What Actually Moves: A Volatility Ranking of Competitive Intelligence, Feb–Aug 2026

Editor's note, 1 October 2026: the sections on Valona and Crayon below have been updated with facts from each vendor's own October 2026 materials (Valona's "Agentic AI for CMI" whitepaper; Crayon's State of Competitive Intelligence 2026 survey). These additions are clearly marked as externally sourced and are not part of the measured battlecard corpus described in this report.

Most competitive intelligence advice is a list of things you should be watching. It is rarely a ranking, because ranking requires the same competitor profile to be rebuilt repeatedly over time so consecutive versions can be diffed. That is what this corpus is: 2,433 battlecards produced between 23 February and 9 August 2026, of which 2,296 had an earlier snapshot of the same competitor to compare against. Across 147 competitors, every section was compared with its own prior version, producing a ranking of how often each competitive surface actually changes.

n = 2,296 consecutive comparisons across 147 competitors, 23 Feb, 9 Aug 2026.

The top of the ranking is the noisy end, not the important end

Advertising activity moves in nearly six out of ten comparisons. That is unsurprising once stated: paid campaigns are the cheapest thing a company can change, they are changed by a different team than the one setting strategy, and they are visible to anyone who looks. Content and social strategy sits just behind for the same reason. These two surfaces are the most likely to trigger an alert and the least likely to mean anything on their own.

The practical implication is uncomfortable for anyone running a competitive alerting programme: the surfaces that fire most often carry the lowest information value per event. If alerts are configured uniformly across a competitor profile, the top two rows of that table generate most of the volume. A single advertising change is close to noise; a sustained shift across several consecutive comparisons is a signal, and distinguishing the two requires exactly the kind of longitudinal comparison that alert-per-event tooling does not perform.

The middle band is remarkably tight

Messaging and positioning (50.1%), product strategy (49.3%) and pricing and packaging (48.6%) sit within 1.5 percentage points of each other. Three surfaces that product marketers habitually treat as operating on different clocks (positioning as slow, product as medium, pricing as rare) move at effectively the same rate over this window.

The pricing figure deserves particular attention, because pricing is the surface most often assumed to be stable. Roughly one in two comparisons showed a pricing or packaging change, whether through tier restructuring, seat-model changes, add-on repackaging or ordinary published-price churn. The operating assumption that a competitor's pricing page is something you check quarterly does not survive contact with this data.

The bottom of the ranking is the finding

Win/loss themes changed in 7.5% of comparisons. That is not a rounding error against a middle band clustered near 50%, it is a different order of magnitude, and it is the single most quotable number in this corpus.

Read it plainly: competitive positioning churns constantly while win/loss narratives barely move. The things a competitor says about itself (its ads, its content, its messaging, its packaging) are in near-continuous flux. The reasons buyers actually chose one vendor over another are close to static across a six-month window.

There are two defensible interpretations. The first is that win/loss is genuinely stable: the underlying reasons deals are won and lost are structural (integration depth, procurement fit, incumbent inertia, a specific missing capability) and those do not change in six months no matter how much the messaging around them does. If that is right, win/loss is the highest-leverage thing to invest in precisely because it is stable: work done once retains its value, whereas a teardown of an ad campaign is stale within weeks.

The second is that win/loss themes appear stable because they are the hardest surface to observe. Advertising is public and pricing is usually published, but win/loss lives in CRM notes, call recordings and the heads of account executives, surfacing publicly only when a vendor chooses to publish it. Low observed volatility may partly reflect low observability rather than genuine stasis.

Both readings converge on the same recommendation, which is why the finding is actionable either way: invert your effort allocation relative to your alert volume. The surfaces generating the most alerts deserve the least individual attention and the most trend-level aggregation. The surface generating almost no alerts is the one where deliberate, sourced, first-party work compounds, and where, notably, several vendors in the category have concentrated their product investment, which is itself a signal about where they believe the value is.

A caveat about a curve that is easy to misread

The number of distinct competitors this platform scraped per week rose over the observation period, from 39 in the week of 4 May to 144 in the week of 3 August. It would be easy, and completely wrong, to present that as accelerating market activity. It is IndustryLens's own customer onboarding curve (more accounts means more competitors tracked) and it is not even monotonic, dipping to 86 in the week of 13 July. It is noted here only so no one mistakes it for a market signal. Every rate in the table above is a share of comparisons, not a count, which is what makes it robust to a growing denominator.

If the volatility ranking is what the corpus says about competitive surfaces in general, the competitive intelligence category is what it says about one market in particular, and it is a useful case study: a handful of enterprise incumbents, a fresh wave of entrants, and a gap in the middle. The story in this window is not incumbents reaching down into the mid-market. It is entrants attacking upward from below, and doing so by naming incumbent price points explicitly in their own positioning.

The entrants are pricing at the floor

CompeteIQ has published a fixed tier starting at $7,000 with unlimited seats, a structure that removes per-seat expansion entirely, which is the mechanism by which enterprise CI contracts grow. BridgeStag positions itself as an interpretation-led alternative to Klue and Crayon, and in its more aggressive framing as a "Klue and Crayon killer" aimed specifically at startups and teams of five to twenty people. Crucially, BridgeStag names its wedge out loud: Crayon's "$20K–$40K annual price point" is cited as the thing it is undercutting. That is not a competitor inferring a rival's pricing, it is a challenger building its go-to-market on the public existence of an incumbent price floor.

Below that, the floor drops to zero. Signal Labs operates a freemium $0 tier. PriceGhost sits at $0–$99 per month. Jenova AI prices at $20 per month and positions explicitly against what it describes as AlphaSense's "$10,000+ anchor."

The incumbents are holding an enterprise price floor

On the other side of the gap, the published benchmarks are an order of magnitude apart. Crayon's enterprise deal-size benchmark sits at $50K.

AlphaSense shows the shape of the incumbent floor from the demand side, with reviews reporting pricing rigidity around a roughly $12,000-per-year anchor and pricing complaints appearing in 11.5% of reviews, alongside an observed shift toward consumption-based pricing that reads as an attempt to soften exactly that rigidity. Comintelli's modular pricing is a variation on the same theme: entry points without lowering the ceiling.

The gap is the mid-market

Put the two ends together and the structure is clear. Between a $0–$99/month freemium floor and a $50K–$180K enterprise floor sits a wide band with comparatively little dedicated product. That band is the mid-market. The entrants are below it, building volume; the incumbents are above it, defending ACV. The category is being pulled apart at both ends, and the interesting commercial question is which side reaches the middle first.

Everyone is moving pricing, and we can measure it

The corpus-wide pricing volatility rate (48.6%) is echoed inside the CI vertical. Every established CI vendor with a meaningful comparison history repriced repeatedly across the window:

Rival-role vendors only, 23 Feb, 9 Aug 2026. Four recently onboarded entrants: BridgeStag, PriceGhost, Rocket Intelligence and Signal Labs, have only two comparisons each and are excluded; two comparisons is not enough to characterise pricing behaviour, and no claim is made about them here.

Klue and AlphaSense stand out on rate rather than volume: seventeen and fifteen pricing changes across just twenty-three comparisons each. For the vendors with the highest public price floors, that is a lot of movement in how the price is expressed, exactly what you would expect from incumbents squeezed from below and looking for packaging answers that do not require cutting the anchor.

What a mid-market PMM should take from this

The transferable lesson is that a published price is a competitive asset your challengers will use against you by name. BridgeStag did not have to guess Crayon's pricing; it quoted a band and built a wedge from it. If your pricing is discoverable and your floor is high, assume it will appear on a challenger's comparison page and that the challenger will define the category boundary around it. The defensive move is not secrecy, it is having a deliberate answer for the segment below your floor before someone else supplies one.

The largest single structural event in this category during the window was not a product launch or a funding round. It was the arrival of the inaugural 2026 Gartner Magic Quadrant for Competitive & Market Intelligence.

A first Magic Quadrant is a category-legitimation event. Before one exists, a buyer proposing a purchase has to argue that the category is real; afterwards, the category is a line item an analyst has blessed, and the internal conversation shifts from "why do we need this" to "which one." For anyone in a young B2B SaaS category, that transition is the single most consequential thing that can happen to your sales cycle, and it happens to your competitors at exactly the same moment it happens to you.

The placements

• Crayon: Leader. Notably, the evaluation's focus was AI and autonomous agents, and Crayon's enterprise users named in connection with the recognition include Dropbox, ZoomInfo, Workday and Zapier.
• Klue(Leader.)
• Valona(Leader.)
• Contify(Visionary.)
• Comintelli(Visionary.)
• Watchmycompetitor, entry into the Magic Quadrant.

The interesting part is what happens next

Several Leaders and Visionaries at once means that, within weeks of publication, a mid-market buyer evaluating this category encountered multiple vendors each accurately describing themselves as recognised at the top of the same grid. Every one of those claims is true. Collectively they cancel out.

This is the predictable consequence of an inaugural Magic Quadrant, and it generalises beyond competitive intelligence. The recognition that was supposed to differentiate becomes table stakes almost immediately, because in a young category the placements cluster, there has not yet been enough consolidation to spread vendors across the quadrants. What survives is the second-order effect: not the badge, but the traffic the badge routes to you before everyone else finishes updating their homepage.

What this means if your category gets its first MQ

The differentiating window is measured in weeks, not quarters, and it belongs to whoever operationalises it fastest, into paid search, sales collateral, outbound, and the comparison pages that will rank for the category term for years. The badge is not the asset; the traffic spike is, and it decays.

The defensive implication follows. If several competitors can truthfully claim Leader status, "analyst-recognised" is no longer available to you as a differentiator, and any battlecard built around it is obsolete on publication, a case where messaging and positioning, the surface that changed in 50.1% of comparisons, moved across an entire category in the same month.

One vendor in the corpus operates at a materially different scale from the rest, and its trajectory is the clearest available answer to what happens to a market intelligence company that keeps compounding.

AlphaSense has exceeded $700M ARR and has initiated IPO preparations, with revenue growth reported at 40% year over year and 7,600 customers. The prior milestone in the corpus is instructive as a rate check rather than a headline: ARR moved from $500M in October 2025 to $600M in Q1 2026, alongside a $7.5B valuation and a $350M raise. The $600M figure is the earlier waypoint on that curve; the current position is $700M-plus and pre-IPO.

From search to execution layer

The product shift underneath the financials matters more to a product marketer than the ARR does. AlphaSense's SuperAnalyst AI agent represents a move away from passive search and chatbot interaction toward what the company frames as an execution layer, a system that produces work rather than answers. The supporting evidence is consistent with that framing: AlphaSense has been shipping PowerPoint and Excel work products as agent outputs, and has positioned itself as model-agnostic, spanning GPT-5.6 and Sonnet-5 rather than committing to a single provider. A "Key Debates" skill extends the same pattern, the unit of output is a structured analytical artefact, not a chat response.

This is the most important strategic distinction visible anywhere in the corpus. A search product's value is bounded by how quickly a human can read; an execution layer's value is bounded by how much work it removes. The pricing consequence follows directly, and it is visible: AlphaSense's observed shift toward consumption pricing is what you would expect from a vendor whose product now does units of work rather than serving units of access. Seat pricing prices the reader. Consumption pricing prices the work.

Accenture as investor and channel simultaneously

AlphaSense's first strategic global channel partnership is with Accenture, and Accenture's role is dual: investor and channel partner, with AlphaSense embedded into what the partnership describes as "agentic systems."

That structure is worth naming because mid-market companies frequently misjudge it. When a systems integrator takes an equity position and a distribution role at the same time, its incentives align to the vendor's growth rather than to margin on resale. It is an enterprise distribution mechanism no mid-market company can replicate, and it does not compete on product, it competes on the buyer never running an evaluation at all, because the vendor arrives pre-selected inside a transformation programme. Against a competitor with an SI partnership of that shape, the battle is upstream of the shortlist.

The counterweight: scale has not solved usability

The same corpus that documents AlphaSense's scale documents its friction. UI overload and a steep learning curve recur as independently observed themes; pricing rigidity around a roughly $12,000-per-year anchor shows up in 11.5% of reviews; and a challenger priced at $20 per month cites AlphaSense's "$10,000+ anchor" directly as its reason to exist. That is the category's whole dynamic in one vendor, extraordinary scale and analyst-recognised execution at the top, and a price-and-usability seam at the bottom that a $20/month product can name in its homepage copy.

Every significant vendor in this category shipped agent functionality during the window, which is precisely why agent functionality is no longer a differentiator. When the Magic Quadrant's evaluation focus is AI and autonomous agents and every serious participant has them, "we have an AI agent" carries the competitive weight of "we have a dashboard."

What the corpus shows instead is that differentiation has migrated to a different question: where does the agent live?

In the workflow tool

Crayon's Field Agent for Slack is the clearest instance. Sales reps submit voice notes or screenshots from inside Slack and the agent structures them into a searchable database. The insight is not the AI, it is that field intelligence is lost at the moment of capture, and the only reliable capture surface is the one the rep is already in. Crayon promoted it through blog content from 3 June 2026, and a Glean integration extends the same logic to enterprise search.

In the deal

Klue's investment is concentrated on win/loss: Blindspots for win-loss analysis, Win-Loss Clips, and a "Zombie Agents" positioning line aimed at agents that produce output nobody acts on. Read against the volatility table, this is a coherent bet: win/loss themes changed in only 7.5% of comparisons, so an artefact built on that surface retains value, unlike competitive content built on advertising activity at 59.6%.

In the deliverable

AlphaSense's agents emit PowerPoint and Excel work products. The agent does not live in a chat window; it lives in the file the analyst was going to have to build anyway, the execution-layer thesis made concrete, and the most direct threat to seat-based value, because a deliverable is countable in a way a conversation is not.

In the protocol

The most forward-looking placement is no interface at all, and by October 2026 Valona had made that pitch explicit: its own whitepaper frames competitive intelligence as an "intelligence layer for enterprise AI," delivered via MCP directly into Microsoft Copilot, Claude and ChatGPT, a sharper claim than the headless-layer groundwork visible earlier in this window. Watchmycompetitor operates 14 domain agents. Signal Labs, the freemium entrant, ships an MCP server. Comintelli's Telli is its agentic entry.

Notice who is furthest along on protocol-level access: a challenger at $0. MCP-style integration is structurally deflationary for incumbents, because it decouples the intelligence from the interface, and the interface is what seat pricing charges for. A vendor whose value is a dashboard has an incentive to be slow here; a vendor with no dashboard to protect has none.

UX debt is the entrants' actual opening

The price attack from below would not work if the incumbent products were frictionless, and the corpus shows they are not. The recurring, independently observed complaints cluster tightly: AlphaSense UI overload and learning curve; Kompyte alert noise, stale data, export gaps and UI issues; Crayon adoption difficulty and news noise; Owler data inaccuracy. Rocket Intelligence has built explicit de-positioning content against Klue, Crayon and Contify, and there is even internal sentiment at Klue questioning how replicable the product is with large language models.

The pattern is unmistakable: the dominant incumbent complaint is not missing capability, it is too much undifferentiated output. Alert noise, news noise, UI overload and adoption difficulty are all the same failure, the product surfaces more than a human can process and pushes the triage cost onto the customer. That is exactly what an execution-layer agent is designed to remove, and exactly the seam a lightweight entrant can attack without matching incumbent data coverage.

Vertical depth is the moat that entrants cannot cheaply copy

Against that backdrop, the defensible incumbent position in the corpus is not breadth, it is vertical depth. Valona has built out chemicals, food and beverage including GLP-1, seed and agriculture, and an agrifood themes database. Contify runs a pharma practice with 350 staff plus diagnostics and procurement coverage. Sedulo Group works in life sciences and oncology monitoring and leads sponsorship of Pharma CI USA. Watchmycompetitor covers manufacturing, government and regulatory; AlphaSense has explicit life-sciences positioning. By October, Valona's published focus had widened on paper to manufacturing, automotive and medical devices alongside chemicals and food/dairy, backed by a claimed 200,000 curated multilingual sources plus financial and trade data: Valona's own claim, not independently verified here.

Vertical depth resists the price attack because it is expensive in a way that does not scale down. A $20/month product can replicate a general-purpose agent; it cannot replicate a 350-person pharma practice or a regulatory monitoring taxonomy. The parallel move is services re-attached to software: Comintelli's "Intelligence as a Solution" framing, its Proactive Worldwide alliance and reseller network, Klue's advisory positioning against Clozd, and Sedulo's position as a pure consultancy. That services signal is thinner in the corpus and should be read as an emerging pattern rather than an established one, but it points the same direction: when the software layer commoditises, incumbents re-bundle the human layer.

What the category's own buyers say (external benchmark, not this corpus)

Crayon's 2026 State of Competitive Intelligence survey: self-reported, sample size undisclosed, read as correlational rather than causal, adds color this battlecard corpus can't measure directly. 65% of teams now run a dedicated CI platform, up from roughly a third in 2022. Teams sharing competitive intelligence weekly or faster are nearly twice as likely to report revenue impact as those sharing monthly or slower (79% vs 41%). Teams running AI agents in production or pilot report revenue impact at roughly double the rate of teams with none (82% vs 42%). And, the finding most relevant to the "where does the agent live" question above, internal knowledge (employee input, internal docs, call recordings) has overtaken competitor websites as the single most-cited CI source, 54% to 48%. That lines up with this report's own observation: public surfaces, pricing, ads, content, are the noisy, highly observable end. Internal data is the input no automated public-signal tool, including this one, can see.

The findings above describe one category in detail, but the operating lessons transfer to any mid-market B2B SaaS company with a competitive programme.

1. Allocate attention inversely to alert volume

Advertising and content/social are the two highest-volatility surfaces (59.6% and 56.8%) and will generate most of your alerts. Treat them as aggregate trend inputs reviewed on a cadence, not as individual events requiring response.

2. Check competitor pricing far more often than you think you need to

Pricing and packaging changed in 48.6% of comparisons, essentially the same rate as positioning and product strategy. If your process treats competitor pricing as a quarterly or annual refresh, your comparison pages are wrong for most of the year. Inside the CI vertical, all eleven established vendors registered repeated pricing changes, with Klue and AlphaSense each recording more than a dozen across just twenty-three comparisons.

3. Build on the stable surface, not the volatile one

Win/loss themes changed in 7.5% of comparisons. Whether that reflects genuine stability or low observability, the consequence is identical: work invested in first-party win/loss retains its value far longer than competitive content built on advertising or social activity. It is also the surface where you hold a structural advantage over any vendor, because you have access to your own lost deals and they do not.

4. Assume your price floor will be quoted back at you

BridgeStag names Crayon's "$20K–$40K annual price point" in positioning aimed at teams of five to twenty; Jenova AI names AlphaSense's "$10,000+ anchor" while pricing at $20/month. If your pricing is discoverable and your floor is high, a challenger will build a category boundary around it. Decide your answer for the segment beneath your floor before someone else defines it for you.

5. Treat the first analyst grid in your category as a sprint, not a badge

Several Leaders and Visionaries appeared simultaneously in the inaugural Competitive & Market Intelligence Magic Quadrant, so the recognition became table stakes almost immediately. What retained value was second-order: the attention the badge routes before the rest of the field updates its homepage. If your category gets its first grid, the differentiating window is short and belongs to whoever converts it into search, collateral and comparison pages first.

6. Stop selling that you have an agent; decide where yours lives

Agent capability is now uniform across serious vendors in this category, which means it no longer differentiates. What differentiates is placement: in the workflow tool where capture actually happens (Crayon's Slack Field Agent), in the deal (Klue's win-loss agents), in the deliverable (AlphaSense's PowerPoint and Excel outputs), or in the protocol layer with no interface at all (Valona's headless direction, Signal Labs' MCP server). Note that a freemium entrant is among the furthest along on protocol access, protocol-level integration is deflationary for anyone whose pricing depends on owning the interface.

7. Expect MCP to stop being a differentiator, and provenance to take its place

By October 2026 Valona was pitching MCP delivery into enterprise AI tools outright, and Crayon's own benchmark was telling buyers to put AI agents into the sales motion. That makes MCP access table stakes, not a wedge. The fight moves to two places instead: trust and provenance (can every claim be traced to a source, and is the underlying data licensed for AI use at all: Valona's whitepaper leans hard on this), and in-deal sales activation (Crayon's own data puts average rep readiness for competitive deals at 6.3 out of 10, the real gap any agent still has to close). A platform that cannot show its sourcing, or that stops at a battlecard instead of what a rep actually says on the call, loses on the metric buyers track: competitive win rate.

The honest limits of this report

This corpus measures whether a section changed, not whether the change mattered. A repositioned headline and a repriced platform both register as one change. It observes public and semi-public surfaces, which is why win/loss (the least public surface) may be understated. It covers 13 verticals, and not all of them are B2B SaaS. And it begins on 23 February 2026 because that is when the first battlecard in this system was generated, not because anything happened in the market on that date. Those limits are stated here rather than buried in the methodology because a report that could not be checked is worth less than one that can.