This is a static sample report. The chat prompts shown below demonstrate how you can question the analysis in your connected ChatGPT or Claude conversation; they are not interactive on this page.
Transparent MMM · Decision report
Media effectiveness and next actions
The report shows the model’s estimates, your original assumptions, and the uncertainty around each conclusion. It is designed to be interrogated, not simply accepted.
Ask Me Anything about this report
Every conclusion has an evidence trail: observed data, your starting assumption, and model uncertainty. You can probe any number or recommendation. I’ll explain what supports it—and say plainly when the evidence is weak.
Ask in chat about a channel, assumption, interval, recommendation, or budget…
WIP — Please ask in chat directly.
What is marginal ROI?Which estimates still rely on my prior?What should I test first?Continue in chat. Answers are limited to evidence available from the fit and its derived analyses.
Channel attribution, ROI, and uncertainty
Prior ROI is the initial assumption.
Posterior ROI is the result after the model used the data.
When they are close, the data mostly agreed with that channel’s starting assumption. When they are far apart, the data pushed back.
| Channel | Spend share | Share of total KPI attributed | Prior ROI Median (P50), 90% CI (Initial Assumption) |
Posterior ROI
Median (P50), 90% CI (After MMM Fit) |
Confidence |
|---|---|---|---|---|---|
| All Paid Media | 100.0% | 31.1% | 3.72x 2.9–4.9x | ||
| Facebook Performance | 24.6% | 7.8% | 6.12x 3.6–10.3x | 2.75x 0.9–9.2x | Low |
| Search | 16.1% | 6.1% | 6.12x 3.6–10.3x | 3.37x 1.0–10.2x | Low |
| Youtube | 14.7% | 4.4% | 6.12x 3.6–10.3x | 2.69x 0.8–7.7x | Low |
| 14.2% | 5.6% | 6.12x 3.6–10.3x | 3.70x 1.1–10.3x | Low | |
| TV | 10.7% | 0.5% | 6.12x 3.6–10.3x | 0.54x 0.2–1.0x | Low |
| Display | 9.9% | 5.4% | 6.12x 3.6–10.3x | 5.66x 1.9–12.6x | Low |
| Facebook (brand) | 9.8% | 1.2% | 6.12x 3.6–10.3x | 1.42x 0.9–2.4x | Medium |
Posterior ROI estimates and 90% credible intervals
Each dot is the posterior median. Each horizontal line is its 90% credible interval. Wider lines mean the model cannot pin the channel down as precisely.
Prior vs Posterior
What we assumed, and what the data concluded
TV — started from 3.6–10.3x, the data moved it to 0.2–1.0x. This is not a small adjustment: the two ranges don’t overlap, so the data rejected the starting assumption for this channel.
Facebook (brand) — started from 3.6–10.3x, the data moved it to 0.9–2.4x. This is not a small adjustment: the two ranges don’t overlap, so the data rejected the starting assumption for this channel.
5 of 7 channels (Facebook Performance, Search, Youtube, Instagram, Display) stayed close to the starting assumption and remain imprecise. Their numbers reflect your starting belief more than a finding from the data — ask me before treating them as settled.
Budget Reallocation Plan
Not more than 20% increase or 50% drop in any channel.
Budget moves from lower- to higher-return channels until their next-dollar returns equalize. Total spend is unchanged.
RECOMMENDED NEXT MOVE
Reallocate cautiously
Treat the proposed moves as testable decisions and validate low-confidence changes first.
MODELED KPI_REVENUE CHANGE
₹+823,555,497
Versus the current allocation; not a guarantee.
PORTFOLIO ROI
3.72x
90% credible interval: 2.9–4.9x
| Channel | Current spend | Recommended spend | Change | Marginal ROI | Confidence | Recommended action |
|---|---|---|---|---|---|---|
| All Paid Media | ₹3.65B | ₹3.65B | +0% | Not blended | — | Same total budget |
| Facebook Performance | ₹897.9M | ₹931.8M | +4% | 2.49x | Low | Test before scaling |
| Search | ₹587.6M | ₹705.2M | +20% | 2.95x | Low | Test before scaling |
| ₹518.3M | ₹622.0M | +20% | 3.13x | Low | Test before scaling | |
| Display | ₹361.4M | ₹433.6M | +20% | 4.34x | Low | Test before scaling |
| Youtube | ₹536.5M | ₹583.5M | +9% | 2.50x | Low | Test before scaling |
| Facebook (brand) | ₹357.7M | ₹178.8M | -50% | 1.19x | Medium | Reduce modestly; monitor |
| TV | ₹390.6M | ₹195.3M | -50% | 0.56x | Low | Validate before reducing |
Marginal ROI is the modeled return on the next unit of spend, not the historical average return. For auction-based digital channels, complement this plan with live signals such as CPM, CTR, and frequency. Marginal ROI is very sensitive to audience saturation and competitive pressure.
Recommended Experiments
Above-even share of paid spend, but the model can’t pin these down. Listed biggest spend share first.
| Channel | Spend share | Why it needs testing |
|---|---|---|
| Facebook Performance | 24.6% | Its range is too wide to act on and the estimate still leans on your prior. |
| Search | 16.1% | Its range is too wide to act on and the estimate still leans on your prior. |
| Youtube | 14.7% | Its range is too wide to act on and the estimate still leans on your prior. |
Ask me in chat to discuss experiment design for any of these — geo-lift, conversion lift, or another approach depends on the channel.