How to Turn Competitor Mentions in Gong Calls into Product Gap Signals
A Gong tracker firing on a competitor's name tells you the name got said. It doesn't tell you whether the prospect brought it up because your product is missing something specific, or because comparing vendors is just what a buyer does at that stage of a deal. Turning a pile of those mentions into an actual gap signal means adding a step Gong doesn't do for you: recording what was actually said, next to the account and the deal it happened in, every single time, and then reading that record for the same gap recurring across different deals.
The setup is a tracker scoped correctly, a log with four fields, and a monthly read-through, not a monthly tally of how many times a competitor's name got said. The tally is noise. The recurring, specific reason underneath it is the signal, and it only shows up once someone has read enough of the actual quotes to notice the pattern.
What a tracker hit actually tells you
Gong ships two kinds of trackers, and which one you're using changes what a hit means. A Keyword Tracker matches literal words. Gong's own guidance for building one to watch competitors is specific about scope: "if you're creating a keyword tracker to track competitors, you probably only want to track when competitors are brought up by people not at your company" (see creating keyword trackers). Skip that scoping and every internal mention, a rep prepping a battlecard, someone asking in a coaching call how a deal against that competitor went, counts the same as a prospect saying the name unprompted.
A Smart Tracker is the other kind, built to catch a concept regardless of the exact words used. Gong describes Smart Trackers generally as "AI models that enable you to find when concepts are mentioned in conversations, even when they are said in diverse ways," which sounds like the fix for the phrasing problem entirely. The volume ceiling that's often quoted for Smart Trackers belongs to one specific version of the builder, though, and Gong's own docs flag it as legacy. The example-based Smart Tracker, the one you train by feeding it "at least 5 examples of real sentences from real calls," requires the concept to already show up in "at least 50 calls" in your dataset, on an account with at least 500 recorded calls total and 1,500 recommended for solid results. A specific competitive loss reason, "their attribution model handles multi-touch and ours doesn't," said in exactly that shape, rarely clears 50 calls before someone's already lost three deals to it. But Gong now defaults new accounts to the question-based Smart Tracker builder instead, and its own docs are explicit that this version "does not require a minimum call volume." A team on the current default builder isn't blocked by the volume problem described above; a team still relying on the older example-based builder, or working with a specific reason that's too rare to describe as a training concept at all, is back to a Keyword Tracker on the competitor's name, which catches the mention and nothing about the reason behind it.
Either tracker gets you the same limited output, a flag that the name was said, on this call, at this timestamp. What was actually meant by saying it is still sitting in the transcript, unread.
The pattern Imogen almost missed at Cutlass Analytics
Cutlass Analytics sells marketing-attribution software to mid-market marketing teams, and Imogen Rasmussen, who runs product there, is the one who has to make sense of every tracker hit. Cutlass has a Keyword Tracker watching for "Vantik," the competitor its reps run into most in competitive deals, set to catch the name from anyone on the call rather than scoped to prospect and customer speech the way the guidance above recommends. Over six weeks it fires on 11 calls, and Imogen's first reaction is to read that as one flat number, a hot competitor this quarter, eleven mentions and counting.
She listens to enough of the calls to notice they're not the same conversation repeated. Four of the 11 name a specific reason:
Prospect (evaluation call, mid-market retail account): We're leaning toward Vantik because they show attribution across every touch in the journey. You're still last-click only, right?
Three more are general vendor comparisons with no specific capability named ("we're also looking at Vantik, wanted to see both"). The rest are Cutlass's own reps mentioning Vantik in internal coaching moments, exactly the noise an unscoped tracker picks up and a properly scoped one is built to filter out. Only the four with a named capability are telling her anything she can act on, and the tracker's single number, 11, buried that distinction completely.
"I almost took the count of 11 straight to the roadmap review," Imogen told her team afterward. "The real number was four, and all four were the same specific gap. That's the one that mattered."
issue filed·
report writtenBuilding the log that survives more than one quarter
A spreadsheet or a Notion table with four columns does most of the work a tracker can't:
- Account and deal stage. Pull this from the call itself or your CRM; a mention in a first call reads differently than the same mention right before Closed Lost.
- The specific capability or reason cited, in a few words you'd recognize again next quarter: "multi-touch attribution," not "attribution stuff."
- A verbatim quote with a link back to the call timestamp, so nobody has to trust a paraphrase months later.
- Outcome, once it's known: won, lost, or still open.
Every time the tracker fires, someone opens the call, decides which bucket the mention falls into, and logs it, or discards it if it's internal noise or a comparison with no reason attached. Once a month, group the log by the reason column. A capability that shows up under three or more different accounts is the actual signal; a single account mentioning something once, however dramatically, usually isn't.
Two related pieces worth reading alongside this: why Gong's Smart Trackers miss feature requests covers the same training-data ceiling from the feature-request side, and how to search every Gong call for one request, not just one goes deeper on what Trackers and Ask Anything each do and don't reach.
Why this stops working past a certain volume
This works cleanly for a team with a handful of competitive deals a month and one person willing to listen to every tracker hit. It stops working for a predictable set of reasons:
- Someone has to listen to every hit. The tracker doesn't tell you which of the 11 calls had a real reason behind it; a human has to open each one and decide.
- Nothing dedupes across channels. The same missing-attribution objection that showed up on a Gong call might also be sitting in a lost-deal note in Salesforce or a Slack thread from the account's AE, and the spreadsheet has no way of knowing those are the same signal.
- The log lives apart from where the roadmap gets prioritized. A spreadsheet nobody but product opens tends to go stale the same way any manually maintained sheet does, updated in a burst before planning and ignored the rest of the quarter.
Where Modem replaces the spreadsheet
That's the point where we'd bring up Modem, and we're upfront about what it is: not a better transcript-miner than Gong's own AI, and not trying to out-parse a single call. What it does is read the transcript from your connected Gong workspace as calls come in, pull out what was actually said about a competitor rather than just that the name was said, and roll it into a topic tied to the account. Connect Salesforce alongside it and that topic carries deal stage and outcome too, so a capability mentioned across four separate accounts shows up as one pattern with four quotes attached, whether those calls happened three weeks apart or a competitor is compared once and never named again. The same objection sitting in a Salesforce lost-deal note or a Slack thread from the account's AE rolls into that same topic too, instead of sitting alongside the Gong quote as three disconnected mentions nobody's linked together. The log above still gets built. It just stops depending on someone remembering to listen to every hit and type the reason in by hand.
Below a handful of competitive deals a month, the tracker-plus-spreadsheet system in this guide is a genuinely reasonable way to run this. It's the volume, not the concept, that eventually makes a manual log too slow to keep up with.
Set this up in twenty minutes
Scope one Keyword Tracker to the competitor that comes up most, restricted to prospect and customer speech only. Create a four-column log: account, specific reason, quote link, outcome. Put 20 minutes on the calendar at month-end to read through whatever fired and group it by reason. The first time three unrelated accounts turn out to be citing the identical gap, the difference between a competitor mention and a gap signal stops being theoretical.
