Predictive Analytics

CAI 3801 – Week 5 Lab Assignment

Predictive Analytics (Forecasting) with Tableau Public + GenAI (ChatGPT/Gemini/Copilot)

This lab focuses only on predictive analytics. Prescriptive analytics will be covered later.|

Submission: Tableau TWBX file + filled template (DOCX or PDF) | Note: change both file names

and include your name as: (e.g., Firstname_Lastname_Week5_Lab.twbx and docx or

pdf)

Academic integrity: You may use GenAI tools for drafting and iteration, but you must verify all

numbers in Tableau and disclose your AI use in the template. No sensitive data in prompts.

Learning goals (what you should be able to do after this lab)

Translate a vague business ask into a measurable predictive question (what to predict, by

when, at what granularity).

Build a time-series view in Tableau and generate a forecast with uncertainty (prediction

intervals).

Assess forecast quality using Tableau’s forecast diagnostics (not ‘vibes’).

Write a short executive-ready narrative that ties forecast results to a business decision and

clearly states assumptions and risks.

Use GenAI as a productivity tool while keeping ownership of context, evidence, and

decisions.

Tools and data

Tableau Public (free) – install or use Tableau Public Desktop.

GenAI tool: ChatGPT/ Gemini / Copilot (choose one).

Dataset: Sample – Superstore (provided in the Week 5 module on Canvas).

Business scenario (choose ONE)

Pick one executive ask below, or write a similar one that fits Superstore. Your job is to turn it

into a predictive question and build a forecast that supports a decision.

A) Operations: ‘We need a sales forecast for the next 4 weeks to plan staffing and inventory.

Where should we prepare for growth or risk?’

B) Finance: ‘Profit has been volatile. Forecast profit for the next month and explain where the

risk is highest (category/region).’

C) Marketing (proxy): ‘Orders are our demand signal. Forecast order volume for the next 4

weeks and identify which customer segments are likely to drive the change.’

D) Your own: A Superstore-friendly question with a clear decision attached (inventory, staffing,

budget, promotion planning).

What you will submit (deliverables)

1) Tableau Public workbook in TWBX

2) Filled student template (DOCX or PDF) with: problem framing, prompt log, screenshots,

forecast diagnostics, and executive brief.

3) AI use note (inside the template): what you used AI for + what you verified + what you

changed.

Step-by-step instructions

Part 1 – Turn a vague ask into a predictive question

1. Choose a scenario (A-D). Write the decision in one sentence (e.g., ‘allocate inventory across

regions for next month’).

2. Define the predictive question using these fields: Target metric (Sales/Profit/Orders),

Forecast horizon (next 4 weeks or month), Granularity (weekly or monthly), and Segment

(overall or by Region/Category/Segment).

3. Define the success metric/constraint for your decision (example: ‘minimize stockouts’ or

‘prepare for regions with >10% forecasted growth’).

4. Use the RTC-OC-QC prompt template below to ask an AI tool for a draft analysis plan. Save

the prompt and 5-8 lines of the output (you will paste excerpts into the template doc).

5. Refine the plan in your own words. You own the final question and scope.

RTC-OC-QC prompt template (copy/paste):

ROLE: You are a business data analyst helping me use Tableau Public.

TASK: Convert this executive ask into a predictive analytics plan.

CONTEXT: I am using Sample – Superstore (Orders). I can build time-series charts and

forecasts in Tableau.

OBJECTIVE/CONSTRAINTS: My forecast horizon is 4 weeks. Budget/time is limited. I must

include uncertainty and validation.

QUALITY CHECKS: Ask 3-5 clarifying questions. Then produce: (1) a final predictive

question, (2) required charts, (3) how to check forecast quality, (4) how to

communicate assumptions and risks.

EXEC ASK: <paste your scenario here>

OUTPUT FORMAT: Use headings and bullets. Keep it under 250 words.

Part 2 – Build the forecast in Tableau Public

6. Open Tableau Public and connect to the Sample – Superstore dataset

7. Create a time series view: drag Order Date to Columns and change it to WEEK (or MONTH).

Drag Sales (or Profit or Quantity for demand) to Rows. This is your baseline trend line.

8. Create a forecast: Analytics pane -> Forecast -> drag ‘Forecast’ onto the view (or Analysis ->

Forecast -> Show Forecast).

9. Open forecast details: right-click the forecast -> ‘Describe Forecast’. Capture the key metrics

(e.g., MAPE/RMSE) and the model notes. Take a screenshot for your submission.

10. Adjust forecast options (if needed): Forecast -> Forecast Options. Set forecast length to 4

weeks or months (depending on your selection of forecast time) and keep seasonality as

Automatic unless you have a clear reason to change it.

11. Add (Region or Category or Segment) based on your scenario i.e., drag your selected

dimension (one of these 3) to color.

12. Add caption (you can use AI to generate the insights on the view as you did in Week 4 Lab

but make sure to edit to reflect your own observations).

— Write 3 factual observations from Tableau (include numbers): current level, trend

direction, and the forecast range etc.

Part 3 – Interpret results and communicate uncertainty

13. Use an AI tool to draft a short executive brief, but only after you provide the verified

Tableau numbers (do not let the AI invent them).

14. Add a ‘Self-check’ section: list what could make the forecast wrong (data gaps, seasonality,

outliers, promo events) and what you would verify next week.

Suggested ‘draft the brief’ prompt:

ROLE: You are my executive writing assistant.

TASK: Draft a 1-page executive brief based ONLY on the verified facts I provide.

CONTEXT: Superstore forecast for next 4 weeks.

VERIFIED FACTS (from Tableau):

– <paste 5-8 bullet facts with numbers, including forecast range/interval>

OUTPUT: 1) 2-sentence summary, 2) 3 insights, 3) 3 recommended actions, 4)

assumptions/risks (at least 3), 5) what to verify next.

CONSTRAINTS: No new numbers. If unsure, say what to verify.

Submission checklist (quick)

Tableau Public workbook link works and is viewable.

Template includes screenshots of: baseline trend, forecast view with intervals, at least one

segmented view, and ‘Describe Forecast’ diagnostics.

Executive brief includes numbers that match Tableau.

AI use note completed (what you used AI for + what you verified).

No sensitive or personal data was used in prompts.

Grading rubric (100 points)

Category What excellent looks like Points

Problem framing Clear predictive question (target, horizon, granularity,

segment) + decision context; success

metric/constraint is stated.

20

Tableau build (forecast) Correct time-series view(s) + forecast shown; at least

one segmented comparison; forecast options are

sensible.

40

Forecast quality +

uncertainty

Uses ‘Describe Forecast’ diagnostics; interprets

prediction intervals; identifies limitations and what to

verify.

20

Executive brief Concise, executive-ready, actionable; all numbers

consistent with Tableau; clear assumptions/risks.

10

Documentation + AI use

note

Prompt excerpts included; transparent AI use; clean

screenshots and labeling.

10

WRITE MY PAPER