Executive cockpit KPIs: what leadership should see every week

Why most dashboards fail

In many SMBs, every leader arrives at the committee with their own version of the truth: an ERP extract, a spreadsheet, sometimes a CRM slice. Numbers disagree; the meeting becomes an argument about sources instead of a decision. Adding “one more report” or “one more web page” without data foundations solves nothing: you multiply divergent views.

A KPI cockpit works when it sits on a recognized layer (warehouse or marts), a limited set of metrics tied to real levers, and an explicit committee agreement on definitions (margin, backlog, AR, cost of goods). Without that social contract, the prettiest tool is ignored.

At Consultation Génosa, the cockpit is part of executive decision support: it exists to surface gaps early enough to fix the quarter — not to impress.

KPI families that actually matter

There is no universal list, but manufacturers, distributors, and B2B services firms often converge on four families. Finance: gross margin, contribution by customer or project, budget variance. Cash: expected collections, customer delays (DSO), critical payables. Operations: service level, delays, scrap or AP-exception rates. Commercial: backlog, qualified pipeline, margin mix — not volume alone.

Pragmatic rule: if a metric cannot trigger a committee action within seven days, it probably does not belong on the primary cockpit. Park it in a specialist view.

• Finance: margin, customer / project P&L, budget variance
• Cash: 4-week forecast, at-risk AR, supplier due dates
• Ops: lead times, quality, exception queues (AP, support, production)
• Commercial: backlog and margin mix, not only revenue

Thresholds, alerts, and the role of AI

A KPI without a threshold is decoration. Define green / amber / red bands with leadership — for example customer margin under 18% for three weeks, or AR over 60 days above a ceiling. AI wired to your data can then watch those thresholds continuously and prepare the “why” (four orders driving it, a customer stretching payments) without replacing human judgment.

ChatGPT, Claude, Gemini, or Grok become useful when they have eyes and hands inside your perimeter — through the Génosa connection — rather than inventing from the public web. Data stays local or on your private network under the chosen frame.

Leadership cadence: weekly before monthly

The monthly pack arrives too late to save a quarter. Install a 30–45 minute weekly review focused on cockpit exceptions, not reading every line. Once a month, go deep; every week, clear exceptions.

Génosa’s free AI diagnostic exists to pick the 2–3 views worth investing in before you build. Then come data foundations, the cockpit, and signals.

Traps to avoid

Avoid vanity dashboards (web traffic, likes) when your real issue is margin and cash. Avoid “real time at all costs” if data quality cannot keep up: a reliable daily refresh beats a false instant stream. And do not confuse email automation with decision support — different lanes.

Agreeing definitions with the committee

Before writing any code, run a short workshop (90 minutes) with the GM, finance, and ops. For each candidate KPI, write in one sentence: formula, system source, owner, amber threshold, red threshold, and typical action if breached. If you cannot fill those six fields, the KPI is not ready for the primary cockpit.

Also document exclusions: returns, credit notes, internal projects, intercompany. Those are what make spreadsheets diverge. Once the committee agrees — even informally by email — you have a mandate to build. Without it, every BI sprint becomes a political negotiation.

In Montréal and across Québec, SMBs that succeed treat this step as a leadership deliverable, not a technical detail left to analysts alone.

Data foundations under the cockpit

The cockpit reads a truth; it does not create it. Warehouse or marts, quality rules, refresh timestamps, and a clear link to the ERP (Dynamics 365 Business Central, QuickBooks, Sage 50, Acomba, Xero, Zoho…) are the condition for lasting trust. A “mostly right” metric destroys credibility faster than a missing metric.

Plan refresh to match the ritual: daily for AR and cash, weekly for some production variances, monthly for complex allocations. Threshold-watching AI inherits that cadence: no point alerting in real time on data frozen since yesterday.

Consultation Génosa deliberately sequences diagnostic → foundations → cockpit → AI signals. Reversing the order (chatbot first) yields seductive demos and fragile decisions.

Measure adoption, not only go-live

A cockpit nobody opens does not exist. Track simple proxies: attendance at the weekly review, exceptions cleared, time from alert to decision. After six to eight weeks, remove tiles never discussed. Add only what the committee requests with an action in mind.

Related guides — customer-project P&L, cash & AR, super AI assistant — deepen views that often feed the same cockpit. The flagship remains executive decision support; KPIs are its weekly face.

Frequently asked questions