Add an independent extension and complete the final review
Learn the reasoning. Try the task. Compare your answer.
Prepared for DataCareerHub · Curriculum version 2026-10-15.6
Save your Northstar evidence, compare five reference outputs and record an explicit self-review using this lesson’s existing permissions. No SQL is executed and written work is not automatically graded.
Open or resume saved investigation →By the end of this lesson
- Choose a bounded extension without inventing new evidence.
- Test whether the conclusion changes under a stated scenario.
- Review the complete project using a practical evidence-based rubric.
Fictional guided portfolio · Source from Module 4 · Local spreadsheet or SQLite · Open the worked case study ↗
Understand the method
The final stage makes the project more than a reproduction of worked answers. Choose one bounded extension, state its question before calculating and show how you checked the result. You do not need to invent a new dataset or a large predictive model. A carefully defined extension can demonstrate independent judgment while remaining grounded in the supplied evidence.
Choose a focused extension
One option is to compare April’s mean with its median and with a clearly labelled A020-exclusion scenario. Another is to examine delivery measurement coverage by channel and period. A third is to redesign the case-study explanation for a different reader, such as a source-data owner, while keeping the analytical facts unchanged. Choose a question that the available fields can answer and explain why it matters.
For the coverage extension, the eligible population is still completed orders. April web has 12 measured durations among 18 orders, while store has two among two. May web has one measured duration among two, while store has 18 among 18. These figures show where collection coverage differs. They do not establish that one channel delivers faster, because the unmeasured durations remain unknown and group sizes vary.
Keep the baseline separate
If you extend the outlier analysis, April’s baseline is 20 orders, €2,900 and mean €145. Excluding verified A020 changes April to 19 orders, €1,900 and mean €100. Compare that scenario with May’s €171 AOV only with the changed population disclosed. The baseline result is not overwritten, and the scenario does not prove the large order was erroneous.
Similarly, a hypothetical equal-mix comparison is not a corrected actual result. State weights, period, eligible records and the purpose of the scenario. When a conclusion changes under a scenario, explain the sensitivity rather than choosing the more attractive answer. A project can show uncertainty and remain useful.
Verify a changed input
Use a disposable copy to change one verified amount or a missing duration. An amount change should update the corresponding monthly and channel totals without changing order count. Replacing an unknown duration with a known zero should increase measured coverage and retain zero in the observed mean. Restore the original fixture before packaging baseline results. Record the test rather than leaving an unexplained altered source in the final folder.
Review the complete submission
Use five categories worth zero to two points each: scope/provenance, preparation/reproducibility, metrics/checks, dashboard/explanation and independent extension/review. Award two for complete and correct evidence, one for partial evidence and zero for missing or materially wrong work. Aim for eight or more with no zero in preparation/reproducibility or metrics/checks. This is a self-review, not an automatically verified assessment.
Finish with a two-minute walkthrough: question, method, main finding, qualifying context, next action and independent extension. Keep the evidence pack editable and its limitations visible. Completing the six-module sequence produces guided practice you can refine further; it does not imply an accredited qualification, a job outcome or permission to present fictional business impact as real.
Practical exercise
Choose one extension and write its question, population, method and checks before calculating. Produce a small result table and a 60–100 word interpretation. Perform one changed-input test in a disposable copy and record expected/observed behaviour, then restore the baseline source.
Final deliverable: your complete project, independent extension, ten-point self-review and two-minute presentation outline. Use review_rubric.txt. Aim for at least eight with no zero in preparation/reproducibility or metrics/checks; no course upload, payment or public registration is needed.
Open the worked solution
| Extension: delivery coverage | April | May |
|---|---|---|
| Web measured / eligible | 12 / 18 = 66.7% | 1 / 2 = 50% |
| Store measured / eligible | 2 / 2 = 100% | 18 / 18 = 100% |
| All measured / eligible | 14 / 20 = 70% | 19 / 20 = 95% |
A worked interpretation is: “In the supplied completed-order population, missing delivery measurements are concentrated in web orders. April web coverage is 12/18 and May web coverage is 1/2, while both store groups are fully measured. The overall coverage increase therefore depends partly on changing group composition. The small May web group and unknown durations prevent a claim about channel delivery speed. Investigate the collection process before comparing all-order service outcomes.”
For the outlier extension, April excluding A020 has 19 orders, €1,900 and AOV €100; the €145 April baseline remains. A full submission earns two points in each rubric category when the evidence supports it. A copied case study with no implemented method or independent extension cannot be treated as a complete independent project merely because it looks polished.
One more question.
How these learning materials are prepared and maintained · Report a correction: support@datacareerhub.io
Read and practice without an account. Sign in with an existing account to save lesson completion.
Review the module checklist →Review the complete course →Fictional training data. Portfolio practice is not a claim of client work or an accredited qualification. Explore practical lessons, worked examples and projects. Follow a published path, practise the method and build work you can explain.