Map feature co-occurrence for bundling decisions
Customers rarely ask for one feature at a time. They describe a job that needs several features working together — and the way they group those features is a clue about how to package them. NEXT reads where customers talk about your product — calls, support tickets, reviews, onboarding notes — and finds which features and products they mention together. The result is a bundling brief: which features co-occur, how often, in which accounts, and what that suggests for packaging.
When packaging gets decided in a review cycle, this co-occurrence pattern is usually the missing input. Most teams reconstruct it from memory and a few loud deals.
What the bundling brief looks like
Example output based on grouped customer conversations across calls, tickets, reviews, and onboarding notes.
Feature cluster
Automations + Approvals + Audit log
How often they co-occur
Mentioned together in 38% of conversations that reference any one of them — the strongest pairing in the set.
Who talks about them together
41 accounts, weighted toward mid-market and regulated industries.
Commercial exposure
About $2.1M ARR sits in accounts describing these three as one workflow.
What customers say
"We turned on automations, but legal wouldn't sign off until we had the approval chain and an audit trail. We needed all three or none."
"We evaluated you for approvals. The automation stuff was a bonus we found later — it should've been sold as a set."
Demand summary
Customers in compliance-heavy segments treat these three features as a single capability. They're currently split across two plan tiers, so buyers hit a wall mid-evaluation.
Signal strength
Strong and consistent for the three-feature cluster. Weaker for the looser pairing of automations + reporting, which shows up but inconsistently.
The brief arrives already assembled, not reconstructed from a few memorable deals.
How NEXT does this
NEXT reads where customers already talk about your product — sales and success calls, support tickets, reviews, onboarding notes, survey responses. It keeps a continuously updated record of which features and products show up together in those conversations, and how that grouping shifts by segment and account size. When a packaging review comes up, NEXT writes a bundling brief: the strongest feature clusters, how often each appears, the accounts and ARR behind them, and verbatim quotes. It lands where product and pricing already plan, and it can brief PMM with the same evidence. NEXT surfaces the pattern and keeps it current. What to bundle, how to price it, and which tier it lands in stays your call.
Why bundling decisions run on incomplete data today
Most packaging calls run on a mix of memory, a few loud deals, and whatever the analytics happen to show. Usage data tells you which features get used together once customers already have them — it can't tell you which features customers wanted together before they bought, or which combinations they describe as one job. The signal that matters lives in conversations, and it decays at every handoff: a rep hears "we needed all three," it gets compressed into a deal note, and by the packaging review it's gone.
Two familiar tools don't close the gap. A dashboard waits for someone to open it, and already assumes you knew which features to chart together. An AI assistant waits for someone to ask the right question, and tends to return the loudest example rather than the pattern across accounts.
The co-occurrence that should drive packaging isn't in a chart — it's in how customers describe their work, and it's only visible if something is reading every conversation and remembering the pattern.
How this compares to the tools you already know
Approach | Where the evidence lives | What product does at decision time |
|---|---|---|
Product analytics | Usage logs, post-purchase | Infers bundles from how existing customers click — blind to pre-purchase intent |
Customer interviews | A handful of recent calls, in memory | Reconstructs patterns from a small, recency-biased sample |
AI assistant | Wherever you point it, on request | Returns a tidy answer to the question asked, not the pattern across accounts |
NEXT | A continuously updated record of customer conversation | Opens a bundling brief with clusters, frequency, accounts, and quotes already attached |
What changes for the product team
You walk into the packaging review with the co-occurrence pattern already attached. Instead of debating which features "feel" like they belong together, you start from how often customers describe them as one job, in which accounts, and how much ARR sits behind the pattern. The conversation moves from "should these be a bundle?" to "this cluster is real in 41 accounts — does the pricing work?"
The three-feature cluster looked like a tier-design detail until the $2.1M in affected ARR was attached. A pairing you assumed was strong — automations and reporting — turned out to be inconsistent in the actual conversations, so you don't build a package around it. And you no longer reopen a quarter of call notes to reconstruct what customers asked for.
When you brief PMM, they get the same clusters and quotes, so the multi-product narrative is built on the same evidence as the package itself.
The packaging call stays with product and pricing. NEXT supplies the demand pattern; what gets bundled and how it's priced is your decision.
Downstream effects
Pricing starts from a demand pattern, not a guess. When a cluster is backed by named accounts and ARR, the tier conversation has something concrete to test against instead of intuition.
PMM builds multi-product narratives from the same evidence. The story customers hear matches the way they already describe the work, which lowers the risk the bundle reads as a vendor invention.
Weak pairings get dropped earlier. Combinations that look intuitive but show up inconsistently in conversation are visible before they become a package no one adopts.
Where the human stays in control
NEXT writes the pattern; it doesn't approve a bundle. You set how strong a cluster has to be before it appears in the brief — how often features must co-occur, across how many accounts — so the threshold matches how confident you want to be before packaging around something. You can require a human to review the clusters before they're shared to pricing. This is configuration work: you're tuning what counts as a real pattern, not signing off on each match.
What the output depends on
The brief is only as good as the conversations feeding it. A few things to get right:
Source coverage. The pattern needs calls, tickets, reviews, and onboarding notes across segments. If most of your enterprise conversations never get recorded, the brief will under-represent that segment's bundles.
Feature vocabulary. Customers rarely use your internal feature names. NEXT maps the language customers use to the features they mean; getting that mapping right is what separates a real cluster from noise.
Threshold calibration. Set the co-occurrence bar too low and loose pairings clutter the brief; too high and you miss emerging bundles. Start strict for a packaging review, looser for exploration.
Timing. Deliver the brief into the packaging review cycle, not after the tiers are locked.
Where this breaks down
Thin coverage in a segment
If SMB conversations are well recorded but enterprise calls mostly aren't, the brief will look like SMB bundles are the whole story. The pattern is only as representative as the conversations behind it.
Co-occurrence isn't intent
Two features showing up together can mean customers want them bundled — or just that both come up in long evaluations. The brief shows the pattern and the quotes; reading intent into it is still a human judgment.
Internal names don't match customer language
If the feature-vocabulary mapping is off, a real cluster can fragment into three weak ones, or two unrelated features can look paired. This is the most common cause of a misleading brief.
Packaging already locked
If the brief arrives after pricing has committed to a tier structure, it becomes a post-mortem instead of an input. The value is in the review cycle, not after it.
FAQ
How is this different from product analytics?
Analytics shows which features existing customers use together after they've bought — it's blind to what customers wanted together before purchase, or what they describe as one job in a sales call. NEXT reads the conversations themselves, so it surfaces intended bundles, including ones your current packaging makes hard to buy. The two are complementary: usage confirms, conversation reveals intent.
Does NEXT decide what to bundle?
No. NEXT surfaces which features customers describe together, how often, in which accounts, and with what ARR behind them. Product and pricing still decide what to package, how to price it, and which tier it lands in. The brief changes the inputs to that decision, not who owns it.
How many accounts does it take before a cluster is real?
You set the bar. For a packaging review you might require a cluster to appear across dozens of accounts and a meaningful share of conversations; for exploration you can loosen it. NEXT shows the frequency and account count so you can judge whether a pattern is strong enough to package around.
Can it brief pricing and PMM at the same time?
Yes. The same brief can land where product and pricing plan and be used to brief PMM. Because all three start from the same clusters and quotes, the package, its price, and the narrative customers hear are built on one body of evidence rather than three separate readings.
What if customers don't use our feature names?
They usually don't. NEXT maps the language customers actually use to the features they mean, so "the approval chain" and "sign-off step" resolve to the same feature. Getting this mapping right is what keeps a real cluster from fragmenting — it's the main thing to verify when you set it up.