ONE-PAGE GUIDE: HOW TO REFRAME A QUESTION INTO AN OPTIMIZATION MODEL
⸻
1. Start with a precise decision question
Write exactly one sentence:
“Given [resources/limits], how should I choose [things I control] to best achieve [goal]?”
Examples:
• “Given RM1M marketing budget, how should I split it across channels to maximise sales while keeping risk acceptable?”
• “Given my production capacity and ingredient stock, which SKUs and volumes should I produce to maximise profit?”
Keep this sentence as the “title” of your model.
⸻
2. Identify decision variables (what you can control)
Ask: What am I actually choosing?
Make a small table:
Variable Meaning Unit Min Max
x_TikTok Budget assigned to TikTok RM 0 1M
x_Meta Budget assigned to Meta RM 0 1M
x_Influencer Budget assigned to influencers RM 0 1M
These are the levers your solver will set.
In your optimisation tools, these correspond to:
• “items” with quantities (allocation/portfolio/formulation), or
• “tasks” with timing/resources (scheduling).
⸻
3. Define the objective (what “best” means)
Ask: If I could optimise only one thing, what is it?
Common patterns:
• Maximise total profit.
• Maximise expected ROI.
• Minimise total cost.
• Maximise sales / reach subject to profit ≥ target.
• Maximise worst-case performance (robust).
Write it as a formula idea (not necessarily full math):
• “Total expected profit = Σ (budget_channel × margin_per_RM_channel).”
• “Total cost = Σ (quantity_i × unit_cost_i).”
If you have multiple goals, either:
• Combine them into one score:
Score = 0.7 × profit – 0.3 × risk
• Or use priorities:
1. Maximise profit,
2. Subject to risk ≤ R,
3. Subject to minimum presence in key channels.
In your tools, this becomes the objective block (maximize/minimize + weights).
⸻
4. List hard constraints (rules that cannot be broken)
Ask three simple questions:
1. Resource constraints
• Money, people, hours, capacity, inventory.
• Examples:
• x_TikTok + x_Meta + x_Influencer ≤ 1,000,000 (total budget)
• Σ (production_SKU × hours_per_unit_SKU) ≤ available_hours
2. Business / regulatory constraints
• Min/max per channel or SKU, nutrition ranges, legal limits.
• Examples:
• x_Meta ≥ 100,000 (minimum presence)
• Protein per serving ≥ 20 g; sugar per serving ≤ 5 g.
3. Logical constraints
• “Either-or”, “if-then”, group counts.
• Examples:
• Cannot launch both SKU_A and SKU_B together (mutex).
• If we launch SKU_X, we must also allocate pack budget Y.
• At least 3 channels must receive non-zero budget.
Write them first in plain language bullets, then map them into:
• resources + item requirements, and
• additional constraint structures (mutex, if-then, groups) in your solver.
⸻
5. Represent uncertainty (optional but valuable)
Ask: What are the key unknowns that matter?
Examples:
• Demand for each SKU.
• ROAS per channel.
• Ingredient prices.
Choose representation:
1. Scenarios
• Define discrete futures: best / base / worst, or a few named scenarios.
• Each scenario has a full set of numbers (e.g. ROAS_TikTok, ROAS_Meta…).
2. Distributions / Monte Carlo
• For each parameter, define:
• Mean, variance, and possibly min/max or a distribution type.
• Use your Monte Carlo/robust tools to:
• Maximise expected value, or
• Maximise a percentile (e.g. 10th percentile performance).
This turns “I’m not sure” into structured risk.
⸻
6. Select the model pattern (which solver to call)
Map your decision to a standard pattern:
• “How to split limited resources across options?”
→ Allocation model.
• “How to mix ingredients/formulas to hit targets at best cost?”
→ Formulation / blending model.
• “Which projects/SKUs/investments to choose under a budget?”
→ Portfolio / selection model.
• “How to schedule tasks over time with limited resources?”
→ Scheduling / RCPSP model.
• “How to choose a plan that is safe under uncertainty?”
→ Robust / stochastic / Monte Carlo version of the above.
⸻
7. Reusable mini-template (copy/paste for any decision)
Use this structure:
1. Decision question:
“Given ___, how should I choose ___ to best achieve ___?”
2. Variables:
List each decision variable with unit, min, max.
3. Objective:
What are we maximising or minimising?
4. Constraints:
• Resources (budget, capacity, stock).
• Business/regulatory rules.
• Logical rules (if-then, either-or, counts).
5. Uncertainty (optional):
• What’s uncertain?
• Scenarios or distributions?
6. Model pattern:
Allocation / formulation / portfolio / scheduling / robust.
That’s the full “question → model” reframing on one page.
