A PMM at a Series B observability company told us she was getting 240 competitor signals a week — pricing changes, G2 reviews, hiring posts, blog updates, product release notes, podcast mentions. She was reading maybe 30 of them. Of those 30, two or three actually changed what sales said in a deal. The other 27 were noise dressed as intelligence.
The problem isn't the volume. The volume is fine — that's what monitoring tools are for. The problem is that no human reads 240 signals a week and ranks them honestly against "does this change a deal we're in this quarter." So the PMM either skims and forgets, or picks the loudest signal (a competitor's funding round), or punts the whole exercise to "next week."
An analyst — meaning a structured LLM prompt with your context loaded, not a generic chat — is good at exactly this kind of triage. Not because it's smarter than the PMM. Because it doesn't get bored at signal 80, it doesn't anchor on the loudest item, and it applies the same scoring rubric to all 240 inputs without skipping the boring ones.
Here's how to set it up so it works on Monday.
What the analyst actually does
The analyst's job is not to tell you what's important. The analyst's job is to apply your rubric to every signal and surface the top of the rank. You decide the rubric. You decide what counts as "deal-changing." You decide which competitors matter this quarter and which are background noise.
If you skip the rubric step and just ask "what's important this week?", the model will hedge — it will hand back a list weighted toward whatever sounds dramatic in the source text. Funding rounds and CEO departures will rank above a quiet pricing-page change that's actually losing you three deals a month.
Step 1 · Define the rubric in writing
Before you load a single signal, write down — in one paragraph — what makes a signal worth your attention this quarter. Not in general. This quarter.
A working rubric for a PMM at a mid-market sales-tech company might read:
A signal is worth attention if it (a) changes how a competitor describes their ICP, (b) introduces a feature that maps to one of our top three deal-blockers, (c) involves pricing or packaging changes that affect our negotiation patterns, or (d) comes from a customer or prospect of ours by name. Signals about funding, hiring, generic blog posts, or product updates outside the named feature areas are background.
This paragraph is the most important thing you'll write all week. It's also the part most teams skip — they jump straight to "summarize this week's competitor news" and wonder why the output is mush.
The rubric paragraph should name
Step 2 · Load the strategic context once
The analyst needs to know who you are before it can rank what matters to you. This is a one-time setup per quarter, not a per-signal step.
Load: your current positioning brief, your top three deal-blockers from recent win/loss work, your current battle card index (titles only is fine), and the rubric paragraph from Step 1. If you have a Stratridge memory layer, this lives there permanently. If you're working in a vanilla LLM, paste it as a system message at the top of the thread.
The point is the model should never have to ask "who's your ICP?" mid-triage. That context is settled before signal one.
Step 3 · Feed signals in batches with consistent metadata
Don't paste 240 signals as a wall of text. Feed them as a structured list — one row per signal — with at minimum: source, date, competitor named, raw text or summary, and any internal flag (e.g., "this came from an AE forwarding it"). Batch sizes of 30–50 work well; over 80 and the ranking starts to flatten.
For each batch, ask the analyst to score every signal against the rubric on a simple scale (1–5 or low/medium/high), explain the score in one sentence, and propose one of three actions: ignore, file for monthly review, or escalate now.
The first time we ran this, the analyst flagged a competitor's quiet pricing-tier rename as 'escalate now.' I would have skimmed past it. It was the reason we'd lost two deals in the prior month and didn't know why.
Step 4 · Audit the bottom of the rank, not the top
Most teams check whether the top three escalations are right. That's the wrong audit. The top of the rank is usually obvious — a competitor sunsetted a product, a new feature shipped that maps to your weak spot. You'd have caught those without help.
The audit that matters is the bottom of the rank. Pick ten signals the analyst marked "ignore" and read them yourself. If you disagree with three or more, the rubric is wrong — not the model. Update the rubric paragraph and re-run.
This is where the practice actually compounds. The rubric gets sharper every week. By month three, the bottom-of-rank audit takes ten minutes and produces zero disagreements, and you trust the top of the rank without checking it.
Step 5 · Convert escalations into specific artifacts
A signal that scored "escalate now" should produce one of four things in the same week, or it shouldn't have been escalated:
If a signal escalates and produces none of these, it wasn't really an escalation. It was an interesting fact. Move it to "file for monthly review" and adjust the rubric so similar signals don't escalate next week.
What this costs to maintain
Honest accounting: the rubric paragraph takes 30 minutes the first time, ten minutes per quarter to refresh. Loading strategic context is a one-time hour if you don't already have it written down. Per-week triage with a properly set up analyst runs about 20 minutes — most of which is the bottom-of-rank audit, not the ranking itself.
Compare that to the four-to-six hours per week most PMMs spend either skimming dashboards or apologizing to sales for missing a competitor move. The savings are real, but they only show up if you do Step 1 honestly. Skip the rubric and the analyst becomes another source of signals you don't trust.
What to do Monday
Open a doc. Write the rubric paragraph from Step 1 — three to six sentences, naming your competitors, your deal-blockers, and what counts as deal-changing. Show it to one AE and one CSM and ask whether the criteria match what they'd flag. Adjust. Then run your first batch of 30 signals through the analyst with that rubric loaded. The first run will take an hour. The fourth run will take 15 minutes.
Keep reading
Analyst Use Case: Competitive Response Drafting
How PMMs use the Analyst to draft a defensible competitive response in twenty minutes instead of the usual three-day Slack thread
How to Use an Analyst for Battle Card Generation
A working PMM's process for generating battle cards with an AI analyst — what to feed it, what to verify, and where the human still has to do the work
How to Use an Analyst for Competitor Profiles
Competitor profiles are tedious to build and quickly go stale. An AI analyst, prompted correctly, can produce competitor profile drafts in minutes — but the draft quality depends entirely on the context you load. Here's the specific workflow and the six-part profile structure.
Analyst
AI strategy advice grounded in your own context — not generic playbooks.
The Analyst is a chat-based AI strategist that reads your Strategic Context, past audits, and competitive signals before answering. Ask it anything from 'why are we losing to Competitor X' to 'how should we reframe our pricing page' — and get answers that are actually about you.
- ✓Reads your own positioning data before responding
- ✓Grounded in audit findings and competitor signals
- ✓No hallucinated advice — evidence cited inline