Blog
Research-grounded notes for product, marketing, and sales teams.
Posts for product, marketing, and sales teams at software companies. Every figure is cited, with sources such as Microsoft Research, McKinsey, Gartner, Forrester, Loopio, IDC, and Melissa Perri's Product Operations book.
Published
Long-form notes.
What to build, how to position it, and how to win, with primary sources cited in every post.
Game Theory · Engineering
9 min
Your Competitor's Next Move Is a Probability Distribution, Not a Headline.
Feeds tell you what happened. A decision needs three more numbers: how likely each next move is, how much of it you can take, and how wrong it can go. The imperfect-information engine that computes all three — posterior, exploitability, blind-spot amplification, Monte-Carlo band — and runs live in your browser.
READ THE POST →
Analyst Relations · Engineering
8 min
You Shouldn't Find Out Where You Land in the Magic Quadrant From the Magic Quadrant.
The report drops once a year; the evidence it scores accumulates every day, in public. How placements from eleven analyst report types become data in the graph, the grain law that keeps product evidence out of vendor-axis scores, and the next edition's movement — projected before the briefing cycle starts.
READ THE POST →
AI Grounding · Architecture
11 min
Context Is the Missing Layer in Enterprise AI.
Enterprise AI fails not because of data scarcity but because the data lacks embedded context. Adapted with attribution from Andreas Blumauer (Graphwise) on the CDO Matters podcast. The pattern that already runs PubMed, EuroVoc, and SNOMED CT in production, applied to software product data.
READ THE POST →
Foundations
16 min
Why the Software Market Needed a Product Graph (and Why Nobody Built One Until Now).
Why earlier knowledge graphs of the software market stalled (shape drift, unresolved duplicates, and missing constraints) and the 99-node-type backbone PYRAMYD built instead.
READ THE POST →
Graph RAG
13 min
Microsoft GraphRAG, Anthropic MCP, and the Convergence Pattern Nobody Is Talking About.
Three independent research teams arrived at the same substrate conclusion within 18 months. We trace the citations, the benchmark deltas, and the architectural primitives that show up in every published GraphRAG variant.
READ THE POST →
Methodology
10 min
What 'Evidence-Grounded' Actually Means: A Field Guide for Buyers Tired of LLM Wrapper Demos.
Every vendor says 'cited.' We dissect what a real citation looks like, what model-grounded vs. graph-grounded responses actually return, and the 7 disclosure controls a procurement team should require before signing.
READ THE POST →
Product Management
11 min
Why Product Roadmaps Miss, and How Connecting Your Product to the Market Closes the Gap.
Roadmaps don't fail because PMs are bad at prioritization. They fail because the inputs to prioritization sit in seven disconnected systems. The graph substrate solves that.
READ THE POST →
Architecture
9 min
The Five Enrichment Scopes: Taxonomy, Row, Column, Group, Cell.
Why one number on a dashboard can be refreshed without paying to refresh the other million. The runner architecture behind it: one pipeline for a full-taxonomy refresh and a single-cell fix, one advisory lock so bulk runs do not collide, and why you pay in tokens only for the cell, field, or record you ask for.
READ THE POST →
APEX · Agents
11 min
APEX: 20 Specialist Agents, One Supervisor, 150+ Tools.
Why one mega-model loses to 20 routed specialists on questions about the software market. The orchestration architecture inside APEX — supervisor + specialist pattern, MCP access to competitors, signals, and battlecards from the AI client you already use, and the Sources panel that shows what each answer used.
READ THE POST →
Markets · Taxonomies
10 min
Taxonomy Pages: Every Category, Industry, Country, and Role Has Its Own Page.
More than 7,600 pages: 2,546 categories, 463 industries, 249 countries, and 4,349 roles. Each one is the materialized view of the typed graph at query time. The Bloomberg Terminal pattern, adapted for the software market.
READ THE POST →
Studio · Sheets
11 min
Studio Sheets: The Spreadsheet That Reads From the Graph, Not From CSVs.
The spreadsheet your team already knows — except the rows load straight from the graph, a formula can look up a graph node or run a model, and APEX can work the grid. The 25-column CSV stack Ops teams build by hand, replaced by a sheet that loads its rows from the graph.
READ THE POST →
Signals · Pipeline
12 min
The 2,000-Source Signal Feed: How a Competitor Move Reaches the Graph.
How a competitor move reaches your battlecard when you refresh it, not next quarter. Inside the pipeline — 2,000+ source feeds, on-demand refresh, content-hash dedup, resolving names to nodes, and the audit chain that makes every signal traceable back to its source. The engineering that narrows the battlecard refresh gap.
READ THE POST →
Product Operations
14 min
Product Operations Needs a Knowledge Graph: Reading Perri and Tilles' Three Pillars Through a Graph Lens.
Perri and Tilles' Three Pillars — Business Data & Insights, Customer & Market Insights, Process & Practices — are each structurally graph problems. The substrate change is the unlock.
READ THE POST →
Product Development · Engineering
13 min
Stop Building What Already Exists: Graph-Grounded Feature Discovery.
Microsoft Research's GraphRAG benchmark and the wider knowledge-graph literature point the same way: questions that span many systems need a graph, not a pile of documents. Here's the mechanism.
READ THE POST →
Competitive Intelligence
12 min
Battlecards Are Stale by Design: Rebuild Them From the Evidence When the Deal Needs Them.
In Crayon's 2024 survey of 700+ CI professionals, 58% said keeping battlecards up to date is a struggle. The structural cause: CI is a relationship problem. The substrate fix is a graph.
READ THE POST →
Sales Enablement · RFX
13 min
Why RFP Answers Should Start From Your Product's Evidence, Not a Document Library.
Loopio's 2026 benchmark: almost 80% of RFP teams now use generative AI, yet a response still takes 33 hours. The substrate underneath proposal tools has to change from document libraries to knowledge graphs.
READ THE POST →
AI Context Management
14 min
The Grounding Layer Every Agentic AI Project Needs — Or Joins the 40% Gartner Says Get Cancelled.
Gartner predicts more than 40% of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. GraphRAG and MCP point to the same fix: a typed context layer under the agent.
READ THE POST →
Sales · Procurement
12 min
Your Buyer Has Done the Research. Has Your Sales Team?
6sense's buyer research finds B2B buying groups are about 70% through their buying journey before they contact a seller. Procurement now researches vendors from graphs of the market; here is what sellers need before the first call.
READ THE POST →
See it with your own product.
Book a demo. We'll connect your product to its market, competitors, customers, and deals, and show APEX agents answering from that evidence with every source listed.