How to Use G2 for Competitive Analysis

A G2 profile looks like a review page and behaves like a scoring system. Underneath the star rating sit three separate layers that get read as though they were one: text written by customers, structured numbers derived from that text, and a second score built almost entirely from data about the company rather than about the product. Most disappointment with this source comes from taking a figure produced by the third layer and describing it as though it came from the first.

What it genuinely holds is a competitor’s market as they and their customers have jointly described it. The categories a rival appears in were chosen by that rival, which makes the list a positioning claim rather than an observation. The feature-level scores are the closest thing to a like-for-like product comparison available free anywhere, because every product in a category is rated against the same named capabilities. And the comparison pages record which products buyers opened in the same session, so the grouping comes from what the market did rather than from anyone’s opinion.

What it does not hold is anything about commercial performance. No revenue, no customer count, no growth, no churn, and no reliable read on price. It also does not hold a representative sample of a competitor’s customers, because leaving a review is voluntary and heavily influenced by whether the vendor ran a campaign. Both limits are ordinary, and the failures come from forgetting them halfway through an analysis rather than from not knowing them.

1. 1Point yourself at the public profile rather than the seller dashboard. The analytics view of your own scored performance sits behind a seller relationship and is limited to rivals sharing one of your categories. Everything a researcher needs about any competitor is readable without signing in. Establish which of those two jobs you are doing before you spend an afternoon looking for a screen you cannot open.
2. 2Read the category list before you look at the rating. Categories are chosen by the vendor, so the list is a positioning statement they published about themselves. Write down which ones they entered recently and which they appear in that you do not. This takes two minutes and reframes the exercise more often than any number on the page will.
3. 3Build the feature-score row for yourself and two rivals. Within a category every product is rated against the same named capabilities, which makes those scores comparable in a way an overall star rating is not. Pick the ten features that decide your deals and fill one row per product. This is the artefact that survives the meeting; the star rating is not.
4. 4Record the score with its review count and its date, as one cell group. A rating on its own cannot be interpreted six months later, because you will not know whether it moved through new reviews, through the decay curve, or through a category change. Three numbers and a date cost nothing extra and make the next comparison possible.
5. 5Harvest names from the comparison and alternatives pages. Both are assembled from products buyers actually viewed together. Scan them for anything missing from your own competitive set: a rival the market treats as an option while your team has never discussed them is worth more than any badge printed beside it.
6. 6Read only the reviews written since your last visit, and code them as you go. Re-reading a whole corpus every quarter is how this becomes a project nobody repeats. Filter by date and by the segment you sell to, then record themes into your sheet while you read. A coded set of forty recent reviews beats an uncoded pile of four hundred.

This is the fact that changes how you use the platform, and it is published rather than inferred. G2 documents that a review starts losing weight the moment it is submitted, with the calculation refreshed daily. The decline is gradual across the first ninety days. A review holds a stronger weight for roughly its first eighteen months. After that the rate of decay accelerates, and at about three years old a review carries approximately three per cent of its original weight, which it then keeps indefinitely.

Three consequences follow, and each one breaks a habit worth breaking.

• A rival’s score can fall while nothing bad happens to them. If a competitor ran a successful review campaign two years ago and stopped, their excellent reviews are quietly aging out of relevance. The number drops. Nobody wrote a complaint. Read a declining score as a question about their review programme before reading it as a question about their product.
• Recent reviews are worth disproportionate attention. Since the last few months of text carries most of the scoring weight, sorting by recency is not merely convenient, it is closer to what the published number is actually measuring. Reading the oldest reviews tells you about a product that may no longer exist.
• You cannot rebuild the number, so do not try. Between the decay curve, the blend of distinct question types behind satisfaction, and the fact that reviews flagged for a conflict of interest can stay published while being excluded from scoring, the visible corpus and the scored corpus are different sets. A discrepancy between the stars you count and the score displayed is the system working.

The vertical axis of a category chart is the part most often quoted and least often understood. It is not a second opinion about the product. G2’s own documentation enumerates what goes into it: company size, employee count, web presence, social following, website traffic, geographic presence, category-specific product visibility, and the number of reviews. Only the last of those comes from customers at all.

Read that list again as a researcher rather than as a buyer, because two things follow from it. The first is that a competitor sitting high on that axis is being measured substantially on how large and how visible they are. That is legitimate information, and it is a different claim from the one people make when they point at the chart in a sales meeting. A well-funded incumbent with tired reviews outranking a better-liked challenger is the axis behaving exactly as designed.

The second is subtler and matters more. Some of those inputs cannot be measured directly from outside a company and have to be estimated. Website traffic for any domain you do not own is a modelled figure, and the error bands are wide enough that the same site draws materially different numbers from different providers. Employee counts for private companies come from profiles that lag reality. So a vertical position is a composite of counts and estimates, and the estimates carry their own uncertainty into it.

In rough order of how much each one repays the time. The first four are where most of the competitive value sits, and the last few matter mainly when you are checking something specific.

The structured layer: categories, features and comparisons

The category memberships. Start here, because it is the only part of the profile the competitor wrote and the only part that states an intention. Products appear in categories they opted into. A rival who has recently entered a category adjacent to yours has made a decision about where they intend to compete, months before that decision shows up in their marketing. A rival sitting in categories you have never considered is describing a market you may be defining too narrowly.

Feature-level ratings inside a category. Every product in a category is scored against the same named capabilities, which makes these the most directly comparable numbers on the platform. An overall star rating compares two different populations of reviewers answering about two different products. A feature score compares the same question asked about both. Pull the eight or ten capabilities that decide your deals, build the row for yourself and two rivals, and you have a defensible comparison in twenty minutes.

The comparison and alternatives pages. These are generated from what buyers actually viewed together, which makes them a competitive set produced by the market rather than by your team. The names you expect confirm nothing. The name you did not expect is the finding, and it is worth more than any rating on the page. Where your list and theirs disagree, one of you is wrong about who you compete with.

The review layer, and how much of it is verified

The reviews tab, read with the filters on. Filtering by company size, industry and role before reading changes the exercise from browsing opinions to sampling a segment. Read the recent critical reviews from companies that resemble your buyers, and ignore everything else on the first pass. Given the decay curve, recency is not merely a convenience here, it aligns your reading with what the score is weighting.

The verification badges on individual reviews. Reviewers establish identity through a LinkedIn profile, a business email on a company domain, or a personal email accompanied by a product screenshot. Reviews verified through a linked LinkedIn profile carry a badge saying so, and validated screenshots are labelled although the images are never published. When you are weighing a specific claim, the badge is a reasonable tiebreaker between two reviews saying opposite things.

What the vendor controls on their own page

The description, the screenshots, the video, the pricing tab and the feature list are all written by the competitor and edited whenever they choose, with no notification and no visible history. That makes this section useless as a record and excellent as a snapshot of current positioning. It is also the part most worth saving, because the previous version is otherwise unrecoverable.

The chart layer: Grid, badges and seasonal reports

The Grid placement and its quadrant. Useful as a category map, dangerous as a verdict. Read the horizontal position for customer sentiment and the vertical for size and visibility, and note which quadrant the whole field clusters in, because a category where everyone sits together is telling you the market is not differentiated on the axes being measured.

Badges, with their season and category read. Badges are generated per category, per segment, per report cycle, and there are a great many of them. A competitor displaying four is not making four market-wide claims. Read the small print on the image: a leader badge in a narrow segment of a niche category is a real achievement and a much smaller one than the logo suggests.

Seasonal category reports. These are the only convenient way to see where an entire field stood at a fixed past date. Downloading the same category report across three cycles turns a static picture into a trajectory, and trajectory is what almost every conclusion in competitive work actually rests on.

The two surfaces almost nobody opens

The pricing tab. Vendor-supplied, frequently incomplete, and occasionally more current than the competitor’s own pricing page because a different team maintains it. Treat any figure here as a starting point to verify rather than as a number to quote.

The discussions and questions area. Thin on most profiles and worth thirty seconds. Where it is active, it holds the questions prospective buyers asked out loud, which is a different and sometimes better signal than what existing customers chose to write about.

There is no download button and there is not going to be one. The corpus belongs to the platform, and the honest position is that a competitor’s reviews are theirs to publish rather than yours to take wholesale. That turns out to matter less than people expect, because the artefact that changes decisions here is small: three sheets, none of which needs more than half an hour a quarter per rival.

Sheet one: position, which only means anything with its date

One row per rival per quarter, seven columns: product, overall score, review count, the categories they are listed in, the badges currently displayed, the Grid quadrant if the category has one, and the date you looked. The date is not administrative. Because reviews decay on a published schedule, a score without the review count and the day it was read cannot be compared with the next one, and you will not be able to tell whether a fall came from new criticism or from old praise ageing out.

Sheet two: features, the one comparison nobody else builds

One row per product, one column per capability, filled from the category’s feature-level ratings. This is the most under-used artefact on the platform, because every product inside a category is scored against the same named features, which makes the numbers genuinely comparable in a way star ratings never are. Ten features across yourself and two rivals is a twenty-minute job and produces the table a product manager will actually argue with. The product comparison template is already laid out this way.

Sheet three: coded quotes, capped deliberately

Quote, theme, reviewer company size, rating, date. Code as you read rather than collecting first and coding later, and stop at forty recent reviews per rival. A coded set of forty beats an uncoded pile of four hundred, and the cap is what keeps this repeatable next quarter. We know what the uncoded pile costs because we did it once on purpose: our study of 500 coded G2 reviews found the themes had stopped moving long before the last hundred.

Filling them, and what tooling actually helps with

All three are filled by hand from the public profile, because no interface exists to do it for you and the reading is where the value is anyway. What software does help with is the part humans are bad at: noticing that something moved. A competitive intelligence platform watches public pages and reports changes, which is a different job from extracting a corpus and the only one worth automating here. If a vendor offers you the other thing, that is a signal about the vendor.

What the public profile gives you, which is nearly everything

No account is needed to read a competitor’s profile, their star rating and review count, their category memberships, their full review text, their comparison and alternatives pages, or the feature-level scores. For research purposes this is the platform. A free account adds some review filtering and access to the seasonal category reports, which is worth the two minutes it takes to create.

The seller dashboard, and why it is narrower than it sounds

Search results for this topic are dominated by the seller-side competitor view, which sits under analytics in the vendor interface. It reports score components and review trends over the previous six months for a set of rivals you nominate. Two documented constraints define it: you can track up to five competitors, and they must share at least one category with your product, because no data is reported for products outside a shared category.

That makes it a monitoring tool for a competitive set you have already decided on. It will tell you how five known rivals moved. It will not tell you that a sixth exists, and it cannot see a competitor attacking you from an adjacent category, which is the direction most genuine surprises arrive from.

What is not available at any price

There is no export of another company’s reviews, and the workaround people usually reach for is a collection method this site does not recommend and the platform’s terms do not permit. The manual route is sufficient and, done with filters and a coding scheme, produces a better result than an uncoded bulk pull would.

What to save on every visit

Score, review count, date, category list, and a capture of any vendor-written content you might later want to compare against. The first three take seconds. The last one is the only defence against a profile being quietly rewritten, which happens more often than most teams expect.

Reading is uncontroversial. The profile is published for buyers to read and you are, in the relevant sense, a buyer’s equivalent. Four constraints sit downstream of that, and the middle two catch more teams than the first.

• Read manually and record manually. Bulk collection from the platform is governed by its terms rather than by whether a page is publicly visible, and the terms are the operative document. Everything this page describes works by reading with filters and writing into your own sheet.
• Never write or arrange a review of a rival’s product. The platform screens explicitly for whether a reviewer works for the vendor or for one of its competitors, and reviews carrying that conflict are excluded from scoring. Beyond the platform rules, in several jurisdictions a fabricated review is a consumer-protection matter for the company that arranged it, not a marketing tactic that went wrong.
• Quoting a rating in your own material makes it your assertion. Put a competitor’s score in a battlecard and your company is now the one publishing a comparative claim. That lands under the rules governing comparisons between named products, not under anything to do with review platforms. Date the figure, state the category it came from, and refresh it, because these numbers move daily and a stale one is a claim that has become inaccurate.
• Leave anonymous reviewers alone. The firmographic fields exist to be read in aggregate. Attempting to work out which named individual wrote a critical review, or contacting them about it, is against the platform rules and is the kind of thing that ends up being described accurately by somebody else later.

• How many customers they have. A review count is a count of people who wrote something, and the ratio of customers to reviewers varies by an order of magnitude between vendors depending on how hard they ask. Proxy: named logos, case studies and filed disclosures.
• Whether their customers are renewing. Reviews are written by people who still use the product. The ones who left are silent here by construction, which makes this the wrong place to look for the most valuable signal. Proxy: retention disclosures from listed rivals and your own inbound switchers.
• What a deal actually closes at. The pricing tab shows what the vendor published, which is list price at best. Proxy: quotes surfaced in your own pipeline and buyer-reported figures.
• Whether a low rating is hurting them. Plenty of commercially strong products carry mediocre ratings, particularly where the buyer and the daily user are different people. Proxy: whether the rating appears in deals you are actually losing.
• What they are building next. Reviews describe the product as shipped. Requests hint at pressure, not at plans. Proxy: their documentation, their job adverts, and their released notes.

Three rhythms, because the profile changes in three different ways and a single check date catches only one of them. Scores recalculate daily, badges and charts move seasonally, and vendor-written content changes silently whenever the competitor feels like it.

A quarterly pass handles the first two comfortably. Record score, review count, category list and date for each rival in the same sheet you use for everything else, then read the reviews that arrived since your last visit rather than the whole corpus. Twenty minutes per competitor is enough once the sheet exists.

The third rhythm is the one that needs a deliberate habit, because there is no notification and no version history. Save a copy of any vendor-written content you might later want to compare against: the description, the pricing tab, the category list. Reopen off-cycle when a rival appears in a category they were not in before, when their review flow visibly changes pace, or when a badge appears in your sales conversations.

This source has an unusual property: its central number is designed to change while you are not looking. Reviews lose weight every day on a published schedule, which means a competitor’s score has a decay built into it rather than merely drifting. A figure copied into a battlecard in March is not just possibly out of date by September, it is arithmetically guaranteed to have moved, and the categories, badges and vendor-written description around it will have moved on their own separate clocks.

The cost lands on whoever repeats it. A seller quoting a rating from a battlecard nobody refreshed will eventually say it to a buyer with the profile open, and being corrected in front of a customer costs more than the point was worth. Holding a set of public pages under observation, so that movement announces itself rather than waiting to be discovered on the next manual pass, is the least contested argument for competitive intelligence software. In practice that means keeping a log rather than running a review: Flares records what each rival alters across their public surfaces and stamps it, which lets a competitive claim be tested against this week instead of last quarter. Which movements deserve a reaction stays a human call. A number that slipped because old reviews aged is noise; the same rival appearing in a category they had never entered is a decision about your market, and only somebody inside it can tell the two apart.