Where we start
You have the map: every ask tagged external, internal or both, and every data C written as a deliverable with a date. You have the front door: an intake template and a service menu. What is left is the conversation itself. The running company is still Kettle & Co, constructed, and every script below uses its asks and its modelled days from the triage desk. The words are meant to be borrowed and rewritten in your own voice; the order of the parts is the part to keep.
★ What you walk out with tonight. A four-part order for any pushback (deliverable, date, decision, boundary with its reason), six scripts you can print and adapt, a three-step escalation path that always arrives with an offer, and a short list of cases where saying yes is the right call.
Clear is kind; a vague yes is not
5 min live"We'll help with the market research" feels generous. It is the most expensive sentence a data team says. Nobody wrote down what help means, so the requester hears "data is doing the research", the analyst hears "I'll pull a few numbers", and three weeks later both are right and both are angry. A clear answer with a boundary costs a slightly harder minute now and saves the month.
LiveWhy "we'll help" grows3 min▶
An unbounded yes has no edge, so every follow-up lands inside it. "Could you also check the competitors in Japan?" is reasonable once data is "helping with the research", and so is "could you draft the survey?", and so is "could you present it to the board?". Each step is small. On the triage desk, the Japan ask costs 3 modelled hours as a consulted page and 40 if data takes all of it; nobody ever agreed to 40, they agreed to "help".
Resentment arrives from both sides. The analyst feels used, because the owned work is late and nobody seems to notice. The BD lead feels let down, because the research arrived late and was shaped by somebody who has never met a Japanese distributor. Neither of them is wrong, and both were set up by the first sentence.
Self-study"No" is not the opposite of "we'll help"2 min▶
The opposite of a vague yes is not a flat no; it is a specific yes with an edge. On the triage desk, the flat no ("say no to anything not ours") kept the hours low and left nine decisions on anecdote. The specific yes gives the requester something real, on a date, and keeps the part that is theirs with them. It is harder to say than either, which is why it is worth scripting.
The research project that became a quarter. An analyst agrees to "help with some market research" for a new region. Two months later she is running a survey, buying a panel and writing recommendations, and the dashboard rebuild she owns is a quarter late. Her manager asks why. She does not have an answer that does not sound like an accusation, because nothing was ever written down. Constructed example; the pattern is common.
Deliverable, date, decision, then the boundary
6 min liveEvery script below has the same four parts in the same order. What they get. When. Which decision it serves. Then the boundary, with the reason, which is almost always the same reason: they own the outcome, so they own the decision. The order matters because people hear the first sentence loudest. Open with the boundary and they hear "no"; open with the deliverable and they hear "yes, and here is its shape".
LiveThree scripts for the loudest complaints4 min▶
"Why isn't data doing our market research?" (BD lead, the Japan ask)
"Just write our UK supermarket penetration strategy." (BD lead)
"Why is data so slow?" (marketing lead)
Self-studyThree more: the tracker, the platform number, and "by tomorrow"4 min▶
"Can't data just own the brand tracker?" (brand manager)
"Meta says ROAS is 4.1, your number is wrong." (marketing lead)
"We need it by tomorrow." (anyone)
Every date in these scripts is a Kettle & Co working day from the triage desk, constructed. The structure is the part that travels.
The same boundary, two orders. "The strategy is yours, not ours. We can give you some numbers." and "By day 18 you get repeat, basket and price sensitivity by region, so your strategy starts from our evidence; the strategy stays with you." say the same thing. The first is heard as a refusal and gets escalated. The second is heard as a plan and gets a thank-you. Constructed example; try both on a colleague.
The escalation ladder, and the cases for yes
4 min liveSome pushbacks do not land, and some should not be made. When a script does not settle it, the ladder has three rungs, and every rung arrives with an offer, never only with a complaint. And some asks deserve a plain yes, even when they are inconvenient.
LiveHow to climb a rung without it becoming a fight3 min▶
Rung 1, analyst and requester. The script. Most asks end here, because the requester gets something real on a date.
Rung 2, both leads. When the requester needs more than the menu offers, the data lead and the requester's lead look at the queue together. The data lead arrives with one concrete offer: "we can do the Japan page and the sizing half this month if the dashboard moves to the first week of next month". The point is to make the trade visible, not to win it.
Rung 3, both heads. A shared priority call, in writing, when two fixed dates collide or capacity drops. The triage desk's break button is this rung's case: with one analyst on leave, the right split starves six data-owned asks, and the fix is a re-planned month agreed by both heads before the old dates pass, not a quiet slide into refusing the split asks.
Self-studyWhen saying no is the wrong answer2 min▶
Internal asks are data's job, and "we're busy" is not a boundary, it is a capacity conversation (rung 2). The evidence half of a split ask is data's job too: refusing it is the say-no rung on the desk, which looked efficient and left nine decisions on anecdote. A real emergency (a broken checkout, a campaign sending the wrong price) gets a yes, with what moved said out loud. And when only data can do something, a holdout read, a join of queries to orders, the answer is yes even when the timing is bad, because the alternative is that nobody does it.
The boundary in this whole track is about ownership of the decision, not about effort. Data can work very hard for marketing and still never own the marketing decision.
The escalation that arrived with an offer. Two fixed dates collide: Black Friday readouts and the budget round. Instead of telling the marketing head "we can't do both", the data lead brings one page: the queue, the two dates, and two options, each saying what would move. The marketing head picks one in five minutes. The same conversation without the page takes an hour and ends with both heads annoyed. Constructed example; the page is the point.
Four scripts of your own
★ 22 min · everyone works it throughNo code. Pick the four complaints you hear most, and write one script for each in the four-part order. Then read each one out loud to the person next to you, who plays the requester and says what they heard first. Rewrite until they hear the deliverable before the boundary. Your scripts will differ from the Kettle & Co ones; keep the order.
Step 1 · List your four complaints. In the requester's words, not yours. If you are stuck, start from the six above.
Step 2 · Tag each one. External, internal or both, from Partnership session 2. An internal complaint ("why is data so slow?") gets a capacity script, not a boundary script.
Step 3 · Write the deliverable and date first. Use your service menu from Partnership session 3 for the turnaround. If there is no deliverable at all, the ask is a redirect: say who owns it and offer an introduction.
Step 4 · Name their decision. The "so you can" line. If you cannot name it, go back to the four questions from Angles session 1.
Step 5 · Write the boundary with its reason. One clause. The reason is nearly always "because you own the outcome". Avoid "it's not our job": it is accurate and lands as a door closing.
Step 6 · Read it out loud. Your partner says what they heard first. If it was the boundary, reorder. Print the four finished scripts and keep them next to your service menu.
Print the six scripts and your four. Use your browser's print on this page with the cards expanded (Expand all, top right). Scripts work best when they are rehearsed before the hard conversation, not composed during it.
After the track
30 min- Use one script this week on a real ask, and write down what the requester said first in reply.
- Find one vague yes you gave in the last month and rewrite it as a script with a deliverable and a date. Send the rewrite if it is not too late.
- Agree the escalation ladder with your manager: who sits on rung 2 and rung 3, and what each rung needs to bring.
- Put the whole Partnership kit in one place: the stakeholder map (session 1), the RACI matrix (session 2), the intake template and service menu (session 3), and these scripts.
Three questions
Clear all three and Partnership session 4 gets stamped in your passport.
1 · The BD lead asks data to write the UK supermarket penetration strategy. Which reply follows the order from Part 2?
Deliverable, date, decision, then the boundary with its reason. A is a door closing, B is the vague yes, C opens with the boundary and has no real date.
2 · Why does a vague "we'll help" usually end in resentment on both sides?
On the triage desk the Japan ask is 3 modelled hours as a consulted page and 40 if data takes all of it. "We'll help" never says which.
3 · Which of these deserves a plain yes, even when the timing is bad?
Internal work only data can do gets a yes; the conversation is about what moves, which is a capacity conversation, not a boundary. The other three are external and belong with brand.
What this session teaches, and where it came from
Every fact on this page, and its verification tier, is recorded in the course's source map.
Partnership session 4 cheat sheet · pin this
That is the Partnership track: their shoes, the ownership map, the front door and the scripts. If you have not done the Angles track, start at Angles session 1, how marketers think, and read every channel in four beats; if you have, the last stop is the triage desk again, this time with your own team's asks in it.