WesBid: From Excel Calculator to an Explainable Amazon PPC Decision Engine
Context
Most Amazon PPC operators start with spreadsheets.
We did too.
The original workflow was simple:
- Export reports
- Paste data into Excel
- Review bid recommendations
- Manually update campaigns
The calculator worked.
But eventually a new problem appeared.
The calculator could tell us what bid to use, but it could not explain:
- Why the bid changed
- Whether the signal was trustworthy
- Which search terms were causing waste
- Which targets deserved scaling
- Whether performance was stable or accidental
The spreadsheet provided answers.
The operator still had to provide judgment.
Problem Statement
Amazon PPC optimization is rarely limited by math.
It is limited by context.
Traditional bid calculators struggle because they treat every target equally:
- Exact Match
- Broad Match
- Auto Targets
- ASIN Targets
All receive similar treatment despite behaving differently.
This creates several issues:
- Over-scaling unstable winners
- Cutting keywords too early
- Missing waste hidden inside search terms
- Ignoring placement opportunities
- Treating learning-phase targets as failures
The result is reactive optimization.
Core Idea
WesBid was built around a simple principle:
A bid recommendation without context is not intelligence.
Instead of acting as a calculator, WesBid became a decision engine.
Every recommendation is attached to:
- Performance state
- Confidence level
- Learning phase
- Placement opportunity
- Search term behavior
The goal is not merely to calculate bids.
The goal is to help operators make better decisions.
Evolution of the System
Phase 1: Excel Calculator
The first version was heavily inspired by traditional PPC spreadsheets.
Inputs:
- Clicks
- Orders
- Spend
- Sales
Outputs:
- Suggested bid adjustments
Effective.
But limited.
Phase 2: Bid Logic Engine
The next step was building automated recommendation logic.
The system introduced:
- Win states
- Neutral states
- Loss states
Targets were classified automatically.
This removed much of the repetitive analysis work.
Phase 3: Explainable Decision Engine
The real breakthrough happened when explanations became first-class outputs.
Instead of saying:
Suggested Bid = $0.84
The system now explains:
Inefficient ACOS detected. Stable volume confirmed. Trimming bid toward target CPC.
This dramatically improves operator trust.
Architecture Overview
Input Layer
Amazon exports:
- Campaign reports
- Targeting reports
- Placement reports
- Search term reports
Intelligence Layer
WesBid processes:
- ACOS
- ROAS
- CPC
- Click volume
- Conversion volume
- Search term behavior
- Placement performance
Decision Layer
Every target receives:
- State classification
- Confidence score
- Suggested action
- Bid recommendation
- Supporting explanation
State-Based Optimization
The engine operates using behavioral states.
🟢 WIN
Strong performance signals.
Examples:
- Orders generated
- Profitable ROAS
- Stable conversion behavior
Suggested action:
Protect and scale.
🟡 LEARNING
Insufficient evidence.
Examples:
- Low clicks
- Limited spend
- Early-stage exploration
Suggested action:
Hold and monitor.
🔴 LOSE
Clear waste signals.
Examples:
- Significant spend
- No conversions
- Poor profitability
Suggested action:
Reduce, negate, or pause.
Confidence Weighting System
One conversion is not the same as ten conversions.
The system now adjusts its aggressiveness based on confidence.
Low Confidence
1 Order
The engine slows scaling behavior.
Reason:
A single conversion may be accidental.
Medium Confidence
2–3 Orders
The engine cautiously scales.
Reason:
Evidence exists but remains limited.
High Confidence
4+ Orders
The engine unlocks full optimization speed.
Reason:
The target has demonstrated repeatable behavior.
Top-of-Search Intelligence
Modern Amazon PPC increasingly rewards premium placement ownership.
WesBid monitors:
- Top-of-Search Impression Share
- Conversion performance
- Stability requirements
The engine can identify targets that deserve additional placement pressure.
However:
A placement boost is only applied if stability conditions are met.
This prevents reckless scaling.
Search Term Intelligence
One of the biggest hidden PPC problems is query-level waste.
A keyword may appear profitable overall while a handful of search terms consume budget with no conversions.
WesBid exposes this layer directly.
Observing
0–2 clicks
No action required.
Weak Waste Signal
3–5 clicks without orders
Monitor closely.
Hard Negative Candidate
6+ clicks without orders
Strong candidate for negative exact implementation.
This allows operators to isolate waste without harming productive traffic.
Manual Bid Override
One unexpected lesson emerged during development.
Operators still need control.
Amazon's live bids frequently differ from exported reports.
To solve this, WesBid introduced manual bid overrides.
Benefits:
- Align recommendations with live account data
- Instantly recalculate recommendations
- Maintain export consistency
- Preserve operator control
This feature became one of the most useful additions to the platform.
The Unexpected Insight
WesBid did not replace the spreadsheet.
It absorbed it.
The original calculator still exists inside the system.
The difference is that the calculator now understands context.
The spreadsheet answered:
What should my bid be?
WesBid answers:
What should my bid be, why should it change, how confident are we, and what happens if we are wrong?
That is a completely different class of tool.
Technical Stack
- TypeScript
- React
- Vite
- State-driven optimization engine
- Explainable recommendation framework
- CSV ingestion pipeline
- Export-ready workflow architecture
Impact
Before
- Spreadsheet calculations
- Manual interpretation
- Hidden search term waste
- Static recommendations
After
- Explainable bidding decisions
- Confidence-weighted scaling
- Search term intelligence
- Placement-aware optimization
- Operator-first workflow
Live System
WesBid Dashboard:
Closing Perspective
WesBid is not an automation tool.
It is an operator tool.
The objective was never to replace judgment.
The objective was to eliminate repetitive analysis so judgment could be applied where it matters most.
The journey started with a spreadsheet.
The destination became a decision engine.