Smart, Sharp, Stuck — Diagnosing a Stalled Data-Analytics Pod

A four-round, intermediate professional-services simulation set inside Selat Analytics Sdn Bhd, a Kuala Lumpur data-and-strategy consultancy (RM 48M fee revenue, 130 consultants). It is 12 March 2026: a flagship RM 1.4 million, eight-week engagement for Bumi Telekom is three weeks from its 9 April board readout and visibly stalled. The churn-prediction model is strong — 0.84 AUC against a 0.75 target — yet the client steering committee has rejected two readouts as “technically impressive, practically useless.” The pod has consumed 62% of budgeted hours for ~40% of client value, the margin is sliding toward 9% versus a 35% target with an RM 180k projected overrun, team cohesion sits at 54/100, and one star data scientist is a flight risk. Playing the newly assigned Project Lead, you must diagnose the stall through Loredana Padurean's Smart (human-collaboration) vs Sharp (analytical/technical) skills lens, re-role the pod to the gap rather than the titles, choose a learning intervention under a hard deadline, and deliver a decision-led readout. Round 1 — diagnose: is the dominant blocker smart, sharp or both? The firm's instinct and hiring strength is sharp, but the real gaps are smart — no single client owner, conflict-avoidant consultants who won't push back on a shifting-requirements sponsor, and a model nobody translated into telco decisions. Round 2 — plan: name a single accountable owner of the Bumi relationship, decide whether the perfectionist data scientists keep tuning or are frozen and redirected to the missing cost-benefit sizing, and pick a learning intervention (action learning through the live readout prep, a coaching course, a heavy three-day offsite, or none). Round 3 — decide the four unblocking moves: skill emphasis (smart vs sharp effort), the conflict move, effort-vs-fit (weekend work or redeploy to the binding constraint), and the intervention dose. The math rewards skill-fit over raw effort and punishes over-investing in the already-abundant sharp work. Round 4 — recover and deliver: hold the diagnosis under a final-week sponsor demand for one more model variant, then deliver a readout that translates the 0.84-AUC model into concrete decisions Bumi can act on. Sticky penalty flags model the five classic errors — sharp-tunnel vision, effort over fit, an ownerless client relationship, conflict avoidance, and data-dump readouts — so that the headline trap of treating a human-collaboration stall as a technical one cannot reach the top verdict, the third readout fails, and the flagship (plus RM 3–4M of follow-on pipeline) is lost. Final KPIs track project progress (%), skill-fit score (/100), team cohesion (/100) and engagement margin (%).