Campaign Intelligence System
9 min read
Campaign Intelligence is Smart Cycle Discounts’ decision-making layer. It looks at every campaign in your store and answers one question for you: “What should I do with this campaign right now?”
The answer is always one of five clear states (a “verdict”) — never a vague score you have to interpret. The same verdict is shown wherever you encounter the campaign: in the wizard, on the campaigns list, on the dashboard, and in the overview panel.
The Five Verdict States #
| State | Color | What it means | What to do |
|---|---|---|---|
| Blocked | Red | The campaign cannot launch. Something is fundamentally broken — no products selected, no discount rule, deleted product catalog, etc. | Open the campaign and fix the blocking issue. The verdict tells you exactly which step to go to. |
| Risk | Orange | The campaign is running and actively losing money or causing problems — discount eating margin, stock depleted, severe overlap. | Review immediately. Pause if needed. |
| Caution | Yellow | A genuine quality issue needs review. Health is fair, a degraded warning is active, or a conflict is worth a look. | Review when you have time. Not urgent, but worth checking. |
| Good | Blue | Campaign is healthy and running within safe limits. | Nothing — keep an eye on it. |
| Ready | Green | All pre-launch checks pass. Safe to activate. | Click the verdict’s CTA to launch. |
Important: The verdict replaces the older 0–100 health score (Excellent / Good / Fair / Poor / Critical) as your day-to-day signal. The numeric health score still exists internally and on the dashboard, but it now describes store-wide health, not per-campaign decision-making (see Store-Wide Health).
What Intelligence Looks At #
Every time you open a campaign page, Intelligence collects four kinds of signals:
| Signal Type | What it covers |
|---|---|
| Planner | What the campaign is set up to do — products, discount type, schedule, priority |
| Health | What’s wrong right now — conflicts, stock issues, schedule overlaps, missing fields |
| Store | What’s happening store-wide — how many active campaigns, catalog coverage, seasonal trends |
| Economics | (Pro) Real order data — revenue attributed to the campaign, discount spend, break-even threshold |
From those signals it computes:
- Four scores (0–100): composite quality, opportunity upside, launch readiness, and downside risk
- Five operational metrics: margin impact, inventory burn, overlap severity, predicted uplift, timing quality
- Confidence (low / medium / high): how sure Intelligence is, based on signal completeness, data freshness, consistency, and evidence quality
- Reliability: whether the data behind the verdict is fresh or stale, and whether evidence is limited (e.g., a brand-new campaign with no orders yet)
- Top 2–3 actions ranked by severity, each with a CTA button that takes you to the right place to fix the issue
Reason Codes — The “Why” #
Every verdict is backed by a reason code that explains why Intelligence reached its decision. Reason codes are grouped into families so the message stays plain-English:
| Family | What it detects | Examples you might see |
|---|---|---|
| Conflicts | Two or more campaigns overlapping on the same products | “Two campaigns share products at the same priority”, “Blocked by a higher-priority campaign”, “Reduce overlap to free up sales” |
| Stock | Inventory problems on discounted products | “Some products are low on stock”, “All products are out of stock”, “Stock will likely deplete before the campaign ends” |
| Coverage | Product selection issues | “No products match”, “Selected products were deleted”, “Category filter returns nothing”, “Selection is too narrow” |
| Discount | Discount configuration problems | “No discount rule set”, “Discount eats too much margin”, “90%+ discount detected”, “BOGO is missing required configuration” |
| Schedule | Timing issues | “No schedule set”, “End date already passed”, “Recurring campaign has a large idle gap”, “Recurring extends beyond 6 months” |
| Reliability | Data-quality concerns | “Data is stale”, “Estimated values until activation”, “Too few orders to draw firm conclusions” |
| Review | Setup completeness and live-ops guidance | “Ready to launch”, “Health is fair — review”, “Campaign is performing well”, “Usage limit exhausted” |
Each reason code carries up to seven pieces of plain-English copy: what happened, why it matters, what to do, why that helps, and the button text for the CTA.
Where You See the Verdict #
The same verdict appears on every surface, formatted for context:
| Surface | Where | What it shows |
|---|---|---|
| Wizard (Step 5) | Campaign creation/editing | Critical issues with affected products, inline guidance, operational signals — used to validate before launching |
| Campaigns List | Campaigns table | Verdict badge per row, primary reason, CTA link for campaigns needing attention |
| Overview Panel | Slide-out panel from the campaigns list | Verdict badge, reason, warning detail, CTA button, confidence indicator |
| Dashboard — Verdict Row | Top of the main dashboard | Focus campaign verdict, recommended action with CTA, operational signals, economic signals (Pro), promotion timeline |
| Dashboard — Campaigns Strip | Below the verdict row | One verdict badge per active/scheduled campaign, sorted by urgency (Blocked first), capped at 5. Hidden when everything is healthy. |
All five surfaces consume the same verdict. None of them invents its own opinion — so what you see in the wizard matches what you see on the dashboard for the same campaign.
Store-Wide Health vs. Per-Campaign Verdict #
The dashboard shows two different things and they should not be confused:
| Store-Wide Health Score | Per-Campaign Verdict | |
|---|---|---|
| What it is | A 0–100 number describing the overall health of your campaign portfolio | A state (Blocked / Risk / Caution / Good / Ready) for one specific campaign |
| Where it shows | Dashboard health gauge | Wizard, list, overview panel, dashboard verdict row, dashboard campaigns strip |
| Use it for | “Is my promotional strategy in good shape overall?” | “What should I do with this campaign right now?” |
How store-wide health is calculated: active campaigns are weighted twice as heavily as scheduled ones. The weakest active campaign pulls the average down. Blocking issues hard-cap the score at 40, and degraded warnings apply a soft penalty.
The dashboard health gauge no longer tracks a single campaign’s readiness — it answers a portfolio-level question.
Confidence and Reliability #
Intelligence is honest about how sure it is. Two pieces of context surface alongside every verdict:
Confidence #
A score that lands in one of three bands:
- High: Lots of fresh signals, consistent data, real order evidence
- Medium: Enough data to act on, but with some gaps
- Low: Limited evidence — typically new or scheduled campaigns with no order history yet
Confidence is built from four weighted factors: signal completeness (24%), data freshness (22%), cross-signal consistency (18%), and evidence quality (28%).
Reliability #
Tells you whether the data behind the verdict is trustworthy right now. You may see one of these chips on the dashboard or overview panel:
- Stale: Data hasn’t been refreshed recently — verdict may be based on outdated signals
- Estimated: Some figures are projections (e.g., scheduled campaign that hasn’t activated yet)
- Limited evidence: Not enough orders attributed yet to draw firm conclusions
When evidence is limited and the structural severity would be “critical”, Intelligence intentionally softens the display severity to “warn” — it doesn’t sound the alarm based on data it isn’t confident in.
Recommended Actions and CTAs #
Every verdict that needs your attention comes with a recommended action and a CTA button that takes you to exactly the right place to fix the issue:
| Issue Type | CTA Goes To |
|---|---|
| Priority conflicts | Wizard Step 1 (Basic) |
| Product / coverage / stock issues | Wizard Step 2 (Products) — or directly to the WooCommerce product list when stock needs fixing |
| Discount configuration problems | Wizard Step 3 (Discounts) |
| Schedule issues | Wizard Step 4 (Schedule) |
| Review / monitor / launch | Wizard Step 5 (Review) |
If the verdict is Ready, the CTA is “Activate” — a single click that launches the campaign, with a confirmation dialog and guardrail check.
Operational Signals #
Beyond the verdict, Intelligence surfaces five operational metrics on the wizard, dashboard verdict row, and overview panel:
| Metric | What it tells you |
|---|---|
| Margin impact | How much the discount eats into profit |
| Inventory burn | How fast stock is depleting under the campaign |
| Overlap severity | How much conflict exists with other campaigns |
| Predicted uplift | Expected revenue increase from running the discount |
| Timing quality | Whether the schedule is well-positioned (seasonality, competition, store activity) |
These are inputs to the verdict — they’re shown so you can see why Intelligence reached its conclusion, not as standalone numbers to optimize.
Custom Guardrails #
You can override three Intelligence boundaries from Settings → Guardrails:
- Maximum discount load percent (default 22%, range 1–95%) — share of catalog revenue that can be discounted at once
- Maximum override intensity percent (default 40%, range 1–100%) — how aggressively higher-priority campaigns can override others
- Minimum stability score (default 70, range 10–95) — promotion stability floor
When you explicitly override a guardrail warning, Intelligence records the event so it can tune itself over time. There is no penalty for overriding — Intelligence advises, it does not punish.
How Intelligence Decides the Verdict #
The decision logic, in plain English:
Default verdict is GOOD. Is there a hard block or blocking critical issue? → BLOCKED Is severity critical, or readiness below 50? → RISK Is severity warn, or any warning active, or health is fair (readiness 70–89)? → CAUTION Is readiness ≥ 90 with no critical issues and no warnings? → READY (or GOOD if just-activated with low confidence) Otherwise: → GOOD
Data-quality signals (stale, low evidence) do not directly upgrade or downgrade the verdict. They surface through the reliability chip and confidence band so you can weigh the verdict appropriately.
Caching #
To keep page loads fast, Intelligence caches each verdict for a short window:
| Surface | Cache window |
|---|---|
| Wizard | 90 seconds |
| Dashboard | 120 seconds |
| Overview panel | 180 seconds |
Caches are flushed automatically when you save, activate, pause, or delete a campaign — so you’ll always see a fresh verdict after taking action.
What Intelligence Does Not Do #
- It is not Analytics. Intelligence answers “what should I do now?” Analytics answers “what happened?” Revenue, conversions, ROI, and historical charts live on the Analytics page.
- It does not auto-fix campaigns. It surfaces the issue and the recommended action, but you decide whether to apply it.
- It does not punish you for overriding guardrails. Override telemetry is collected to tune the engine, not to lower your scores.
- It does not score per-campaign on a 0–100 scale anymore. The 0–100 number you see on the dashboard is store-wide health, not the verdict.
Quick Reference #
| Question | Where to look |
|---|---|
| Is this campaign safe to launch? | Verdict in the wizard Step 5 or overview panel |
| Which of my campaigns need attention? | Dashboard campaigns strip, or the campaigns list verdict column |
| Why did Intelligence reach this verdict? | Reason text under the verdict badge — and the operational signals row |
| How sure is Intelligence? | Confidence band and reliability chip on the verdict |
| What should I do about it? | The CTA button on the verdict — it takes you to the right wizard step |
| How is my overall promotional strategy? | Store-wide health gauge on the dashboard |
| How did past campaigns perform? | Analytics page — not Intelligence |
