Executive Pulse: product approach
A single-page decision cockpit for the leadership of a multi-country, multi-brand, multi-channel consumer goods company. It is built around one sequence: business health → deviation → drivers → implications → action.
The problem
CPG leaders get dozens of reports: sales cubes, trade promotion management (TPM) exports, syndicated share decks and finance packs. Each answers a different question at a different grain and on a different refresh cycle. Executives spend their review time reconciling numbers instead of deciding. The hard part is not showing data. It is choosing what deserves attention, linking a financial outcome to the driver that caused it, and separating "us" from "the market", all on one page.
Executive users and the decisions they make
| Role | Primary question | Typical decision | Lens defaults |
|---|---|---|---|
| CEO | Are we on track? What are the 3 issues that move the year? | Where to intervene; which leader owns it | NR, CM, share, FY outlook |
| CFO | Are we growing profitably? Where is margin leaking? | Re-forecast, cap trade spend, approve price moves | CM bridge, GM%, trade % of gross sales |
| CMO / Brand | Are we winning vs competitors? Is A&P paying back? | Reallocate A&P, respond to competitor moves | Share, growth vs market, ROMI, promo ROI |
| Chief Sales Officer | Which channels, customers and markets are short of target? | Redeploy field effort, fix distribution, renegotiate terms | NR and volume vs target, outlook |
| COO / Supply chain | Are we losing sales to availability? Is cost-to-serve rising? | Expedite, reallocate stock, change service levels | Fill rate, lost sales, cost per case |
| Procurement | How is input inflation hitting margin vs budget? | Hedge, re-source, trigger price or pack changes | COGS per case, GM% |
| Revenue growth mgmt (RGM) | Are price, promo and mix delivering the plan? | Re-design promo calendar, price-pack architecture | Price realization, promo depth and ROI, mix |
The role lens changes emphasis: which KPIs appear, how exceptions rank and which cards are marked "your focus". It never changes the numbers. Every role sees one version of the truth.
Metric prioritization
- Tier 1, outcomes (always visible): net revenue, contribution margin, FY outlook, market share. These are what the board asks about.
- Tier 2, drivers (one click away): volume, price, mix, trade spend, COGS, cost-to-serve, A&P. These explain Tier 1 through an additive bridge.
- Tier 3, diagnostics (inside the drill drawer): SKU, city, channel, promotion event, fill rate, competitor signal.
Ranking rule: an item reaches the front page only if it is material (it moves the $ outcome), actionable (it has an owner) and unexpected (it deviates from plan or from the market).
Dashboard structure: five bands, one decision sequence
- Health: an AI briefing (status, what changed, why, outlook) plus six role-specific KPI tiles, each shown against target, last year and recent trend.
- Deviation: a weekly or cumulative trend vs target and LY with the forecast to year-end, and a variance bridge that splits the gap into volume, mix, price, trade, COGS, logistics and A&P. Beside them, a ranked exceptions feed.
- Drivers: a brand × country (or channel) performance map, and a growth-vs-margin quadrant that answers "are we buying growth?", with the RGM plan-vs-actual scorecard.
- Implications: our growth vs category growth (share gain or loss), marketing and promotion ROI, and external signals such as competitor moves, input costs, demand and supply.
- Action: recommended actions with an owner, value at stake and status, each traced back to the exception that produced it.
Interaction model
- Global context bar: role, period (last week, rolling 4 weeks, QTD, YTD, FY outlook), comparator (target or LY), geography (region → country), brand (category → brand) and channel. Every panel responds to the same context.
- Progressive disclosure: clicking any KPI, bridge bar, heat-map cell, quadrant bubble, exception, promotion or action opens a drill drawer with a diagnosis, mini-bridge, trend and breakdown by SKU, city, channel or country. "Focus dashboard on this" pushes that slice into the global filters, so the whole page re-cuts without navigating away.
- Exception-led: the default view shows the most material deviations, risks and opportunities, not every metric.
Where AI earns its place
- Materiality ranking and anomaly detection: scores every brand × market × channel slice by $ gap, weighted to the role, and suppresses noise.
- Driver attribution: an additive volume / mix / price / promo / cost decomposition, so every "why" ties back to dollars.
- Causal hypotheses: joins internal drivers with external signals (competitor price index, market growth, fill rate, commodity costs) to propose the likely cause, each with its evidence.
- Narrative generation: the executive briefing and the drawer diagnosis, where every sentence cites a number on the page.
- Forecasting: a run-rate-adjusted FY outlook in the prototype; production would use ML forecasting with promo and seasonality features.
- Marketing mix model (MMM) and post-event promo evaluation: ROMI and promotion ROI (incremental vs baseline) feed the "repeat, optimise or stop" calls.
- Next-best action: rule and model-generated recommendations with owners. Closed actions feed back to improve ranking.
- Later: conversational Q&A ("Why is Germany behind?") grounded on the same semantic layer.
Guardrails: AI text is labeled, shows its evidence and is never the only source of a number. Actions have a human owner, and confidence is shown where models estimate (ROMI, forecast).
Data foundation and governance
Sources: ERP (shipments, COGS), TPM system (trade spend, promo calendar), distributor management system (DMS) and retailer point-of-sale data (secondary sales), syndicated retail audit (market size, share, competitor pricing), MMM outputs, supply-chain planning (fill rate, OOS), and plan/budget (targets, RGM plan). They are unified in one semantic layer with governed metric definitions. Access is controlled through role-based access control (RBAC) with SSO and MFA, row-level security (for example country managers see only their markets), audit logs on exports, and refresh cadence shown on screen: weekly sell-in, weekly or monthly syndicated data.
Success measures
- Adoption: % of the leadership team active weekly; share of business reviews run from the Pulse instead of slide packs.
- Time to insight: minutes from opening to identifying the top 3 issues (target under 5).
- Decision velocity: days from a deviation appearing to an owned action; % of actions closed on time.
- Business impact: trade spend ROI, forecast accuracy (FY outlook error), share of gap recovered after action.
- Trust: number of "my number is different" disputes; count of legacy reports retired.
About this prototype
Kestrel Consumer Group, its brands, competitors, events and every figure are synthetic. They are generated in the browser from a seeded model covering 6 brands, 3 categories, 9 countries, 7 channels and 52 weeks, with planted business stories: a competitor price cut, promo-funded growth, commodity inflation, a supply disruption, a declining category, an e-commerce breakout and general-trade expansion. Narratives and actions are rule-generated from that data to simulate the AI layer.