Lifesight MCP Prompt Library
24 prompts. Organized by the decision you're trying to make, not the role you have. Every prompt is copy-paste runnable in Claude or ChatGPT once Lifesight MCP is installed.
How to use these prompts
Each prompt is paired with:
- When to use it - the trigger that should send you here
- The prompt - copy as-is, or edit the bracketed slots
- What you'll get back - the shape of Mia's answer
- What to do next - the suggested follow-up
Tools the prompts call:
- list_models - workspace inventory
- earch_knowledge_base - methodology, past experiments, Lifesight docs
save_insight - bookmark an answer back to Lifesight.
1. Channel deep-dive
The decisions you have to make in the next eight weeks.
1.1 Where to move the next 20% of Q4 spend
When to use it. You're locking Q4 budget this week and want a defensible reallocation before the auction tightens.
PromptStress-test my [Quarter] plan. Where am I over-investing relative to
incrementality, and what would moving [%] of that spend to higher-iROAS
channels look like in incremental revenue and CAC?
What you'll get back: A ranked table of over-invested channels with current iROAS, recommended deltas, and the projected incremental revenue lift. Methodology footnote underneath.
Next step: Paste the table into your Q4 plan doc. Save the answer with save_insight so you can re-pull it after the change is live.
1.2 The retargeting trap check
When to use it. Your team is defending a retargeting line item because the platform ROAS looks great.
PromptFor my last [weeks] of retargeting spend, show me platform-reported
ROAS vs Lifesight iROAS by ad set. Flag every ad set where the gap is
greater than 2x.
What you'll get back: Two-column comparison with gap highlighted. Mia adds a sentence on which retargeting audiences are most likely to be cannibalizing organic conversions.
Next step: Send the table to your media buyer with a "pause or cap?" tag on each flagged row.
1.3 The discount-or-brand decision
When to use it. Finance wants a discount; brand wants upper-funnel; you have to choose.
Prompt.Simulate two [Quarter] scenarios on my live model: (A) repeat last year's
[%] sitewide discount, (B) hold price and shift the discount budget
to upper-funnel video. Show incremental revenue, gross margin impact,
and projected [Quarter] carryover for each.
What you'll get back: A two-column scenario comparison with the carryover number isolated - this is usually the line that wins the argument.
Next step: Take both scenarios into the planning meeting. Save the answer so you can revisit it in February.
1.4 The Q1 cliff stress test
When to use it: You want to make sure your Q4 plan doesn't strand you in January.
PromptBased on my current [Quarter] plan, project the [Quarter] baseline assuming we cut
upper-funnel investment by [%] in week 1 of [month]. Where does the
revenue floor land? What's the brand-led investment in [Quarter] that would
raise that floor by [%]?
What you'll get back: A projection chart description with two scenarios and a callout of the specific Q4 line items that drive the Q1 floor.
Next step: Use this as your Q1 insurance line in your Q4 board memo.
2. Channel deep-dive
When one channel is acting weird and you need to know if it's real.
2.1 The "why did this channel spike" question
When to use it. A channel jumped 30%+ week-over-week and you don't know if it's incremental or correlational.
Prompt[Channel] spend was up [X]% last week and reported ROAS was up [Y]%.
What does my causal model say about the actual incremental change?
Was the lift real, was it pulled from another channel, or was it
correlation with [seasonal event / promo]?
What you'll get back: A causal decomposition: incremental contribution, cannibalization from adjacent channels, and external-factor contribution.
Next step: If the lift is real, plan to test scaling. If it's cannibalization, flag the channel mix for review.
2.2 The cross-channel cannibalization map
When to use it. You want to know which of your "winning" channels are stealing credit from others.
PromptFor my top [5 channels] by spend, show me the cannibalization matrix:
which channels are taking credit for conversions that other channels
caused? Rank by total mis-credited revenue.
What you'll get back: A 5×5 matrix or ranked list. Most teams discover that 1-2 channels are credit-thieves and 1-2 are credit-victims.
Next step: Recalibrate channel-level KPIs to net of cannibalization. Save the matrix for the next attribution debate.
2.3 The "is this new channel ready to scale" question
When to use it. You're testing a new channel (CTV, retail media, podcasts) and need to decide whether to 5x it.
PromptFor [new channel], summarize the incremental contribution since launch.
At what spend level does my causal model expect saturation to begin?
What's the recommended next-quarter budget cap based on the saturation
curve?
What you'll get back: A spend-vs-iROAS curve description, a saturation point, and a recommended cap.
Next step: Set the next-quarter cap below the saturation point. Plan a holdout test if confidence in the model is below 80%.
2.4 The channel deprecation check
When to use it. Before you kill a channel, make sure your platform numbers aren't hiding its real value.
Prompt[Channel] looks like an underperformer on platform ROAS. What does my
causal model say about its true incremental contribution, including
brand halo, view-through impact, and effect on downstream channels?
What you'll get back: True incremental contribution plus second-order effects on other channels.
Next step: If second-order effects are large, propose a 50% cut instead of a full kill. Save the analysis so you can re-test in 90 days.
3. Anomaly triage
Something looks off. Was it real or was it noise?
3.1 The morning anomaly check
When to use it. Slack lit up at 9am because a number moved.
PromptFind every channel where incremental contribution deviated more than
2 standard deviations from forecast in the last 7 days. For each,
say whether the deviation crosses the threshold for action and what
the recommended next step is.
What you'll get back: A short list of real anomalies, each tagged "act" or "noise" with reasoning.
Next step: Triage the "act" list with your team. Save the response as your morning briefing.
3.2 The campaign-level fatigue check
When to use it. Performance is degrading and you suspect creative fatigue.
PromptFor [campaign], show me incremental CPA week-over-week for the last
[8 weeks]. Identify when fatigue began and which creative cohorts are
the primary contributors.
What you'll get back: A trajectory analysis with the fatigue inflection point and the at-fault creatives.
Next step: Route the at-fault creative IDs to your creative team with a "refresh by [date]" note.
3.3 The "did the platform change something" question
When to use it. A platform changed an algorithm or attribution window and you want to isolate the impact.
Prompt[Platform] [reported / made an attribution change] on [date]. Decompose
my [platform] performance into change attributable to the platform
change vs change attributable to our actual marketing decisions over
that window.
What you'll get back: A clean split of the variance into "us" and "them".
Next step: Communicate the split internally so the team doesn't over-react to platform-side noise.
3.4 The competitor pressure check
When to use it. CPCs spiked and you suspect a competitor entered the auction.
PromptIn the last [4 weeks], where did my CAC rise without a corresponding
change in our spend, bidding strategy, or creative? Are these signals
consistent with new competitive pressure in specific auctions?
What you'll get back: A list of suspected competitive-pressure channels with confidence ratings.
Next step: Adjust bidding strategy on flagged auctions or shift budget to lower-pressure channels.
4. Board / CFO prep
The questions you need answered before the executive meeting.
4.1 The board-ready iROAS summary
When to use it: You're prepping the marketing slide for the board deck.
Prompt.Draft a board-ready summary of last month's marketing performance using
iROAS and incremental revenue (not platform-reported). Include channel
mix, biggest mover, biggest concern, and a one-line methodology footnote
my CFO can defend. Save it as a board prep note.
What you'll get back: A 3-paragraph summary plus a saved bookmark in your Lifesight workspace.
Next step: Drop it straight into the board deck. Link the bookmark in speaker notes.
4.2 The CFO "prove it" answer
When to use it: Finance is challenging a marketing budget and wants a defensible number.
PromptFor the last quarter, calculate the incremental revenue attributable
to marketing, the implied iROAS, and the confidence interval. Compare
to platform-reported ROAS so I can show the gap. Frame the response
in P&L language.
What you'll get back: A finance-grade response: incremental revenue, iROAS, confidence interval, P&L framing.
Next step: Bring it to the CFO meeting. Save the response for repeat questions.
4.3 The investor-update marketing line
When to use it: You're contributing to an investor update.
PromptIn two sentences, explain how marketing contributed to last quarter's
revenue. Use causal numbers. Avoid jargon. Audience is non-marketing
investors.
What you'll get back: Two clean sentences in the right register.
Next step: Hand to the IR team.
4.4 The annual planning anchor
When to use it: It's planning season and you're setting next-year marketing budget.
PromptBased on the [last X months] of causal data, what's the marketing
budget range that maximizes incremental revenue for next year?
Show me three scenarios: status quo, [+X%}, and [-X%}. For each,
project incremental revenue, blended iROAS, and the channels most
sensitive to the change.
What you'll get back: A three-scenario planning view.
Next step: Take into the planning meeting. Save as your annual planning anchor.
5. Methodology lookup
Pull the receipts when someone asks "how do you know?"
5.1 Pull the holdout setup
When to use it. You're explaining a recent geo-lift test to finance or to a new team member.
PromptFind the holdout setup we used in the [Q1 / most recent] geo test
and summarize the methodology in language a non-marketer can follow.
Include the markets, the test duration, and the confidence level.
What you'll get back: A plain-English methodology summary pulled from your Lifesight knowledge base.
Next step: Use it as the methodology slide in any deck that cites the test result.
5.2 Reconcile two conflicting numbers
When to use it: Two reports show different numbers for the same channel and you need to know why.
PromptMy [report A] shows [channel] at [number]. My [report B] shows the
same channel at [different number]. Search the knowledge base for the
methodology difference between the two reports and explain in 3
bullets.
What you'll get back: A bullet list reconciling the two methodologies.
Next step: Send to the team that flagged the conflict. Update your reporting glossary if needed.
5.3 Explain incrementality to your CFO
When to use it. Your CFO has asked (again) what "incrementality" actually means.
PromptPull the Lifesight definition of incrementality and rewrite it as a
3-sentence explainer for a CFO who has never run a marketing
experiment. End with the one sentence that distinguishes it from
attribution.
What you'll get back: A CFO-grade definition with the attribution contrast at the end.
Next step: Use it in your next finance review. Save it so you can hand it out next time.
5.4 Audit-trail any number
When to use it. Someone needs to know exactly which model and run produced a number you cited.
PromptThe number I just cited - [paste number or describe the answer] -
which model produced it, when was the model last run, and what's the
confidence interval? Give me the audit-trail summary.
What you'll get back: Model name, run timestamp, confidence interval, and the data window.
Next step: Drop into any document or email that needs an audit trail.
6. BFCM stress test
Pressure-test your peak plan before the auction prices it for you.
6.1 The peak-week incrementality forecast
When to use it. 6 weeks before BFCM, finalizing the peak plan.
PromptFor BFCM week, project incremental revenue and CAC for my current
plan. Identify the 3 channels with the highest probability of
saturation during peak. Recommend the spend cap that maximizes
incremental revenue without crossing into negative incremental.
What you'll get back: Forecast + saturation flags + recommended caps.
Next step: Apply the caps. Save the forecast as your benchmark - re-run after BFCM to score the plan.
6.2 The discount-depth simulator
When to use it: You're choosing between three discount depths for the BFCM offer.
Prompt.Simulate three BFCM offers on my live model: [X%] sitewide,
[X%] sitewide, and a tiered offer (BOGO on top SKUs, [x%] on
everything else). Project incremental revenue, gross margin
impact, and post-BFCM repeat-rate effect for each.
What you'll get back: A three-row comparison with the repeat-rate row often deciding the call.
Next step: Take the comparison into the offer-selection meeting.
6.3 The post-BFCM diagnostic
When to use it. First Monday after BFCM. Time to know what really happened.
PromptFor BFCM week, calculate incremental revenue vs forecast for every
channel. Identify the biggest positive and negative variance. For
each, attribute the variance to (a) our decisions, (b) channel
performance, or (c) market conditions. Save the diagnostic as a
post-BFCM note.
What you'll get back: A variance table with attribution split.
Next step: Use the diagnostic in the post-mortem. Save it as the seed for next year's plan.
6.4 The Q1 carryover read
When to use it. Second week of January. You need to know how much of Q4 is still working.
PromptQuantify the [Quarter] brand investment carryover effect on[Quarter] to date.
Which channels are still delivering incremental revenue from [Quarter]
spend? What's the implied lifetime value of the BFCM-acquired cohort
vs the pre-Q4 baseline?
What you'll get back: Carryover quantification by channel + LTV comparison.
Next step: Use it to defend Q1 baseline spend. Save it for next year's Q4 planning anchor.
A note on safety
Every prompt above is read-only against your causal model. None of them spend money or change a budget. Write actions - propose and apply budget reallocations - ship in Q3 with an explicit approval flow.
If you spot a prompt that should be here but isn't, email it to [email protected]. The library updates monthly.
Updated about 11 hours ago