15 Minutes
There are good glossaries for AI builders. There are good glossaries for product managers. There is no good glossary for the people who sit between customer feedback and competitive movement — the ones who have to explain why churn spiked the same month a competitor shipped a feature.
That gap is the whole reason this exists.
Contents:
The seam (7)
1. Signals and sources
Signal
A single piece of evidence about what a customer or a competitor did, said, or changed. Broader than "feedback," which is customer-sourced by definition. Narrower than "data," which includes things nobody will ever act on.
Source
A system that produces signals. Support tickets, sales calls, review sites, community forums, changelogs, pricing pages, job postings. Most teams have more sources than they read.
Structured vs. unstructured feedback
Structured feedback arrives pre-categorized: a rating, a dropdown, a score. Unstructured feedback arrives as language. Structured is easy to count and tells you little. Unstructured is hard to count and holds the reason.
Signal volume vs. signal density
Volume is how much you collect. Density is how much of it carries a decision. Adding a source immediately raises volume and almost never raises density.
Signal decay (HyperOrbit term)
The rate at which a signal loses decision value as it ages. A churn complaint is most useful the week it arrives. A competitor pricing change is most useful before your next renewal conversation, not after. Decay rates differ sharply by signal type, which is why one refresh cadence for everything is the wrong design.
Sampling bias
The distortion introduced by which customers actually give you feedback. Survey respondents skew toward the very satisfied and the very angry. The quiet middle is where most of your revenue sits and most of your churn starts.
Sentiment
Whether a piece of feedback reads as positive, negative, or neutral. Useful as a filter. Weak as a conclusion. Sentiment tells you the temperature, not the cause.
Intent
What the customer is trying to accomplish, as distinct from what they said. "Add a bulk export button" is a request. "I need to get this into a board deck by Friday" is the intent. Roadmaps built on requests ship buttons. Roadmaps built on intent ship the thing that made the button unnecessary.
2. Voice of Customer
Voice of Customer (VoC)
The practice of collecting, organizing, and acting on what customers say across every channel. Modern VoC is continuous and cross-channel. Older VoC was periodic and survey-led. The difference is not tooling; it is whether the program produces decisions or reports.
Taxonomy
The category system feedback is sorted into. A taxonomy that mirrors your org chart will hide problems that cross teams. A taxonomy that mirrors customer language will surface them and annoy your org chart.
Theme
A cluster of individual pieces of feedback that share an underlying cause. The unit most teams should be acting on. Individual verbatims persuade; themes prioritize.
Verbatim
The customer's exact words, unedited. The thing that survives a summary chain and changes a room's mind. Any system that reduces feedback to counts without preserving verbatims has thrown away the persuasive half.
Closing the loop
Going back to the customers whose feedback drove a change and telling them. The step almost everyone skips, and the one that converts feedback-givers into repeat feedback-givers.
Churn signal vs. churn event
A churn event is a cancellation. A churn signal is the evidence that precedes it — a support escalation, a drop in usage, a competitor name appearing in a call transcript. Programs that measure events are reporting history. Programs that catch signals have something to do about it.
Feature request
A customer's proposed solution. Valuable as evidence, dangerous as a specification. The underlying need is usually adjacent to what was asked for.
Score metrics (NPS, CSAT, CES)
Net Promoter Score, Customer Satisfaction, Customer Effort Score. Standardized numbers that make trends comparable across time and teams. They compress the reason out of the signal by design. Use them to detect that something changed, not to explain what.
3. Competitive Intelligence
Competitive intelligence (CI)
The practice of collecting and acting on information about competitors, buyers, and market movement. Distinct from competitor monitoring, which collects and stops.
Battlecard
A short internal reference on how to sell against a specific competitor: their positioning, their weaknesses, your counters. The most-produced and least-updated artifact in most go-to-market functions.
Battlecard drift (HyperOrbit term)
The gap between what a battlecard claims about a competitor and what that competitor currently does. Drift begins the day the card is published and remains invisible until a rep loses a deal due to an out-of-date objection.
Win/loss analysis
Structured post-mortems on closed deals, won and lost. The highest-signal CI source most companies own and do not read. Loss reasons recorded by the rep who lost are not the same as loss reasons given by the buyer.
Competitor trigger
A discrete, observable competitor action: a launch, a pricing change, a funding round, a repositioning, a leadership hire. The CI unit that can actually start a workflow.
Positioning claim
A specific assertion a competitor makes about what they are and who they are for. Worth tracking verbatim, because claims change faster than products and are the first thing to move when a competitor is repositioning against you.
Share of voice
How much of the conversation in a category a given company occupies, across search, social, review sites, and increasingly AI-generated answers. A proxy for salience, not for quality.
Category seam
The space between two established buying categories, where a need exists but no budget line does. Seams are underserved because incumbents on either side are optimizing for their own category's buyer. They are also where a new category can be created, and where a company can be badly misunderstood for a long time.
4. The seam
This section covers terms for the space where customer feedback and competitive intelligence meet. It is the least standardized vocabulary in this glossary, because it is the least mature practice.
Competitive mention
A piece of customer feedback that names a competitor. The single highest-value signal type in most feedback corpora, and the one most likely to be lost in a taxonomy built around product areas rather than customer language.
Switching signal
Evidence that a customer is evaluating an alternative. Ranges from explicit (a competitor named in a renewal call) to weak (a sudden drop in a champion's usage after a competitor's launch week). Individually ambiguous, collectively legible.
Attribution lag (HyperOrbit term)
The delay between a competitor's action and the appearance of customer feedback caused by it. A competitor ships a feature; six weeks later your support queue fills with requests for it. Teams that read the two streams separately experience this as a mysterious spike in demand. Teams that read them together experience it as a consequence.
Counter-positioning window
The period between a competitor's move and the market settling on how to interpret it. Short. The advantage goes to whoever can connect the move to what their own customers are already saying, before the narrative sets.
Bidirectional signal loop
An architecture in which customer feedback informs competitive analysis and competitive events inform feedback interpretation, continuously and in both directions. The distinction from a pipeline is direction: a pipeline moves signals in one direction and ends at a dashboard.
CIA × VoC loop (HyperOrbit term)
Our name for the specific bidirectional loop between our Competitive Intelligence Agent and our Voice of Customer Agent. We qualify it every time rather than saying "the loop" because several vendors now use "loop" language across different architectures, and the unqualified term no longer means anything.
Signal gap (HyperOrbit term)
The distance between the moment a signal exists somewhere in a company's systems and the moment someone makes a decision with it. Most organizations measure how much feedback they collect. Almost none measure this.
5. Agent mechanics that matter to buyers
There are excellent glossaries covering agent internals for people building them. This section covers only what changes an evaluation decision.
Agent (vs. dashboard)
A dashboard presents what happened and waits. An agent decides what to look at, acts, checks the result, and adjusts. The word has been diluted to near-uselessness in this category — nearly every vendor now claims it. Ask what the agent decides without being asked, and what happens when it is wrong.
Stop condition
The rule that tells a running agent it is finished. The hardest part of agent design and the most revealing question in a vendor evaluation. An agent with a loose stop condition does not fail loudly; it produces plausible output indefinitely.
Human in the loop
A design in which a person reviews, approves, or corrects agent output before it has consequences. Not a weakness. In customer and competitive intelligence, where the cost of a confident wrong answer is a misdirected roadmap, the correct architecture is usually the one.
Grounding
Tying every output back to the specific source signals that produced it. Without grounding, an insight is an assertion. With it, an insight is an argument you can check. The practical test: can you click from a stated theme to the verbatims underneath it?
Permissions wall
The point in onboarding where a tool needs access to systems the buyer does not personally control. Support tickets, CRM records, call recordings. It is the real constraint on adoption of any tool that reads a company's signals, and it is the reason self-serve activation is harder in this category than in most others.
Conclusion
Why we wrote this
Every term here describes something we ran into building HyperOrbit, or something a customer said they had no word for.
If we've defined something badly, or missed a term you use, tell us. We'll update it and credit you.
Further reading on the AI side of this vocabulary: Paweł Huryn's AI Product Manager Glossary covers agent architecture, evals, and model mechanics in far more depth than we do here.
