from pathlib import Path
p=Path('retail-orders.csv')
print(hashlib.sha256(p.read_bytes()).hexdigest())

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Handle a zero denominator explicitly

Python practice · Python 3.10+ standard library · Run locally in a folder containing the supplied files.

Your task

Write a rate function and test both ordinary and undefined cases. Explain the expected output before running the code.

Hint

Read with csv.DictReader so column names are explicit. CSV values are strings until converted. Keep parsing failures visible.

Worked solution

54/900 is 0.06. A zero denominator returns None, meaning unavailable, rather than claiming a 0% measured conversion rate. Validate nonnegative counts separately.

import csv
with open('retail-orders.csv',newline='',encoding='utf-8') as f:
    rows=list(csv.DictReader(f))
def rate(numerator,denominator):
    if denominator==0: return None
    return numerator/denominator
assert rate(54,900)==0.06
assert rate(0,0) is None

Extend the test

Change one input in a disposable copy and predict which check should fail. Restore the original before comparing results.

Resolve a missing value using evidence

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

R05 has a blank status in the messy file. Should you fill every blank with completed? Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

No. Flag R05 for review; the independent clean control file identifies it as completed. Record field=status, old=blank, new=completed, source=clean control, reason and reviewer/date. If no control exists, keep it unresolved and disclose exclusion.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Normalize a category without losing the source

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

The messy export has R02 channel " store ". Define a safe correction. Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

Retain channel_raw, apply trim plus a reviewed case mapping to Store, then check the allowed set {Web,Store}. Reject an unexpected value such as wholesale rather than mapping it silently to Store.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Separate repeated exports from conflicting records

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

R01 occurs twice. What must you check before removing a row? Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

Compare every field, document the duplicate occurrence and preserve its source row number. This fixture has an identical repeated row; a derived clean copy may retain one R01. If amounts differed, stop for a source decision instead of keeping the first.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Declare the row meaning before cleaning

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

Someone asks you to make the order table one row per month. Is that cleaning? Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

No. One row currently means one order. Monthly aggregation creates a new analytical dataset and loses order-level detail. Keep both tables, document the transformation and reconcile completed cents before and after.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Protect cents and identifiers during import

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

Why should order IDs remain text and net_cents become an integer? Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

IDs are labels, not quantities; preserve prefixes and any leading zeros. net_cents must parse as a nonnegative integer. A value such as "12 EUR" should be rejected for review, not coerced to zero.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Produce a complete correction log

Data preparation practice · Use the messy source and independent clean control. Corrections belong in a derived output and a change log.

Your task

Describe the three defects and final clean-output checks for retail-orders-messy.csv. Describe the risk of an automatic correction.

Hint

Distinguish formatting fixes from changes in business meaning. Specify what evidence authorizes each correction.

Worked solution

Log duplicate R01, R02 category spelling/spacing and R05 missing status. After evidence-based corrections: 8 unique IDs, known status/channel sets, nonnegative integer cents, completed total 70000. Keep raw file and rejected/duplicate evidence.

What to hand over

Provide raw source fingerprint, old/new values, rule, evidence, affected row IDs and validation totals. A clean-looking table alone is insufficient.

Choose a defensible sales chart

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

Draw April EUR 300 and May EUR 400 as bars. What happens if the y-axis begins at EUR 290? Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

A truncated baseline exaggerates the relative height of bars. Start bars at zero, label EUR and show both periods. A line chart with a clearly labelled narrower scale can answer a different trend question, but disclose the scale.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

Show counts beside conversion rates

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

Search has 27/270 conversions and Email 18/180. Which is better? Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

Both are 10% in this fixture. Show numerator and denominator alongside the rate. Different attribution, audiences and costs prevent a causal ranking; Email has lower observed spend per attributed conversion.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

Explain chart meaning without color alone

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

Create a chart distinguishing Web EUR 330 and Store EUR 370. Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

Use direct text labels and consistent positions, plus accessible colors. A reader must identify each category without relying only on red/green. Provide a two-row data table as an alternative.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

Present an inventory exception honestly

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

P02 count differs by one unit. Write a chart/table heading and an action. Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

Heading: P02 physical count is one unit below recorded closing balance. Show expected 15, counted 14, difference-1. Action: recount and inspect movement records; do not call it theft or shrinkage without evidence.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

Write a precise comparison headline

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

Write one sentence explaining the April-to-May retail change. Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

Completed order value rose from EUR300 to EUR400, an increase of EUR100 (33.3%) in this fictional extract. The chart does not show why it changed or whether the change will persist.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

Design a decision-focused reporting page

Visualization practice · A useful chart makes its units, population and limitations inspectable.

Your task

Choose three metrics and one action for the retail manager. Sketch the view and write an accessible text alternative.

Hint

Start with the decision and the comparison. Avoid decorative marks that compete with the evidence.

Worked solution

Use completed sales EUR700, completed orders 6 and AOV EUR116.67, with monthly split 300/400 and cancelled-order reconciliation 2/EUR90. Ask whether source completeness and status definitions match reporting policy before drawing operational conclusions.

Review checklist

Can someone recover the values, period, population and denominator from the chart and text? State one conclusion the evidence does not support.

The join that doubled revenue

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

An order worth EUR100 has two item rows. After joining order totals to items, SUM(order_total) shows EUR200.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Order totals are repeated at item grain. Aggregate line amounts or aggregate each side to the required order grain before joining. Reconcile unique order count and pre/post amounts. DISTINCT on a monetary amount can wrongly merge different orders with equal values.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Growth is not a percentage-point change

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A chart labels sales growth from EUR300 to EUR400 as +33 percentage points.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Sales increased EUR100 or 33.3% relative to April. Percentage points compare percentages: a rate changing from 10% to 12% rises 2 percentage points, or 20% relatively.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Averages hide unequal group sizes

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A report averages campaign conversion percentages and presents it as the overall conversion rate.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Use sum(conversions)/sum(clicks)=54/900=6%. Taking the simple mean of Search 10%,Social 2%,Email 10% gives 7.33%, a different equally weighted campaign statistic.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Unknown is presented as zero

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A report replaces a missing campaign spend with 0 and claims free conversions.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Keep unavailable cost separate from known zero. Suppress spend-dependent rates until the amount is verified; flag the missing source and document its reporting impact.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

The model knows the future

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A churn predictor includes account_closed_date while predicting churn before closure.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

The feature is not available at prediction time and leaks the outcome. Define the decision timestamp, remove unavailable information, split by appropriate time/entity and compare against a simple baseline.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Correlation becomes a promise

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

Sales increased after a marketing campaign, so a slide says the campaign caused all growth.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

The timing alone does not isolate the campaign from prices, seasonality or other changes. Describe an observed association and propose a comparison or experiment with a defensible causal design.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Only successful records enter the analysis

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A course survey includes only learners who finished and claims all learners found the course easy.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

The sample excludes people who stopped. State response/completion selection, seek feedback from non-completers where appropriate, and avoid generalizing satisfaction to all starters.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

The test set becomes training feedback

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A learner tries twenty model settings and chooses the one with the best final test score.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

The test set has influenced selection and is no longer an untouched estimate. Use training/validation or cross-validation for selection, then evaluate once on a held-out set. Disclose previous reuse.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

A partial month looks like a decline

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A dashboard compares all April with May 1–10 and announces falling monthly sales.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Compare equal elapsed windows or clearly mark partial periods. A projection needs stated assumptions; it is not an observed full-month result.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

A practice project becomes invented employment

Mistake Clinic · Inspect the reasoning before opening the diagnosis.

The questionable claim

A fictional retail exercise is listed as a paid client engagement with a claimed 30% profit improvement.

Investigate

Identify the unit of analysis, denominator, population or information boundary. What extra evidence would be needed? Write a corrected statement first.

Diagnosis and repair

Label it Independent learning project using fictional data. Describe verified work and observed outputs; do not invent employment, clients or business impact. Resume guidance does not guarantee interviews or a job.

Prevention

Add a test or review question that catches the error before publication. Save the original and corrected explanation so your reasoning is auditable.

Retail: a checked monthly sales brief

A shop manager wants to compare April and May completed order value. Establish the definition before proposing actions.

Define the assignment

Audience: an operational manager. Deliver a reproducible analysis, a source-validation log, one useful visual or table and a 250-word decision note. Suggested duration: 60–90 minutes. All source records are fictional.

Investigation plan

Inspect unique order IDs and allowed statuses; filter completed; group by month/channel; reconcile totals; draft a one-page explanation.

Exercises

1. Write the metric definition and grain. 2. Validate keys and units. 3. Calculate independently using two methods. 4. Show excluded/exception records. 5. Write one recommendation and one unanswered question.

Worked control totals

8 source orders, 6 completed, 2 cancelled; completed sales EUR700. April 300, May 400; change 100 /33.3%. Store 370,Web 330; AOV116.67.

Interpretation and limitations

Do not infer profit, customer retention or campaign effects: these fields are absent. A strong answer proposes the next evidence needed and avoids presenting a descriptive result as an established cause.

Self-review rubric

Score 0–2 each for scope, arithmetic, validation, explanation and reproducibility. For each score link to a concrete artifact. A total score is a self-check, not certification or employment evidence.

Retail dataset pack

Original fictional retail records for repeatable learning, with source meaning and known controls.

What each row means

One row per order. Status determines inclusion in the sales metric.

Fields, units and validation

Read data-dictionary.txt before calculations. Keep integer money in cents; quantities/counts are whole nonnegative values. Treat identifiers as labels. Preserve the original input and log transformations.

Known controls

8 source orders, 6 completed, 2 cancelled; completed sales EUR700. April 300, May 400; change 100 /33.3%. Store 370,Web 330; AOV116.67.

Reuse and attribution

Created by DataCareerHub; permission is granted to use and adapt these fictional fixtures for learning and portfolios with acknowledgement. They are not real clients, employers or commercial results. Do not replace them with confidential records.

Suggested independent extension

Create a second synthetic period with documented assumptions, predict the control totals and challenge your implementation. Clearly distinguish your invented extension from the original fixture.

Inventory: reconcile recorded and physical stock

An operations team wants a list of discrepancies before adjusting stock records. A physical count is evidence to investigate, not automatic proof of loss.

Define the assignment

Investigation plan

Calculate opening+receipts-shipped-damaged; compare counts; list every exception; prioritize a recount and movement-log check.

Exercises

Worked control totals

Expected closing by SKU14,15,13,2,11,40; counted 14,14,13,2,11,40. P02 difference-1; all other differences 0. Total expected 95, counted 94.

Interpretation and limitations

No unit costs, count timestamps or movement evidence are supplied; do not estimate loss value or blame a person. A strong answer proposes the next evidence needed and avoids presenting a descriptive result as an established cause.

Self-review rubric

Inventory dataset pack

Original fictional inventory records for repeatable learning, with source meaning and known controls.

What each row means

One row per product SKU. Compare a movement-based expected closing quantity with an independent count.

Fields, units and validation

Known controls

Expected closing by SKU14,15,13,2,11,40; counted 14,14,13,2,11,40. P02 difference-1; all other differences 0. Total expected 95, counted 94.

Reuse and attribution

Suggested independent extension

Create a second synthetic period with documented assumptions, predict the control totals and challenge your implementation. Clearly distinguish your invented extension from the original fixture.

Marketing: evaluate attributed conversion efficiency

A marketing manager wants an allocation discussion. Separate observed attributed performance from incremental impact.

Define the assignment

Investigation plan

Check campaign-period uniqueness and count domains; aggregate counts/costs; calculate ratios from totals; compare channels and disclose attribution limits.

Exercises

Worked control totals

Spend EUR500, clicks 900, attributed conversions 54, conversion rate 6%, spend/conversion EUR9.26. Search spend 270/27=10; Social 180/9=20; Email 50/18=2.78 EUR/conversion.

Interpretation and limitations

No order values, user overlap or randomized comparison are provided. Do not claim ROAS, profit or causal lift. A strong answer proposes the next evidence needed and avoids presenting a descriptive result as an established cause.

Self-review rubric

Marketing dataset pack

Original fictional marketing records for repeatable learning, with source meaning and known controls.

What each row means

One row per campaign and month. Conversions use a fictional attribution rule.

Fields, units and validation

Known controls

Spend EUR500, clicks 900, attributed conversions 54, conversion rate 6%, spend/conversion EUR9.26. Search spend 270/27=10; Social 180/9=20; Email 50/18=2.78 EUR/conversion.

Reuse and attribution

Suggested independent extension

Create a second synthetic period with documented assumptions, predict the control totals and challenge your implementation. Clearly distinguish your invented extension from the original fixture.

Retail portfolio: from rough draft to evidence

Annotated model structure for an independent learning project. This is an example of presentation quality, not a completed project to claim as your own.

A weak opening

“I analyzed data and made a dashboard.” This gives no question, source, inclusion rule or independently checkable result.

A stronger opening

“In a fictional eight-order retail extract, I compared completed order value across April and May, reconciled excluded cancellations and documented a EUR100 observed increase.”

Why the stronger version works

It names the scope, independent work, method and evidence boundary. It does not claim a real employer, client impact or an outcome outside the extract.

Suggested repository structure

README with question and reproduction commands; data with dictionary and permissions; src with your implementation; tests with expected totals; outputs with chart/table and decision note. Keep answer keys separate from your own attempt.

Evidence to annotate

8 source orders, 6 completed, 2 cancelled; completed sales EUR700. April 300, May 400; change 100 /33.3%. Store 370,Web 330; AOV116.67. Explain why each validation exists. Include one failed check you investigated and how you resolved it, rather than hiding all intermediate problems.

Your independent task

Reproduce the analysis yourself, then add a documented extension or alternative presentation. Credit the original fixture and explain your contribution. Do not infer profit, customer retention or campaign effects: these fields are absent.

Career note

Portfolio examples and resume guidance do not guarantee interviews or a job. Use your own verified work; do not copy the sample as commercial experience.

Inventory portfolio: from rough draft to evidence

Annotated model structure for an independent learning project. This is an example of presentation quality, not a completed project to claim as your own.

A weak opening

“I analyzed data and made a dashboard.” This gives no question, source, inclusion rule or independently checkable result.

A stronger opening

“I reconciled six fictional product records against independent closing counts and isolated a one-unit discrepancy for P02 without assuming its cause.”

Why the stronger version works

It names the scope, independent work, method and evidence boundary. It does not claim a real employer, client impact or an outcome outside the extract.

Suggested repository structure

Evidence to annotate

Expected closing by SKU14,15,13,2,11,40; counted 14,14,13,2,11,40. P02 difference-1; all other differences 0. Total expected 95, counted 94. Explain why each validation exists. Include one failed check you investigated and how you resolved it, rather than hiding all intermediate problems.

Your independent task

Reproduce the analysis yourself, then add a documented extension or alternative presentation. Credit the original fixture and explain your contribution. No unit costs, count timestamps or movement evidence are supplied; do not estimate loss value or blame a person.

Career note

Portfolio examples and resume guidance do not guarantee interviews or a job. Use your own verified work; do not copy the sample as commercial experience.

Marketing portfolio: from rough draft to evidence

Annotated model structure for an independent learning project. This is an example of presentation quality, not a completed project to claim as your own.

A weak opening

“I analyzed data and made a dashboard.” This gives no question, source, inclusion rule or independently checkable result.

A stronger opening

“I compared attributed conversion efficiency across three fictional campaigns, calculated rates from aggregated counts and documented why these comparisons do not establish incremental lift.”

Why the stronger version works

It names the scope, independent work, method and evidence boundary. It does not claim a real employer, client impact or an outcome outside the extract.

Suggested repository structure

Evidence to annotate

Spend EUR500, clicks 900, attributed conversions 54, conversion rate 6%, spend/conversion EUR9.26. Search spend 270/27=10; Social 180/9=20; Email 50/18=2.78 EUR/conversion. Explain why each validation exists. Include one failed check you investigated and how you resolved it, rather than hiding all intermediate problems.

Your independent task

Reproduce the analysis yourself, then add a documented extension or alternative presentation. Credit the original fixture and explain your contribution. No order values, user overlap or randomized comparison are provided. Do not claim ROAS, profit or causal lift.

Career note

Portfolio examples and resume guidance do not guarantee interviews or a job. Use your own verified work; do not copy the sample as commercial experience.

Build an honest facsimile résumé

A facsimile résumé is a sample layout that shows how to organize real evidence. It is not permission to copy someone else’s history. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

1. Choose a role and gather evidence

Read a genuine job description and list three recurring requirements. For each, identify a real work example, independently completed learning project or skill gap. Do not claim a tool because it appears in a job posting.

2. Use a clear structure

Start with name, professional contact details and a relevant portfolio link. Follow with a concise summary, demonstrable skills, clearly labelled projects, actual experience and genuine education/training. Use descriptive headings and a readable single-column layout.

3. Turn work into defensible bullets

Use action + scope + method + checked result or output. Example: “Reconciled eight fictional orders to six completed orders and EUR700 completed sales using an explicit SQL status filter.” This describes evidence, not a profit improvement or paid client result.

4. Tailor without fabricating

Use the employer’s terminology where it accurately describes your work. Explain career changes through transferable evidence. Label coursework and fictional data openly; never invent clients, employers, certificates, dates or performance percentages.

5. Check readability and privacy

Use selectable text and conventional section headings. Open the exported file, copy its text into a plain editor and check reading order. Follow the employer’s file-format instructions. Do not claim any design will guarantee an applicant-tracking-system score or acceptance. Include only contact details you choose to share.

6. Make your own version

Download the fictional sample and evidence worksheet. Replace placeholders with genuine details and remove sections you cannot support. Use the linked Professional Resume Builder for guided fields, a private preview and PDF export. This website processes temporary documents; existing learners can save versions when storage is configured. No external AI service receives your resume.

Before applying

Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment. Read every line aloud, verify links/dates and ask whether you could demonstrate each claim in an interview.

Map a job description to real evidence

Career preparation using your own demonstrable skills. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Method

Build a four-column matrix: requirement, actual evidence, confidence and next practice.

Worked example

A posting requests SQL joins, dashboards and communication. SQL evidence might be your own query plus reconciliation; a dashboard screenshot alone does not demonstrate refresh correctness. Mark unsupported requirements as practice gaps.

Practice

Choose three requirements, link one artifact each and schedule a focused exercise. Do not paste private employer documents into an external tool.

Self-review

Check accuracy, relevance, evidence and clarity. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Explain your analysis in an interview

Career preparation using your own demonstrable skills. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Method

Practise a two-minute answer: question, source/grain, method, validation, finding, limitation.

Worked example

For the retail case: completed sales 300→400 EUR; verified unique orders and exclusions; report 100 EUR/33.3% increase while avoiding a causal claim. An interviewer can inspect the query and output.

Practice

Record yourself locally or rehearse with a friend. If asked about an unfamiliar tool, state your experience accurately and explain how you would investigate.

Self-review

Check accuracy, relevance, evidence and clarity. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Complete a realistic take-home task

Career preparation using your own demonstrable skills. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Method

Use the retail case with a 90-minute budget:15 scope,35 analysis,20 communication,20 checks.

Worked example

A complete small submission contains a README, source attribution, reproducible query/script, checked result and a short decision note. Prioritize correctness over adding a complex model without a reason.

Practice

Use the rubric: scope, correctness, validation, explanation and reproducibility. Respect the actual employer’s instructions about AI assistance and external help.

Self-review

Check accuracy, relevance, evidence and clarity. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Present a project with limitations

Career preparation using your own demonstrable skills. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

Method

Prepare five slides or sections: decision, data, method, evidence, next step.

Worked example

Use one finding per view. Show a reconciled number and a limitation together: “P02 count is one unit below expected; recount before adjusting records.” A polished claim without source evidence is weaker.

Practice

Practise answering: What did you do yourself? What failed? What would change your recommendation? Credit collaborators and teaching sources.

Self-review

Check accuracy, relevance, evidence and clarity. Resume guidance and examples are for learning and preparation only. They do not guarantee interviews, a job offer or employment.

SQL: query grain and validation

A short retrieval session. Try the questions before revealing the explanations; revisit them after a few days and again in a later week.

Recall without looking

What does one result row represent? When does a join amplify rows? Why filter before aggregation?

Check and explain

Result grain must follow the question. A one-to-many join repeats parent values; reconcile counts and sums. WHERE filters source rows, HAVING filters groups.

Transfer task

Write monthly completed sales from memory, then compare with the SQL monthly challenge. Change the task to channel share and explain the denominator.

Record your next review

Write which concept you missed and one new example. Revisit after 2–3 days, then about a week; adjust the interval to your needs. This is a suggested study routine, not an automated reminder or saved score.

Excel: formulas and controls

A short retrieval session. Try the questions before revealing the explanations; revisit them after a few days and again in a later week.

Recall without looking

When does COUNTIFS count orders? Why refresh a PivotTable? Why can an average of rates mislead?

Check and explain

COUNTIFS counts rows, so the source must have unique order grain. Refresh and range checks prevent stale outputs. Aggregate numerators/denominators for a population rate.

Transfer task

Rebuild SUMIFS and an inventory reconciliation without looking. Compare them against case controls, then introduce one duplicated row in a copy.

Record your next review

Python: types and reproducibility

A short retrieval session. Try the questions before revealing the explanations; revisit them after a few days and again in a later week.

Recall without looking

What type does csv.DictReader return? Why hash an input? How should a zero denominator behave?

Check and explain

CSV fields begin as strings. Hashes identify bytes, not truth. A zero denominator is undefined and needs an explicit unavailable result; zero is a different claim.

Transfer task

Read the retail file, validate IDs/status and sum integer cents from memory. Test a malformed amount and describe the failure policy.

Record your next review

Analysis: evidence and interpretation

A short retrieval session. Try the questions before revealing the explanations; revisit them after a few days and again in a later week.

Recall without looking

What population does the metric cover? Is a month complete? Does an association establish a cause?

Check and explain

Define exclusions and observation windows. Partial periods need comparable windows. A descriptive comparison cannot isolate causal effects without an appropriate design.

Transfer task

Write two supported findings and two unsupported claims from the marketing case. For each unsupported claim name the missing evidence.

Record your next review

ML: leakage and evaluation

A short retrieval session. Try the questions before revealing the explanations; revisit them after a few days and again in a later week.

Recall without looking

When is a feature available? Which data choose hyperparameters? Why compare a baseline?

Check and explain

Prediction-time availability prevents leakage. Use training/validation or cross-validation for selection and reserve an untouched test. A simple baseline establishes whether complexity adds useful evidence.

Transfer task

Sketch a churn prediction timeline, flag closure-derived features and define an evaluation metric aligned with errors. This exercise does not train a model.

Record your next review

Write a useful executive summary

Business Communication Lab · Translate a technical result into a decision someone can assess.

Method

Lead with the answer, then scope, evidence, limitation and next action.

Worked example

“Completed order value increased EUR100 (33.3%) from April to May in this fictional extract. Six completed orders total EUR700; cancelled orders were excluded. The extract does not identify the cause. Confirm source completeness and investigate product mix before changing plans.”

Your exercise

Reduce that paragraph to 80 words without losing the comparison or limitation. Compare your wording to the source totals.

Review criteria

Check audience, units, period, inclusion rule, evidence and next action. Avoid claiming causation, certainty or measured impact that your source does not establish.

Ask better stakeholder questions

Business Communication Lab · Translate a technical result into a decision someone can assess.

Method

Ask what decision is pending, who will act, which population counts and what evidence could change the decision.

Worked example

“Before comparing campaigns, do conversions use the same attribution window? Is spend complete? Are audiences overlapping? Is the decision about attributed cost efficiency or incremental profit?”

Your exercise

Write five questions for the inventory manager. Separate necessary metric definitions from questions that would merely be interesting.

Review criteria

Check audience, units, period, inclusion rule, evidence and next action. Avoid claiming causation, certainty or measured impact that your source does not establish.

Explain uncertainty without hiding the answer

Business Communication Lab · Translate a technical result into a decision someone can assess.

Method

Separate observed facts, assumptions and unverified explanations. Avoid replacing uncertainty with vague confidence.

Worked example

Fact: P02 counted 14 versus expected 15. Assumption: movement records and count refer to the same time. Unverified explanation: an omitted shipment. Next action: recount and align timestamps.

Your exercise

Rewrite “Our stock process is unreliable” as a scoped observation and a proportionate next action. Do not assign blame from a single record.

Review criteria

Check audience, units, period, inclusion rule, evidence and next action. Avoid claiming causation, certainty or measured impact that your source does not establish.

Write a reproducible handover

Business Communication Lab · Translate a technical result into a decision someone can assess.

Method

Record exact sources, required tool/version, run command, expected totals, exceptions and known limits.

Worked example

A colleague should run case-reference.py beside the three named CSV files and compare its JSON to case-expected-results.json. If the source changed, they should inspect differences rather than overwrite expected results.

Your exercise

Write a handover for one case that someone can follow without asking where the input lives. Keep confidential data and credentials out of it.

Review criteria

Check audience, units, period, inclusion rule, evidence and next action. Avoid claiming causation, certainty or measured impact that your source does not establish.

Verify AI-generated SQL

Responsible AI Practice · No API integration is enabled; exercises can be completed by inspecting the supplied example.

Scenario

An assistant suggests SELECT SUM(net_cents) FROM orders and calls the result completed revenue.

Evidence-based response

The query lacks the completed filter and returns 79000 cents. The defined metric is 70000. Ask for assumptions, inspect schema/grain and independently reconcile before using the result.

Practice

Request an explanation of the status rule, then write the corrected query yourself. Treat generated code as an untrusted draft; run only in a disposable local practice database.

Boundary

The platform does not automatically verify external AI answers. You remain responsible for checking source meaning, calculations, privacy and permission before using a result.

Challenge an invented explanation

Responsible AI Practice · No API integration is enabled; exercises can be completed by inspecting the supplied example.

Scenario

An assistant says Email caused the sales increase, using only the two separate fictional files.

Evidence-based response

The files have no campaign-to-order key or causal design. Attributed efficiency and retail sales are separate descriptive facts. Ask the assistant to mark unsupported statements and cite exact fields.

Practice

Write a revised paragraph separating measured values from hypotheses. A fluent explanation does not create evidence.

Boundary

The platform does not automatically verify external AI answers. You remain responsible for checking source meaning, calculations, privacy and permission before using a result.

Prepare safe learning prompts

Responsible AI Practice · No API integration is enabled; exercises can be completed by inspecting the supplied example.

Scenario

You want help with a work spreadsheet containing names and commercial records.

Evidence-based response

Use the original fictional fixtures or a genuinely non-sensitive schema/example. Merely removing names can leave identifying details. Follow your organization’s policy and tool settings before sharing any work data.

Practice

Describe the task using invented columns/values. Do not upload credentials, private resumes, personal information or customer records into this learning app.

Boundary

The platform does not automatically verify external AI answers. You remain responsible for checking source meaning, calculations, privacy and permission before using a result.

Document AI assistance and your contribution

Responsible AI Practice · No API integration is enabled; exercises can be completed by inspecting the supplied example.

Scenario

A project uses an assistant to draft a formula and summarize findings.

Evidence-based response

Record tool/model if known, date, task assisted, output you adopted, checks you performed and your independent contribution. Respect any assessment/employer restriction. Do not claim generated work as independently produced.

Practice

Add an assistance note to your case README, or explicitly say no AI assistance if accurate. Explain one suggestion you rejected and why.

Boundary

The platform does not automatically verify external AI answers. You remain responsible for checking source meaning, calculations, privacy and permission before using a result.

Set up a disposable SQLite practice database

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

SQLite command-line tool or a local SQLite editor; no production database access required.

Steps and expected result

Expected 8. If sqlite 3 is not found, install from the official SQLite distribution or use a trusted local editor, following your device’s policy. A second import into the same database fails because the table already exists; use a fresh practice database. Never point the exercise at a work database.

sqlite3 practice.db < retail-setup.sql
sqlite3 practice.db "SELECT COUNT(*) FROM orders;"

Verification

Check the input row count and case control totals before attempting a larger exercise. Record your operating system/tool version when asking for support. Never include passwords or private source records.

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Run a local Python practice script

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

Python 3.10+ and the standard library. These supplemental case references require no pip packages.

Steps and expected result

Run from the extracted pack directory. Expected JSON contains retail sales_cents 70000. On Windows the launcher may be py -3. A missing file error means the three CSV files are not beside the script; do not fix it by deleting validation.

python3 --version
python3 case-reference.py

Verification

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Import CSV without corrupting fields

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

Excel desktop or another spreadsheet editor with CSV import. Exact menus vary by version.

Steps and expected result

Import using UTF-8 and comma delimiter. Set IDs to text and money fields to whole numbers. Confirm 8 retail rows. If all data appears in one column, choose the correct delimiter. Dates should remain the documented YYYY-MM labels. English formula names/separators may differ by locale.

Verification

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Keep a reproducible practice folder

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

A local folder with separate raw, working, outputs and notes areas.

Steps and expected result

Extract the complete pack rather than opening one file inside a ZIP. Keep original sources unchanged and put edits in working. Record commands relative to the project folder. If output differs across systems, first check input bytes, parsing and versions.

Verification

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Read an error before changing code

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

A failed check is useful evidence; capture its message and the smallest reproducible input.

Steps and expected result

For ValueError converting net_cents, inspect the exact offending value and source row. For KeyError, compare required header spelling and accidental spaces. For no rows, inspect filters and status case. Do not use a blanket try/except that turns every error into 0.

Verification

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Document and isolate tool environments

Tool Setup & Troubleshooting · Work on your own device with fictional practice data.

Prerequisites

Use official tool documentation and a dedicated local environment when a course needs packages.

Steps and expected result

The supplemental enrichment case-reference.py needs no third-party packages. Existing deep learning/BI/cloud courses have different prerequisites and explicitly documented runtime limits; a successful Python setup does not prove Spark, Docker or Power BI is installed.

python3 -m venv .venv
# Activate using the command for your OS and shell.
# Install only the requirements supplied by the specific course.

Verification

Independent check

Explain which output proves the setup works and which capabilities remain untested. Do not disable security settings to make an exercise run.

Grain

The meaning of one row: order, order line, customer-month or another explicitly defined unit.

Example

An order-total column repeated after joining to items cannot be summed as if each row were still an order.