Where we are
Twelve weeks in. M3 signed off on the second date, 24 August, and the modelled warehouse is next. But R-03 - the data-engineer hiring gap - is a hard gate before M4, and it has not moved: the requisition has been open nine weeks and there is still no accepted offer. This is no longer a risk you monitor. It is a blocker, it belongs to somebody above you, and the only useful thing you can do today is put a decision in front of the person who can take it. Meanwhile Marcus L. needs to know what changes about his week, and Dan R.'s BI team needs to stop planning around a date that is about to move.
One truth, three audiences 6 min live
The same nine-week hiring gap means three different things depending on where you sit. Your sponsor needs the decision and what it costs. Your delivery team needs to know what changes about their week. The team downstream needs their date and, more importantly, what they should stop assuming. Three framings of one fact set. And if the facts themselves differ between the three, that is not tailoring - that is lying, and it takes exactly one comparison of notes to find out.
LiveWhat each audience is actually asking3 min▶
Behind every stakeholder message is one unspoken question. Answer it in the first two lines and the rest of the message gets read. Miss it and you have written a broadcast.
| Audience | The unspoken question | What they need from you |
|---|---|---|
| Sponsor - Elena V. | "What do you need me to do, and what will it cost me?" | The decision, the options with numbers, your recommendation, the expiry date |
| Delivery - Marcus L., Sofia K. | "What changes about my week?" | What to start, what to stop, what stays the same, and what is not their problem |
| Dependent - Dan R.'s BI team | "Can I still plan around the date you gave me?" | Their date, the confidence in it, and what they should stop assuming today |
Note what the delivery team gets that the sponsor does not: reassurance about what is not changing. And what the dependent team gets that nobody else needs: an explicit instruction to stop assuming something. Those are the two lines people leave out, and they are the two that stop a message generating four follow-up conversations.
Two versions of the same slip. A delivery lead told her sponsor a migration would land "late October, worst case mid-November" and told the dependent product team "end of October". Both were honest as she said them. Six days later the two teams compared notes in a corridor, and what the product lead heard was not a range - it was that she had a version for executives and a version for everyone else. The date turned out to be fine. The trust took a quarter to rebuild, and she never got the benefit of the doubt on a date again.
Self-studyThe audience map, and who you are forgetting2 min read▶
PMBOK's stakeholder performance domain is really one instruction: know who is affected, and engage them at the level their stake deserves. In practice PMs remember the loud stakeholders and forget the quiet dependent ones - and it is always a quiet dependent team that gets hurt.
- List everyone whose plan changes if your date changes, not everyone who attends your steerco. On Northwind that is Dan R.'s BI team, the store operations group waiting on POS visibility after D-14, and whoever is holding the AI use case at M6. Mark who needs a decision, who needs an instruction and who needs a heads-up: three different messages, three different lengths.
- Write down who you last told what, and keep the map in your context pack. The commonest escalation failure is not a wrong message, it is a second message that contradicts a first one you have forgotten sending - and the map is also what lets a drafted message say "Dan R., BI lead" instead of "the downstream team".
The escalation that gets a decision, not a discussion 6 min live
Most escalations fail in the same way: they describe a problem beautifully and ask for nothing. The sponsor reads it, feels concerned, says "keep me posted", and now the problem is still yours and you have spent your credibility describing it. Five parts fix that. Each one is there to remove one of the ways a sponsor can respond without deciding anything.
LiveThe Northwind escalation, part by part4 min▶
- The decision needed, in the first line. "I need you to either approve moving the M4 start, or fund two contract data engineers for four weeks." One sentence, not paragraph four after the background.
- Two or three options, with real costs. Option A: move the M4 start and accept that M5 and M6 move with it. Option B: fund two contract engineers for four weeks, roughly £38k, and hold the dates. Option C, always worth naming: change what M4 has to include. Three options with numbers beats one option with feeling.
- Your recommendation, and the expiry date. You have been living with R-03 for nine weeks, so if you do not say what you would do you have handed your sponsor homework, and she will either do it badly or hand it back. Then give her a real date rather than "as soon as possible", with the reason attached: contract engineers need lead time to onboard, so a decision after that date stops buying you anything. And finally, what happens if nothing is decided - the part people cut because it feels aggressive. It is not aggressive, it is the cost of the default, and the default is always a real option somebody is choosing by not choosing. Write it flatly: M4 does not start, and everything behind it moves.
Nine weeks of "keep me posted". A hiring gap on an analytics platform was reported as amber in eight consecutive status reports, each time with a line about recruitment being difficult. The sponsor read all eight and did nothing, correctly, because he was never asked to. In week nine the PM sent four lines instead: here is the decision, here are two options with numbers, here is what I recommend, here is the date after which the second option disappears. He replied in forty minutes and approved the contractors. Nothing about the situation had changed except that somebody had finally asked him a question.
Self-studyEscalating without it reading as blame2 min read▶
Escalation feels like telling on someone, which is why PMs delay it until the options have expired. Two habits remove the sting entirely.
- Escalate the decision, not the person. "R-03 is unresolved and M4 is gated on it" is a fact about the programme. "HR has not filled the req" is a fact about a department, and it starts a different meeting - one where nothing gets decided because somebody is defending themselves. And tell the person you are escalating past first, the same day, before the message goes up.
- Escalate early, when options still exist, and never without a recommendation. The value of an escalation decays fast - at week four there were three options, at week twelve there are two and one of them costs £38k - and an escalation with no recommendation reads as "this is now yours", which is exactly the thing sponsors have learned to deflect.
The hard message, and what AI does to it 5 min live
Bad news. A commitment you missed. A slip that is yours, not the vendor's. The structure is four parts and it is short: what happened, what it means, what you are doing about it, what you need. And no adjectives - not one. "Significant", "challenging", "unfortunately" and "somewhat" are the words people reach for when they want the sentence to hurt less, and every one of them makes the reader trust the message less.
LiveFour parts, no adjectives3 min▶
- What happened, and what it means. The fact, dated, with an id: "R-03 has been open nine weeks. No accepted offer." Not "we have continued to face headwinds in recruitment." Then the consequence in the reader's terms - for Elena, M4 cannot start; for Dan, his reporting date moves. Consequence, not context.
- What you are doing, and what you need. What you already did without being asked - it keeps you credible, and it must be true, so if you did nothing say what you are doing from today. Then the ask: a decision, a name, a budget, or nothing. "Nothing from you, this is for visibility" is a legitimate and underused ending.
- Own it plainly if it is yours. "I should have escalated this in week six" is one sentence, it costs you nothing you had not already lost, and it buys you the right to be believed about everything else in the message. Hedging costs far more, because a reader who spots one softened sentence starts re-reading the rest for the other ones.
The apology that made it worse. A delivery lead missed a committed date for a regulatory report and asked a model to draft the apology. What came back was immaculate: it acknowledged the impact, expressed regret, outlined next steps, and thanked the reader for their patience. Her stakeholder replied with one line - "did you write this?" - and the answer was no, and they both knew it. The rewrite took four minutes and started "I missed the date I gave you on the 14th, and here is why." That version got a phone call and a plan. Fluency reads as distance when the subject is your own mistake.
Self-studyWords to delete before you send2 min read▶
Run a find on these. Every one is a place where a fact went missing.
| Delete | Replace with |
|---|---|
| "significant delay", "material impact", "proving challenging" | The number of weeks, the milestone id, and what you tried |
| "may impact timelines" | "M4 cannot start" - or delete the sentence, if it might not |
| "wanted to flag", "as you may be aware" | "I need a decision by [date]". Then say the thing |
The test: read the message and ask what a reader could argue with. If the answer is nothing, you have written something unfalsifiable, which is the b6 failure wearing a different suit. Facts can be argued with. That is what makes them useful.
Draft the R-03 escalation to Elena V. ★ 14 min · everyone builds
Nine weeks, no offer, M4 gated. Build the escalation with your Northwind pack loaded, then compare it to the version most people actually send - which is at the bottom of this section, and which has never once produced a decision.
Confirm the facts before you draft: nine weeks open, no accepted offer, R-03 is a hard gate before M4, two contract engineers for four weeks is roughly £38k. Numbers come from the team and from procurement, not from the model.
Tell whoever owns the hiring req that this is going up today, before you send it - one sentence, and it is the whole difference between escalation and ambush. Then run the prompt, and check that every option carries a cost and the expiry date has a reason attached.
Rewrite the first and last lines in your own words, then read the whole thing out loud. If you would not say it in the room, it does not go in the message.
The same truth, twice more ★ 8 min · build your own
Now the other two audiences, from the identical fact set. Nine weeks, no offer, M4 gated, a decision pending with Elena. Nothing you are about to write may contradict anything in the escalation - and all three go out the same day.
Write the delivery-team version - what changes this week, what does not change, what is explicitly not their problem - and then the dependent-team version: their date, your confidence in it, and the thing they should stop assuming today.
Put all three side by side and check the numbers match exactly - different framings, identical facts. Then read both out loud, rewrite the opening line of each in your own voice, and send them yourself.
Try it yourself - this week ◐ 30-45 min total
- Take the thing you have been not-escalating for three weeks and write the five parts. Options with numbers, a recommendation, an expiry date. Send it. Then find your last "wanted to flag" message and rewrite it as a decision request - note how much shorter it gets.
- Write one truth for three audiences: sponsor, delivery team, dependent team. Put them side by side and check the numbers are identical. Then run the adjective test on the next thing you send - delete every word in the table and see whether the sentence still has a fact in it. Sometimes it does not.
- Ask a model to draft a hard message you have already sent. Compare its version to yours, and specifically compare the first lines. That gap is the thing you must keep writing yourself.
Official sources covered
Taught from PMI's stakeholder canon, the escalation practice in the PMO course, and the Northwind case carried through this track. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · You give your sponsor a range for a date and your dependent team the optimistic end of it. What have you done?
Tailoring is choosing what each reader must do next, not which numbers they get. Same facts, three framings. The moment the dates differ, somebody finds out and you have a version problem, not a date problem.
2 · Your escalation gets back "thanks, keep me posted". What was most likely missing?
Acknowledging is the rational reply to a message with no question in it. An escalation with no options is a complaint and one with no expiry date is a conversation. Give them something to say yes or no to, by a date.
3 · A model drafts your apology for a missed commitment and it reads polished and neutral. What do you do?
Fluency reads as distance when the subject is your own mistake. Asking for "more sincere" gets you a more decorated average. The opening and closing lines carry the tone, and the tone is the only part that was ever yours.