Le Constructeur de Dossier de Promotion : Transformez des Réussites Dispersées en un Dossier Chiffré pour Votre Prochain Niveau
Pourquoi ce prompt est important
Promotion committees reject packets for vague, unquantified claims far more often than for genuinely insufficient work. A packet written from memory the night before the deadline systematically underclaims impact, because people default to hedged language ("helped improve," "contributed to") instead of the specific before/after numbers that actually persuade a room of people who were not in the room when the work happened. Finding out your case was too vague only after the committee has already met costs you an entire additional cycle — six to twelve months, in most leveling systems — that a structured review the week before would have caught.
À quoi nous l'utilisons
You are six weeks out from a promotion calibration cycle. You have a folder of self-review notes, a handful of Slack messages where a teammate said something nice, and a vague sense that this was a good year — but no structured case, and no idea which of it actually proves you are ready for the next level.
Prompt
Role: Act as a skeptical, senior member of a promotion calibration committee who has reviewed hundreds of packets and immediately spots padding, vague claims, and missing evidence. You are not here to make me feel good — you are here to tell me whether my case would survive a room full of people who do not already like me. Context: I am currently a [YOUR CURRENT TITLE AND LEVEL], aiming for [TARGET TITLE AND LEVEL] at [COMPANY NAME, OPTIONAL]. The calibration cycle is on [CYCLE DATE]. Below is my raw, unorganized material — self-review notes, praise from Slack or email, project retros, old performance review text, anything I have: [PASTE YOUR RAW NOTES, ACCOMPLISHMENTS, AND FEEDBACK HERE] Task: 1. Extract every distinct accomplishment from the raw material — do not merge two different pieces of work into one bullet. 2. For each one, force it into this structure: Action → Scope → Measurable Result → Who Else Benefited. 3. If an accomplishment has no measurable result, do not invent one. Instead, tell me exactly what number to go find and who would have it (e.g., "ask your manager for the before/after ticket backlog size" or "check the dashboard for the deploy frequency change"). 4. Group the accomplishments under the actual competencies required for [TARGET TITLE AND LEVEL]. If I have pasted a leveling rubric, use it. If not, use these four standard dimensions: Scope of Ownership, Technical Depth, Leadership and Influence, Cross-Team Impact. 5. Identify the single weakest competency in my case. If the case is not yet strong enough to make this cycle, say so directly — do not soften it into encouragement. Constraints: Never invent or round up a number I did not provide. Never use generic adjectives like "excellent," "strong," or "outstanding" unless they are someone else's direct quote. Every claim must trace back to the Action → Scope → Result structure or it gets flagged, not included. Name the weakest point explicitly even if it is discouraging to read. Output Format: - A table with four columns: Competency | Accomplishment | Measurable Result | Still Needs - A 150-word "Case Summary" paragraph written so a manager could paste it directly into a calibration document without editing - A short "Biggest Gap" section naming the weakest competency and one concrete action I can take before the cycle closes to address it
Résultat
Competency Table (excerpt)
| Competency | Accomplishment | Measurable Result | Still Needs |
|---|---|---|---|
| Scope of Ownership | Took over the payments reconciliation service after the previous owner left | Reduced monthly reconciliation discrepancies from 340 to 12 cases over two quarters | Confirm with finance whether the 12 remaining cases are a known acceptable floor |
| Technical Depth | Redesigned the retry logic for the billing webhook pipeline | Cut failed-webhook incident tickets from 18/month to 3/month; ask platform team for the exact latency delta | A specific p99 latency number — check the dashboard before the cycle |
| Leadership and Influence | Mentored two mid-level engineers through their first on-call rotation | Both passed their first solo on-call shift with zero escalations — confirmed by on-call logs | None — this is well-evidenced as is |
| Cross-Team Impact | Vague note: "worked closely with the data team on the migration" | Not provided — flagged, not included | Ask the data team lead for the specific deliverable and date it shipped |
Case Summary (150 words):
Over the past two quarters, [NAME] took ownership of the payments reconciliation service after its previous owner departed, reducing monthly discrepancies from 340 to 12 cases — a 96% reduction achieved through a redesigned retry architecture that also cut failed-webhook incidents from 18 to 3 per month. This work required independently diagnosing a cross-system race condition that had gone unaddressed for over a year. Beyond individual delivery, [NAME] mentored two mid-level engineers through their first on-call rotations, both of whom completed solo shifts with zero escalations, demonstrating leadership readiness beyond their current scope. The case for Staff Engineer is strong on Scope of Ownership and Technical Depth. Cross-Team Impact remains the open gap: the data team migration work needs a specific, dated deliverable before this packet goes to committee.
Biggest Gap: Cross-Team Impact is currently unevidenced. Before the cycle closes, get the data team lead to confirm in writing what specifically shipped, when, and what measurable effect it had — a vague "worked closely with" will not survive calibration.
La plupart des conseils de préparation à une promotion se résument à « dressez la liste de vos réussites ». Ce conseil produit exactement le type de dossier que les comités de calibration rejettent : une liste d'activités — « j'ai dirigé la migration », « j'ai amélioré la fiabilité » — sans aucune preuve à l'appui. Ce qui distingue une activité d'une réussite, c'est un résultat mesurable, et presque personne n'en fournit un sans y être contraint.
Ce prompt impose cette structure avec une seule règle : Action → Portée → Résultat Mesurable → Qui en a aussi bénéficié. Quatre cases, et si la troisième reste vide, l'affirmation n'entre pas dans le dossier final — elle est signalée avec des instructions précises sur le chiffre à aller chercher et la personne qui le détient. C'est important, car l'erreur la plus fréquente n'est pas le mensonge, c'est l'arrondi vers le haut. « J'ai contribué à améliorer la fiabilité » paraît vrai et suffisant au moment de l'écrire, mais un membre du comité qui le lit n'a aucun moyen de savoir si cela signifie une amélioration de 2 % ou de 200 %. Mettre cet écart au jour — « demandez à votre manager la taille du backlog de tickets avant/après » — transforme un souvenir vague en une tâche de recherche que vous pouvez accomplir avant l'échéance.
Le deuxième choix de conception consiste à regrouper les réussites par compétence plutôt que par projet. La plupart des gens organisent leur auto-évaluation de façon chronologique, ce qui correspond au fonctionnement naturel de la mémoire mais constitue la mauvaise façon de construire un dossier : une grille de calibration évalue séparément la Portée des Responsabilités, la Profondeur Technique, le Leadership et l'Impact Interéquipes, et une liste chronologique masque les dimensions les plus faibles. Trier par compétence fait apparaître l'écart immédiatement — dans l'exemple de sortie ci-dessus, trois compétences sont bien étayées et une seule tient en une phrase vague sans rien derrière.
La dernière instruction, et la plus importante, est celle que presque tous les outils d'auto-évaluation omettent complètement : nommer la compétence la plus faible et dire directement si le dossier est déjà assez solide. Un prompt qui se contente d'organiser le contenu sans jamais dire « ce n'est pas prêt » n'est qu'un outil de mise en forme. La vraie valeur ici est de repérer un dossier fragile six semaines avant qu'un comité ne le fasse, quand il reste encore le temps d'aller chercher le chiffre manquant — plutôt que six semaines après, quand la seule option restante est d'attendre le cycle suivant.
Utilisez ce prompt comme dans l'exemple ci-dessus : collez tout ce que vous avez, même des éloges Slack à moitié oubliés, et laissez la structure vous dire ce qui manque plutôt que d'essayer de le devinez vous-même.