In AI, everyone’s talking about context — prompting, RAG, and bigger context windows. We try to engineer the perfect inputs so models behave.
Product marketing lives on the other side of that coin. My day is a relay: messaging review → jump into win/loss notes → hop to a launch stand‑up → back to a competitive teardown. That’s not a lack of focus. That’s the job.
Does this line of thought hold up? A quick look at the research says yes—and also “handle with care.”
- Attention does fragment: average time on a screen shrank from ~2.5 minutes in 2004 to ~47 seconds in recent years. Researcher Gloria Mark’s work tracks that slide and links it to how modern work happens. Read more @ Attention Span
- Switching has a measurable cognitive cost. Classic task‑switching studies show we pay “switch costs” each time we change tasks (goal shifting and rule activation), which slows us down and increases errors. Read more @ American Psychological Association
- And the workplace adds a “toggle tax.” One HBR field study observed ~1,200 app/site toggles per day; the reorientation time summed to ~4 hours a week (~9% of work time). Harvard Business Review
So yes, the PMM “context switching” instinct is real. But here’s the twist: the answer isn’t to pretend we’re ICs with single‑threaded backlogs. It’s to treat context switching like a product—with constraints, interfaces, and operating principles.
A new lens: the PMM Context Budget
Think of your week as a budget across three context types:
- Core context (the narrative you own): positioning, value prop, proof points. This is your “source of truth.”
- Timed context (time‑bound pushes): a Tier‑1 launch, a market category play, a pricing change.
- Ongoing context (ongoing flow of info): sales questions, analyst pings, competitive chatter.
I believe, great PMM weeks spent deliberately across all three—without paying compound interest on switch costs.
Five simple practices that can help manage context switching
-
Mode the day, not the quarter. Block 90–120 minutes for one context mode at a time (e.g., “Narrative Mode” AM, “Launch Mode” PM). Even splitting the day into two halves also works!Protect two “Core context” blocks early in the week when your energy is best. Research shows attention and error rates improve when you cut forced switches.
-
Ship a “State Packet” before you switch. Before you leave a task, write three lines: What just changed, What is true now, What’s the very next move. You’re doing human “context engineering”—preloading the next session with the right inputs (the way in‑context learning supplies examples to a model).
-
Run “Human RAG.” Well, I may be a force-fitting here :) But like how Retrieval‑augmented generation feeds models the exact docs they need; do a human version: for each active push, keep one living doc with the 7 items you always need
-
Problem statement
-
Target segment
-
Latest proof
-
Two risky assumptions
-
Top objection
-
Competitor counters When you re‑enter, you “retrieve” fast. (For the AI folks: yes, this mirrors context engineering—curating the right external facts for better answers.)
-
Single‑thread owners, multi‑thread visibility. For each context (Core/Timed/Ongoing), name a directly responsible individual who holds the narrative and a weekly cadence (15 minutes is enough). You reduce ad‑hoc toggles while keeping a line of sight for sales, PM, and comms.
-
Reset rituals > willpower. Between modes, do a some resets: Recap → Reframe → Ready → Re‑enter. Even short, intentional breaks help your brain switch more cleanly. (Gloria’s work points to the value of designing “empty space” into your day.)
AI’s “in‑context learning” is a fancy way of saying: answers depend on what you feed the system right now. PMMs live that reality. We’re the human routers of context—what to include, what to mute, what to sequence. Understanding the mechanics of model context helps us design better human context: tighter briefs, clearer prompts for internal copilots, and saner cadences for our teams.
Try this next week: a one‑page Context Ledger
-
Core: 2 blocks (90–120 min).
-
Goal: sharpen or validate the narrative.
-
Output: 1 truth we can prove.
-
Timed: 2–3 blocks.
-
Goal: move the launch one irreversible step.
-
Output: a decision, not a deck.
-
Ongoing: time‑boxed daily (30 min inbox + 30 min sales/field).
-
Goal: respond without reopening big threads.
-
Output: answers logged to the RAG doc.
At the bottom of the Ledger, keep a short “State Packet” for each context. That’s your bridge over the switch cost.
Bottom line: Context switching is baked into PMM work. The win isn’t eliminating it; it’s budgeting it.