What Does This GEO Case Study for SaaS Cover?
This GEO case study covers a mid-market SaaS platform that raised its Perplexity citation share from 4 percent to 27 percent of a 220-prompt question set in six months, inside a category with four entrenched competitors and two dominant review aggregators. The win did not come from outspending anyone. It came from narrowing to a defensible wedge, rebuilding roughly thirty pages for retrieval, and manufacturing independent corroboration in that specific wedge.
The client is an anonymized composite drawn from engagements with workflow and revenue-operations platforms in the eight to twenty-five million dollar ARR range, selling mid five-figure annual contracts to operations leaders. Figures are directional and rounded to reflect what we typically observe. The transferable parts are the wedge logic, the retrieval rebuild, and the lag structure between publishing and observable citation movement.
Crowded categories change the math in a specific way. When four competitors have larger domains, more reviews, and a decade more content, competing on the same broad prompts is a losing position that can absorb an entire year of budget. The strategy that works is to become unbeatable on a narrow band of questions that the incumbents answer generically, then expand outward from citations you already hold.
Why Are Perplexity Citations Harder to Win in a Crowded Category?
Perplexity citations are harder to win in a crowded category because the engine resolves ambiguity by defaulting to consensus sources. When a category has an established vocabulary and several well-covered vendors, the retrieval layer finds abundant material from review aggregators, comparison sites, and incumbent blogs, and it has no reason to reach further down. A newer or smaller vendor is not excluded on quality; it is excluded on redundancy.
The second constraint is citation slot scarcity. A typical Perplexity answer surfaces roughly five to eight sources, and in mature categories two or three of those slots go consistently to the same aggregators. That leaves a much smaller contested pool than most teams assume. In baseline audits of crowded B2B SaaS categories we typically find that 50 to 65 percent of citation slots are occupied by the same six domains across the entire prompt set.
The third constraint is that Perplexity rewards specificity in a way that punishes generic positioning. Pages that describe a whole product category get passed over for pages that answer one narrow question with numbers, conditions, and edge cases. This is genuinely good news for challengers, because specificity is one of the few advantages that does not require scale, and it can be built in a quarter rather than a decade.
The Baseline: 4 Percent Citation Share Across 220 Prompts
The baseline was built from 220 prompts spanning problem descriptions, tool comparisons, integration questions, pricing and procurement questions, and implementation questions, each run three times and averaged. The client was cited in 4 percent of runs and mentioned without citation in another 11 percent. Two review aggregators appeared in more than 70 percent of runs, and the largest competitor appeared in 58 percent.
The diagnostic detail that shaped the whole program was where the client did appear. Its 4 percent of citations clustered almost entirely in integration and migration questions, not in category or comparison questions. That told us the market already treated the client as credible on a specific technical dimension, and that the existing citations were a foothold rather than noise. Programs that ignore the shape of their baseline usually attack the prompts they are furthest from winning.
We also scored the pages that were winning citations from competitors. Three quarters of them shared four traits: a direct answer in the opening paragraph, at least one number in the first hundred words, question-form subheads, and no gating. Fewer than a fifth were the highest-authority pages available on the topic. Retrieval was rewarding structure and specificity far more than domain strength, which set the technical direction for months three and four.
The Three-Signal Citation Wedge
The framework we used is the Three-Signal Citation Wedge, and it selects the prompt clusters worth fighting for by requiring three conditions to hold simultaneously. The first signal is incumbent genericism: the existing top-cited answer for that cluster is vague, dated, or gives a one-size-fits-all response that would frustrate a real buyer. The second is proprietary evidence: your organization holds data, telemetry, or operating experience nobody else can publish.
The third signal is buying proximity, meaning the cluster sits close enough to a purchase decision that winning it changes pipeline rather than only traffic. Clusters that satisfy all three are your wedge. Clusters that satisfy two are a later phase. Clusters that satisfy one are a distraction, no matter how much search volume they carry, because winning them produces citations that never convert.
Applying the wedge to 220 prompts produced twenty-six clusters, of which seven passed all three tests. Every asset published in months two through five mapped to one of those seven. The discipline of refusing the other nineteen clusters was harder to hold than the execution itself, and it is where most crowded-category programs quietly fail by spreading effort until nothing crosses the citation threshold anywhere.
Months One and Two: How Was the Wedge Narrowed?
Months one and two were spent building the baseline, scoring competitor pages, and running the wedge test with the client's product and customer success teams in the room. Subject-matter access was the rate limiter, not writing capacity. Seven working sessions of ninety minutes each produced the proprietary evidence base: anonymized deployment timelines, failure-mode data from support tickets, and integration behavior nobody else could document.
The wedge that emerged centered on complex multi-system deployments in regulated mid-market organizations, a scenario the incumbents addressed with generic enterprise messaging. The client had roughly four hundred such deployments to draw on, which meant it could publish real distributions rather than claims. That evidence became the spine of eleven assets, each built to answer one cluster question completely enough that a competitor could not answer it better without the same data.
The team also killed work in this phase, which is underrated. Two planned campaigns aimed at broad category prompts were cancelled, along with a comparison hub targeting the two largest competitors head-on. Head-on comparison in a crowded category almost always loses to aggregators that carry hundreds of reviews. The comparison content that shipped instead was scoped to the wedge, comparing approaches for regulated multi-system deployments rather than products in general.
Months Three and Four: What Does Rebuilding for Retrieval Involve?
Rebuilding for retrieval means making each page answer one question completely, in a structure a model can lift from. Thirty-one pages were rewritten to open with a forty to sixty word direct answer, carry question-form subheads, keep paragraphs under one hundred words, and place at least one specific figure within the first hundred words. Marketing abstractions were replaced with conditions, thresholds, and timelines.
Technical work ran alongside the writing. The team added FAQ and Article schema, fixed a rendering issue where key content loaded client-side and was therefore invisible to several crawlers, flattened the URL structure for the wedge clusters, and published a plain HTML documentation section that had previously lived inside a JavaScript application. That rendering fix alone made roughly a third of the site's substantive content retrievable for the first time.
Citation share began moving in week seven, which is typical. Early gains came almost entirely from the integration and migration clusters where the client already held a foothold, confirming that existing citations compound faster than new ones. By the end of month four, citation share had reached 14 percent overall and 31 percent within the seven wedge clusters, while broad category prompts had barely moved at all.
Months Five and Six: What Were the Results and Pipeline Impact?
By month six, citation share across the full 220-prompt set reached 27 percent, and within the seven wedge clusters it reached 58 percent, which put the client ahead of every competitor in that band. Mention without citation rose from 11 percent to 34 percent. The two dominant aggregators still held their slots, but the client had displaced incumbent blog content in roughly half of the wedge answers.
Months five and six added the corroboration layer: 38 written reviews focused specifically on regulated multi-system deployments, four practitioner podcast appearances, two ecosystem marketplace listings, and five guest pieces in operations-focused newsletters. The instruction was consistent throughout, which was to describe a specific situation with numbers rather than to praise the product. Corroboration published in month five was still producing citation gains in month eight, after the formal engagement had ended.
Commercially, self-reported assistant-sourced origin appeared on about 11 percent of new opportunities by month six, concentrated heavily in the regulated mid-market segment the wedge targeted. Those deals closed at roughly 1.4 times the average win rate. That segment concentration is the point. A narrow wedge produces narrow pipeline, which is exactly what you want when the alternative is broad visibility that converts at category-average rates.
What Transfers to Other Crowded Categories?
Four things transfer to any crowded category. Read the shape of your baseline before you plan, because your existing citations tell you where the market already finds you credible. Apply a three-condition test before committing to any cluster. Fix rendering and structure before producing new content, since unretrievable pages make the content budget worthless. Start corroboration by month two, because its lag is the longest variable in the program.
What does not transfer is the assumption that more content wins. This program published eleven new assets and rewrote thirty-one existing pages over six months, which is a modest output by enterprise standards. The incumbents published considerably more in the same period and lost citation share inside the wedge, because volume without specificity produces pages that retrieval treats as redundant with everything already indexed.
The other durable lesson is that generative engine optimization in a crowded category is a positioning exercise wearing technical clothes. At Lemniscate Growth we treat it that way, running the wedge analysis and GEO scoring from the GrowthGPT toolset before any content plan is written, so the first decision is which questions to own rather than how many articles to produce. Teams that reverse that order tend to spend two quarters proving it.
Ready to build measurable pipeline?
30-minute strategy session. No pitch. Just pipeline advice.
Get Your Free Strategy Session