The Birbal loop
Listen. Decide. Execute. Learn. Repeat.
One brief, and every action from then on follows the same eleven steps: perception first, decision gated by policy and risk, execution verified and remembered.
“I am a Gurgaon luxury real estate developer. I want to generate qualified leads for a ₹5 crore property. My budget is ₹1,00,000/month.”
Listens To
The market, competitors, customers, audiences and the live account: read continuously, so the next decision starts from what is actually happening.
- 01Observe
- 02Understand
Nothing here is a one-off. The Learn step hands everything it found back to the next Listen, which audience converted, which hook died, which rupee earned the most. That accumulated account knowledge is the part a competitor cannot copy.
Next best action
What is the single highest-impact thing to do next?
Every candidate action is scored the same way, so the ranking is arguable rather than mysterious.
Scoring
One ranked list. Raise the risk and an action falls, however large its upside.
Explainability
Why did Birbal do this?
Every illustrative action carries the reason, evidence, risk and reversibility a decision-maker needs before approving it.
#1Shift ₹8,000 to Campaign AConf 91%+₹31,000 revenue
Why
Campaign A is returning more marginal revenue than the current donor campaign.
Evidence
- ROAS: 4.7×
- CAC: ↓ 18%
- Audience capacity: High
- Creative health: 92
- Action
- Shift ₹8,000 to Campaign A
- Risk
- Low
- Confidence
- 91%
- Reversible
- Yes
#2Replace fatigued Creative BConf 84%−12% CPA
Why
Performance has softened while reach stayed stable, which is consistent with creative fatigue.
Evidence
- CTR: ↓ 18%
- Frequency: 2.1
- Creative health: 61
- Last 7 days: declining
- Action
- Replace fatigued Creative B
- Risk
- Medium
- Confidence
- 84%
- Reversible
- Yes
#3Test broad audienceConf 72%+18% volume
Why
The current audience is showing capacity constraints while the broader segment has not been tested.
Evidence
- Audience capacity: Medium
- Frequency: 3.4
- Estimated reach: High
- Prior test: None
- Action
- Test broad audience
- Risk
- Medium
- Confidence
- 72%
- Reversible
- Yes
Illustrative ranking. Live figures come from the account’s own data.
Worked examples
What detection actually looks like.
Two problems are planted in the development fixture. Both have to be found from the numbers alone: nothing in the data is labelled.
Signal 01 · Creative fatigue
A creative wearing out
Impressions held flat at 10,000/day for all 90 days, so a falling click-through rate cannot be blamed on reach.
The drop is caught by a rolling median against the trailing window, not a fixed threshold, which would either miss it or fire constantly on ordinary noise.
Signal 02 · Budget misallocation
The winner nobody funds
Share of spend against return, across the fixture’s last thirty days. One campaign earns most per rupee and receives least.
- Campaign 1Spend32%ROAS2.15×
- Campaign 2Spend28%ROAS2.75×
- Campaign 3Starved winnerSpend10%ROAS4.30×
- Campaign 4Spend30%ROAS2.05×
Budget follows marginal return, within hard scaling limits, so the fix is a proposal with an expected impact attached, not a blind reallocation.
Figures from the seeded development fixture used to test detection, not from a customer account.
Don't take our word for it.
Give Birbal a goal.
We'll show you what Viralitea finds, what it recommends and what it can automate.