How to search every Gong call for a feature request, not just one
There is no single search box in Gong that answers "every time anyone mentioned this feature, across every call, with which accounts and how much revenue." Two features get you partway there: Trackers, which flag a phrase in every call going forward, and AI Ask Anything, which can answer a question against a set of calls it picks for you. Neither one, on its own, produces the rollup a product team wants.
In practice you combine them and accept a ceiling. Trackers give you an ongoing flag and a Team Insights view of how often it fires. Ask Anything gives you a fast answer, but only over a capped slice of calls it selects. Getting the full list, tied to accounts and deal size, means exporting and doing the join yourself, or using a tool built to do that join continuously. Here's what each Gong feature covers, where the combination breaks down, and what a small team can reasonably build before that point.
“we need SSO before rollout”
“any update on single sign-on?”
“SSO came up twice on this call”
Linear issue, quotes attached
insight report in NotionWhat Trackers do
Gong Trackers watch for a phrase, keyword, or trained concept and flag it wherever it appears "in calls and emails," going forward from the moment you create the tracker. Set one up for "bulk export" or "SSO," and every future call gets scanned automatically, with the hit surfaced in the call's summary. Creating a keyword tracker takes a few minutes: name it, enter the comma-separated terms and phrases you want flagged, save it.
What Trackers give you at the account level is thinner than it sounds. The aggregate view that exists is Insights > Team > Trackers, which shows how often your own team members bring up a tracked concept, not a count of how many customer accounts raised it or what those accounts are worth. To see which calls a tracker fired on, you filter Gong's Search page by tracker and read the list. That list is useful. It is also a filtered call list, not a report, and it starts from the day the tracker was created. A tracker set up this week has nothing to say about the eleven times customers mentioned the same feature last quarter.
What Ask Anything does
AI Ask Anything is closer to what people picture when they imagine "search all my calls." Ask it a question and it searches your calls and answers based on what it finds, no tracker setup required. The scope is where it stops being a full-library search: a broad question runs against "up to the top 100 calls" that match, a question scoped to a specific deal or account looks at up to 60 calls and 500 emails from that record, and a question about a contact draws on their 10 most recent calls and 80 recent emails.
For a team with a few hundred calls a year, top 100 might cover most of it. For a team recording a few hundred calls a month, it doesn't, and Ask Anything gives no indication of what it left out. It answers from what it retrieved, not from your full call history, which means it can undercount a request that's common but spread across more calls than its search window returns.
Where the API comes in, and where it stops
The remaining option is pulling calls out through Gong's API and searching the transcripts yourself, and Gong's own community forum is a useful warning about the scale of that job. One thread in the Gong visioneers community describes a team with roughly 7,533 calls to pull, hitting a wall where only 100 calls came back per request. Building a script that pages through the whole call library, keeps the pagination cursor from expiring, and re-runs it on a schedule to catch new calls is a real engineering project, not an afternoon task, and it still leaves the harder problem unsolved: matching "can we get bulk export" to "we need to pull our data out in one go" without a human reading both.
A worked example
Dana Okonkwo has led product at Brindle Systems for three years, shipping infrastructure monitoring to a few hundred mid-market accounts. In a QBR, a rep mentions that a customer asked, unprompted, for a way to schedule dashboard exports to land in someone's inbox every Monday. Dana remembers hearing something like that before, months ago, from a different account, but not which one or what was said.
She sets up a Gong Tracker for "scheduled export" and "recurring report" so future calls get flagged, then runs an Ask Anything query: "which accounts have asked about scheduling automated exports?" It returns three calls from the past few weeks, since that's what its search window reached, with reasonable summaries of each. It doesn't turn up the older mention Dana half-remembers, because that call sits outside whatever slice Ask Anything selected. She asks a rep to search manually by account name, and finds it: a call from four months earlier where a director at a different company said almost the same thing in different words.
By the time Dana has the tracker running and has manually confirmed the older mention, she has four data points that all describe the same request, discovered three different ways, none of which would have surfaced together without her doing the digging by hand.
Where this stops working
For an occasional check, tracker plus Ask Anything plus a manual read is a reasonable amount of effort. It stops scaling on three fronts at once. Call volume outpaces what Ask Anything's search window returns; tracker hits accumulate without a running account-level count anywhere; and every past request only shows up if you happen to remember it well enough to ask a targeted question. None of that is a Gong flaw exactly. Gong is built to make one call or one deal legible fast; a cross-account feature-request rollup with revenue attached is a different job.
That's the part we build Modem for. Modem is honest that it doesn't out-transcribe Gong; it doesn't do a better job of reading a single call. What it does is read the transcripts from your connected Gong workspace starting from the moment you connect it, with no backfill of calls that happened before that, and it treats every mention of a feature as one more entry in a running topic, matched against however that same request got phrased in Slack, a support ticket, or an email, whether or not anyone tagged it or thought to ask about it. That wouldn't have surfaced Dana's four-month-old call any faster; she'd still be searching by hand for anything that happened before she connected Gong. What changes is the next mention. When "scheduled export" comes up on call nine from a company Modem hasn't seen raise it before, the topic's count updates and the account gets attached, so the answer to "how many times has this come up, and from whom" is a lookup instead of a research project the next time someone asks. It reads transcripts only, never audio, and never writes back into Gong.
Two guides worth reading alongside this one: the six tools that mine product feedback from sales calls, which compares Modem's Gong integration against Chorus, Productboard, and Enterpret head to head, and how to handle duplicate feature requests, which covers the matching problem underneath this one: recognizing that a Gong call and a Slack message are describing the same request in different words.
Start with one tracker and one running note
Set up one Tracker for the feature request you keep half-remembering, run an Ask Anything query against your most active accounts to see what its search window turns up, and keep a running note, however plain, of every hit with the account name attached. That alone beats relying on memory at the next QBR. It also makes the ceiling obvious fast: the note works until the list is long enough that nobody wants to keep updating it by hand, which is exactly the point where an aggregation layer earns its keep.
