How to Use Gong for Competitive Intelligence

Treat this as two different assets that happen to live in one product. The first is the raw record: recordings and transcripts of conversations your sellers had, in which buyers occasionally said a rival’s name and explained why they were considering them. The second is an analytics layer built on top of that record, which counts those moments, charts them and reports a win rate beside them. Almost everything written about this platform for competitive purposes describes the second and quietly assumes the first is complete.

The record is genuinely strong evidence, and it is unusual in this cluster for being evidence nobody produced for an audience. A buyer explaining on a Tuesday afternoon why they are also talking to somebody else has no incentive to shade the answer, which is more than can be said for a press release or a review. The analytics layer, meanwhile, is only as good as its configuration, and configuration is where competitive readings go wrong here.

What the archive does not contain is the competitor. It has no view of their pricing page, their release notes, their hiring or their funding. It holds a sample of your own market conversations, drawn from the buyers who agreed to talk to you, during the weeks they happened to be evaluating something. That sample is worth a great deal and it is not a substitute for watching the rival directly, which is the split this page keeps returning to.

1. 1Audit the competitor list before trusting a single number. Every competitive figure the platform reports is built on a keyword tracker holding names somebody typed in. Open it, note when it was last edited and who owns it. In most companies nobody has touched it since onboarding, which means the analytics describe the market as it looked at implementation.
2. 2Find what the list is missing. Filter search by the concept tracker Gong ships, which detects competitive discussion without any names, across a fixed quarter. Filter by your named list over the identical window. Open the calls in the first set that are absent from the second and read them: that difference is your blind spot, and it usually contains one real rival plus a lot of buyers comparing you to a spreadsheet.
3. 3Split one tracker per rival, using the words people actually say. A single tracker holding every name produces one undifferentiated line. One tracker per competitor produces a trend you can act on. Matching is literal unless you enable related word forms, so include the shortened name sellers say aloud, the parent company, and the mistranscriptions you found in the previous step.
4. 4Make the calls arrive instead of going to find them. Save the search and attach a stream so matching calls collect themselves and notify you. Competitive listening never fails because the search was difficult; it fails because it depends on somebody remembering. Twenty minutes a week against a queue beats a day a quarter against an archive.
5. 5Pull the quarter into a sheet, one row per call. Not per deal. A rival named in discovery and named again during negotiation are two separate facts about how the deal moved. The columns are set out in the extraction section below, and the discipline that matters is capturing the buyer's own sentence rather than your summary of it.
6. 6Rank rivals with the win rate, then argue with quotes. The competitor mentions insight tells you which rival deserves attention this quarter, and it cannot tell you that they beat you, because deals where anyone names an alternative are larger and more contested to begin with. Use the number to choose where to look and the three best quotes per rival to change what your sellers say.

One fact explains more of this source's behaviour than any other: the product ships two separate competitive instruments, and they do not measure the same events. Gong’s documentation describes a tracker of its own named Competition, which uses its models to detect when competitive topics come up and, in its own words, works without a configured list of competitor names. Alongside it sits a keyword tracker named Competitors, which an admin sets up, which holds specific names, and which is the tracker the competitor mentions insight in win/loss analytics reads.

So one instrument recognises the situation and the other recognises the string. That difference has a consequence nobody writes down: your competitive analytics can only ever see the rivals somebody remembered to type in, while the concept tracker sees the conversation regardless. In a market that has not changed for three years the two agree closely. In a market with new entrants, or one where buyers increasingly compare you against an in-house build rather than a vendor, they diverge, and the divergence is the interesting number.

Running the comparison, which takes about an hour

Filter search by the concept tracker across a fixed period, a quarter is usually enough, and write down the count. Filter by the name-based tracker over exactly the same window and write down that count. Then open the calls in the first set that are missing from the second and read what the buyer actually said. Three things tend to come out of the pile, and all three are worth having.

• Names that belong on the list. Usually one or two, often a regional player or a recent entrant your team has heard of and never formalised. Adding them costs five minutes and recovers their whole history, because the search runs across calls that already happened.
• Alternatives that are not vendors. Buyers comparing you with a spreadsheet, an internal tool or doing nothing at all. These never appear in a competitor list and frequently account for more lost deals than any named rival does.
• Ordinary noise. Someone describing a former employer, or a partner being confused for a competitor. Worth reading once so you know the false-positive rate on your own data rather than guessing it.

Why one tracker per rival beats one tracker for all of them

A single tracker holding every competitor name produces a single line: competitive mentions went up. That is almost never actionable, because the interesting movement is one rival rising while another fades, and a combined count hides exactly that. Splitting the trackers gives a trend per competitor, which is the shape you can take to a pipeline review. Gong states that keyword trackers are unlimited in number, so the only cost is deciding who matters.

Matching is literal unless you turn on the related word forms option, which means the tracker finds the string you gave it and nothing else. Brand names are the weak point: short, unusual or non-English product names come out of transcription inconsistently, and a literal match cannot recover a name it never saw spelled that way. Put the variants in deliberately, including the mangled versions you find while reading, and the shortened form your own sellers say out loud rather than the registered trademark.

The competitor mentions view in win/loss analytics reports the win rate on closed deals where a rival was named at least once, and it can be filtered by team, by period and by the CRM fields that matter to you. It is the number most likely to be quoted in a leadership meeting and the number most likely to be over-interpreted on the way there.

Read it as a statement about your funnel rather than about the competitor. Deals in which somebody names an alternative are systematically different from deals in which nobody does: they tend to be larger, run longer, involve more people and reach a formal evaluation. A lower win rate on those deals is therefore consistent with two entirely different stories, one where you lose head to head against that rival and one where you simply win less often in contested enterprise deals of any kind. The metric cannot separate them and neither can anyone quoting it without the surrounding detail.

What it is genuinely good for is ranking. If four rivals appear in your pipeline and the win rate against one of them is markedly worse than against the others, you have a defensible reason to spend the next month on that one rather than on whichever name is loudest internally. Ranking attention is a real decision and this number supports it. Concluding that a competitor is winning in the market requires evidence the archive does not hold.

Two filters improve the reading immediately. Segment by deal size, because the contested-deal effect largely disappears once you compare like with like. And segment by the period in which the deal closed rather than the period in which the call happened, since a long sales cycle means today’s win rate is describing conversations from two quarters ago. That lag is the quiet property of this source and it is worth stating out loud whenever the figure is presented.

The two instruments that decide what counts as a mention

The Competitors keyword tracker is the list of names an admin maintains, matched word for word, and it is the backbone of every competitive report the platform produces. Read it first and note its edit date. The Competition tracker built by Gong is the concept detector that fires on competitive discussion without needing a name, which makes it the only surface here capable of surfacing a rival you have never heard of. Its mirror limitation is that it tells you a competitive conversation happened and leaves you to read the transcript to find out with whom.

The three ways a mention reaches you

Search with the Trackers filter is where the work actually happens, because tracker filters combine with account, deal stage, team, date and call activity, and searching past calls with a tracker is documented and supported. Streams and saved searches convert that into a queue that fills itself, which is the single change that decides whether this survives a busy quarter. Call summary notifications push an email when a term appears, and are worth reserving for the one or two rivals whose appearance would change what somebody does that afternoon; set for a full competitive set they stop being read within a fortnight.

The four screens that aggregate what was found

Initiative boards plot tracker occurrence over time as a share of deals or calls, and Gong suggests no more than six trackers on one, which is a sensible ceiling for a competitive set as well. Win/loss analytics carry the competitor mentions insight covered above, and depend on deal outcomes flowing in from the CRM data, so their quality is bounded by how honestly closed-lost reasons are recorded. Team insights on trackers measure how often your own people say each term, which answers a question competitive teams rarely ask: a rival raised mostly by your reps is a coaching finding, not a market one. And deal boards attach mentions to open opportunities rather than closed history, which makes them the only forward-looking view here and the one worth handing to a sales manager.

The two routes to the evidence itself

The transcript and the moment hold the buyer’s own wording with the sixty seconds either side, and everything above is a route to here: the surrounding context is what separates a serious evaluation from a negotiating tactic, and no count carries that distinction. The documented API is the supported way to move your own calls and transcripts into your own systems, with published authentication, cursor paging and default limits of three requests a second and ten thousand a day. When each is worth using is the subject of the next section.

The deliverable from this source is not a dashboard screenshot. It is a sheet you can sort, pivot and quote from, and it is worth building even if you never leave the platform to look at it, because a competitive claim that cannot be filtered by segment loses most arguments it enters.

One row per call, not per deal

A rival named during discovery and named again during negotiation are two different facts about how a deal moved, and collapsing them into one row destroys the sequence that makes the archive worth reading. Ten columns cover it: date, account, segment or region, deal size band, the rival named, who raised them, the buyer’s own sentence, the objection attached to it, the stage the call sat at, and the outcome once the deal closes. Only the seventh is prose, and it is the one that ends up in a battlecard, so record what was said rather than your summary of it. The competitor tracking spreadsheet already carries most of that structure if you would rather not design it.

Three ways to fill it, and when each is worth the effort

Reading the search results and typing the rows takes twenty to thirty minutes per rival per quarter, and for most teams that is the right answer, because the reading is the work and the typing is a rounding error on top of it. The documented API becomes worth the engineering only when you want mention counts joined to pipeline data continuously rather than quoted quarterly, which is a reporting problem rather than a competitive one. And where a company wants the watching to be continuous rather than periodic, competitive intelligence platforms connect to a call archive through the vendor’s own integration, using your credentials against your own recordings, which is a different proposition from anything involving a competitor’s website.

Two pivots that repay building the sheet

Mentions by rival by quarter tells you which competitor is rising in your pipeline, and it is the chart that gets a competitive programme funded. Mentions by rival by segment tells you where, which is the one that changes what a team does on Monday. Both are two clicks once the columns exist and neither is available from the platform’s own views, because those aggregate across a workspace rather than across the cut you care about. Keep the quote column alongside them: the pivot decides where attention goes and the sentences decide what anybody says once it gets there.

What you need to be given

A seat, visibility of the calls you intend to read, and ideally the permission that governs tracker management. The last one matters more than people expect. The useful configuration is several trackers rather than one, each maintained as the market changes, and if every edit has to be requested from another team then the competitive set stays frozen by default. Agreeing a naming convention once and holding the permission with whoever owns the competitive programme removes a week of latency from every change.

What it costs

Gong does not publish pricing. There is no plan table and no per-seat figure on their site, so every number circulating in comparison articles is a third-party report of a negotiated deal rather than a vendor statement, and this page will not repeat any of them. The number that matters for the decision in front of you is simpler: if your company already records calls here, the competitive use adds no licence cost at all. It costs the hours somebody spends reading, which is roughly two or three per quarter once the searches are saved.

The retention ceiling

Gong publishes a standard retention period of the lesser of three years and the length of your customer relationship, with an admin setting that extends the policy to calls saved in the library. Two things follow for research. Three years is the outer bound on any trend you can build here, so a longer comparison has to come from a source outside the platform. And because the setting can be shortened, and purged calls cannot be restored, the configured value is worth checking before you plan a multi-year read rather than assuming the maximum applies.

Language coverage

Transcription runs across more than seventy languages by Gong’s own count, but the analysis layer is not uniform, and the documentation says so. Some features return results only for English calls, example-based smart trackers among them, while question-based smart trackers are available in every supported language. Keyword trackers match whatever strings you supply, so a multilingual team gets the most reliable competitive coverage by adding local spellings directly rather than relying on the concept layer to bridge markets.

Coverage of the archive itself

Before comparing counts across teams, check what proportion of each team’s meetings reaches the archive at all. Adoption is uneven in almost every rollout, and a region with lower recording discipline will look competitively quiet through an adoption artefact rather than anything a buyer said. That check takes ten minutes against closed opportunities and prevents the most common false finding this source produces.

The consent question that governs recording a call in the first place is settled before this page starts, and it is worked through under sales call recordings. What belongs here is the narrower question that comes after: a recording made for one stated purpose is now being read for a second one, and that deserves a deliberate answer rather than an assumption.

The honest framing is that competitive research on your own call archive is a normal internal use of a business record, and it stays normal as long as three things hold. The reading is done by people who already have legitimate access to those calls. Nothing identifying a buyer leaves the analysis. And the output describes patterns rather than individuals, so a battlecard says that mid-market buyers in one region report being quoted a lower figure, never that a named person at a named company said so.

The permission model is where that gets enforced rather than merely intended. Access to calls is configurable, and a competitive analyst usually needs a wide view to do the job at all, which makes it worth agreeing the scope explicitly with whoever administers the workspace instead of quietly inheriting whatever the default was. Where recordings touch regulated conversations, financial advice or health information for instance, the sensible answer is to exclude those teams from competitive reading entirely rather than to rely on judgement call by call.

Two things to avoid outright. A customer’s recording or transcript should not travel to an agency, a consultant or any other third party until somebody has re-read what your own recording notice actually promised. And do not repeat a competitor’s claimed shortcomings as fact because a buyer said them: a buyer is reporting what a salesperson told them, which is hearsay about a rival’s product and can be wrong in ways that are your problem once you put it in writing.

The boundary is clean and it is worth stating before anybody builds a programme on this source: the archive contains your market’s conversation about a rival, and contains nothing the rival did. Four specific gaps follow from that, each with a better place to look.

• Anything they changed quietly. A new price, a repositioned home page, a retired plan tier. None of it reaches a call until a buyer happens to mention it, which can be months. Watching the pages directly is the answer, and dated captures of how they looked before are covered under web archives.
• Where they are investing. Hiring, new markets and new functions are visible long before they show up in a sales conversation. The job description is the specification, as set out under competitor hiring.
• Their commercial performance. No revenue, no growth, no retention. A buyer’s impression that a rival is doing well is an impression. Filed accounts and registry documents are the closest thing to an answer for a private competitor, covered under company registries.
• Everyone who never spoke to you. The buyers most worth hearing from are the ones who shortlisted a rival and never called you at all. Your archive is silent on them by construction, and the closest available substitute is the public review record described under competitor reviews.

There is one more limit specific to the counting layer rather than to the corpus. The concept tracker can tell you competitive discussion occurred without telling you who was discussed, and the name tracker can tell you who without telling you it was serious. Neither ever supplies the thing a competitive programme is really asking for, which is whether the buyer was going to choose them. Only the transcript answers that, one call at a time.

Run this on three clocks, because the three jobs decay at different speeds and collapsing them into one quarterly session is what makes teams learn about a rival’s new pitch a season late.

Weekly, someone skims the stream. Twenty minutes is enough to read the competitive calls that arrived since the last look, and this is where a genuinely new claim gets caught while it still matters. Nothing is counted at this stage and nothing is written down beyond the two or three sentences worth keeping.

Quarterly, the counts and trends get read properly, the tracker comparison from earlier on this page gets rerun, and any names it surfaces get added. Log what you added and why, since that log is the record of how fast your market is moving. This is also the point to re-read the tracker list itself and delete the rivals who have stopped appearing, because a competitive set that only ever grows becomes unreadable within two years.

Off cycle, reopen immediately on four triggers: a rival appears in a live deal in a segment they had never touched, a seller reports a claim nobody has heard before, your win rate against one name moves sharply in a single quarter, or the workspace retention setting changes. The last one sounds administrative and is not. If retention shortens, the history behind your trend lines disappears without notice and nothing in the interface will tell you it happened.

Every other source in this cluster updates when the competitor acts. This one updates when a buyer both notices what the competitor did and chooses to mention it to one of your sellers, which means the delay is not a matter of how often you check. A rival can change their pricing in January, and the first trace of it here is a February discovery call, and the second is a March negotiation, and the count that finally looks like a trend arrives in April. The archive is a lagging indicator by construction, and no amount of diligence inside the platform converts it into a leading one.

That lag is expensive in exactly one way. Sellers spend the intervening months answering a pitch nobody has documented, each of them inventing a response alone, and the good answers stay in the heads of the two reps who worked it out. Narrowing that window means somebody has to be reading the rival’s own pages while the calls are being listened to, and doing both at once is the whole proposition behind competitive intelligence software. Flares handles the watching half: a running, dated file of what each rival changes on the pages they control, so a January edit is on record before the February call raises it, and the seller can be briefed rather than left improvising. Deciding what any of it means for your deals stays with a person. A repricing that matters enormously in one segment is irrelevant in another, and the archive described on this page, read by somebody who knows your buyers, is the only place that judgement can come from.