PLG reprices in public, sales-led reprices in private — the visible velocity gap
Public pricing churn is almost entirely a product-led (PLG) phenomenon: of the corpus SKU changes with a known go-to-market motion, 41 came from PLG vendors and just 2 from purely sales-led ones. The gap is real — but it's also a mirror, because sales-led vendors reprice just as often, invisibly, inside contracts.
What's happening — and why
What's happening: sort every logged pricing and packaging change by how the vendor sells, and the split is lopsided — 41 changes from product-led (PLG) vendors, 6 from self-serve-only, and 2 from purely sales-led. The only serial repricers in the corpus (Perplexity, HeyGen, WellSaid, You.com) are all product-led.
Why: PLG pricing lives on a public page, is self-served, and can be A/B-tested, so it gets iterated in the open and often. Sales-led pricing changes happen one contract at a time, at renewal, behind an NDA — so they leave no public trace. The visible velocity gap therefore measures two things at once: PLG vendors genuinely iterate faster, and theirs are the only changes anyone outside the company can watch happen.
How it works
Evidence over time
14 supporting · 4 counter — hover or tap a point for detail, click to jump to the row.
Evidence
| Company | Date | What happened |
|---|---|---|
| Corpus tally | Jun 2026 | SKU-change events (packaging + price-change + deprecation) by GTM tag of the company: PLG 41, self-serve-only 6, sales-led-only 2. Public SKU churn concentrates almost entirely in the PLG/self-serve cohort. |
| Perplexity | Jan 2026 | PLG/dual-motion: three public ladder restructures in the window — API split into Search + Agentic Research, then enterprise split into Pro/Max — the corpus's most frequent public repricer. |
| Windsurf | Apr 2026 | PLG: swapped credits for quotas, raised Pro $15→$20, added a $200 Max — a public, self-serve repricing iterated live (and publicly contested on Reddit). |
| Dust | Jun 2026 | PLG: replaced a two-year-stable flat €29 plan with a credit-metered Free/Pro/Max ladder — the kind of full public restructure self-serve vendors can ship without a sales motion. |
| HeyGen | May 2026 | PLG: two public packaging changes in the window, including a whole-ladder cutover to a unified credit currency. |
| You.com | Apr 2026 | PLG: two public pricing changes in the window (commodity search-call cut into the $5–7/1k band, plus research-endpoint repricing) — iterated openly on the pricing page. |
| Batch tally | Jun 2026 | In the 2026-06-30 batch, all public price/packaging moves were PLG/self-serve (Shortwave, Creatify, lemlist, Lokalise, Qodo, Hume, DeepInfra, Vercel, RunPod, Together); the lone sales-gated mover (Mercor) changed only a public marketing stat and left buyer pricing behind a contact form — a second-cycle confirmation of the visibility gap. |
| Window tally | Jul 2026 | The first window large enough to normalize the claim. 101 customer-facing events (packaging + price-change + deprecation) logged between 2026-07-15 and 2026-07-30, attributed by the mover's sales_motion tag: 84 plg, 9 self-serve-only, 8 sales-led-only. Raw ratio 10.5:1, not ~20:1. Normalized: plg-tagged companies are 224 of 353 (63.5%) and produced 83.2% of events — a 1.31x over-index; sales-led-only companies are 73 of 353 (20.7%) and produced 7.9% — a 2.6x under-index. Event rate per company differs by roughly 3.4x, so a meaningful share of the original 20:1 was corpus composition rather than behaviour. The gap is real and smaller than claimed. |
| Mercor | Jul 2026 | Two more sales-only events, both public and both price-free — the invisibility mechanism demonstrated twice more. On 2026-07-21 Mercor renamed three enterprise product lines site-wide (Business / Enterprise evals / Data partnerships → Enterprise agents / Human data / Data monetization), added an APEX research line and grew its data-integration count 34+ → 50+; on 2026-07-29 it disclosed a $100K company-referral bonus cap on its Data page and updated talent-side stats (average contracted rate $122 → $121/hr, roles created 305.0K → 313.9K). Every disclosed number is talent-side or marketing. Its buyer-side take-rate has never appeared on any surface across the entire record. |
| Window tally | Aug 2026 | Recomputed at 380 companies on the 2026-07-31 to 2026-08-16 window: 82 customer-facing events (packaging + price-change + deprecation) from 67 distinct movers, attributed by the mover's sales_motion tag — 70 plg-tagged, 7 sales-led+self-serve without plg, 5 sales-led-only, 0 self-serve-only. Raw ratio 14:1, up from 10.5:1 at the prior window. Normalized: plg-tagged companies are 238/380 (62.6%) and produced 85.4% of events, a 1.36x over-index; sales-led-only companies are 78/380 (20.5%) and produced 6.1%, an index of 0.30 — a 3.3x under-index. Event rate per company differs by roughly 4.5x, widened from 3.4x. Both the raw ratio and the normalized gap moved AWAY from convergence, which is the direction the published kill condition said would falsify the trend if reversed. |
| Glean | Aug 2026 | The leak channel escalated from meters to dollar rates to rate CUTS, at a vendor whose own pricing page still 301-redirects to its homepage. Glean carries price_transparency gated and a sales-led-only motion, and it produced two of the window's five sales-led-only events — both price-changes. On 2026-08-04 its Model Hub Usage table cut published dollar rates for the first time since Glean began disclosing them in July 2026: GPT 5.6 Terra input $2.50 to $2.00, cache write $3.125 to $2.50, cache read $0.25 to $0.20, output $15.00 to $12.00 (-20%); GPT 5.6 Luna input $1.00 to $0.20, cache write $1.25 to $0.25, cache read $0.10 to $0.02, output $6.00 to $1.20 (-80%). Luna was simultaneously reclassified from Premium to Standard in the Enterprise Flex tier table, moving it inside the included 100/user/week Thinking and Adaptive Reasoning allowance instead of always billing FlexCredits. On 2026-08-11 the same card was expanded rather than repriced: GPT Realtime 1.5, 2 and 2.1 each split from a single ambiguous row (showing only $5.00 input and no output rate) into three modality-specific lines — Audio $32.00 in / $64.00 out, Text $4.00 in / $16.00-$24.00 out, Image $5.00 in — plus new rows for GPT-4o Mini TTS ($0.60 text in / $12.00 audio out) and Deepgram Nova-3 Multilingual ($0.0117/min). The table's own last-updated stamp moved 7/30 to 8/5/2026. A fully gated vendor publishing dollar figures, then cuts, then previously undisclosed output rates. |
| Mercor | Aug 2026 | The invisibility mechanism at its clearest yet: six quantified VALUE claims in two days and still no price. Mercor rebuilt its Enterprise page on 2026-08-11 — deleting the $10B valuation / $2B+ revenue run rate / 100k+ contractors / 400+ employees scale-stats bar and the self-referential AI-support case study, renaming Agent Benchmarking to Agent Optimization, and reframing the offering as Discover to Deploy to Improve to Monetize — then doubled the proof section again on 2026-08-12, from three named client engagements to six. Every one is quantified: $7-9M of upside run-rate savings identified in a private-equity pre-deal Value Creation Assessment, $20M of annual run-rate impact identified by 2031 plus roughly $4M near-term in a second, a 50%+ reduction in median time to mitigation in a technology SRE engagement, 25 candidate AI workflows identified across seven business functions, 132 expert-graded agent runs across real-world HR tasks, and roughly 12,000 support requests handled per week internally. Two page rebuilds, six dated outcome numbers, and Mercor's buyer-side take-rate has still never appeared on any surface across the entire record. |
| Glean | Jul 2026 | A gated vendor publishing a rate card — the second-strongest sales-led disclosure this window. Glean (price_transparency gated, sales-led-only motion) expanded its Enterprise Flex rate card to 13 metered capabilities. It does not publish a dollar-denominated ladder, but it publishes the DIMENSIONS it charges on, which is the part a buyer needs to model. Sales-led vendors disclose meters more readily than prices. |
| Lorikeet | Jul 2026 | The boundary case that shows GTM tag and transparency are separate axes: Lorikeet carries a sales-led-only motion yet price_transparency public, and it published a real mechanic change — capping its 1.50-credit (Start) / 1.20-credit (Scale) voice-resolution rate at resolutions 'up to 3 minutes long', with no list price moved. Sales-led motion does not entail a hidden rate card; 3 of the corpus's 73 sales-led-only companies publish prices. |
Counterexamples
- Anthropic · Jun 2026 — Dual-motion frontier vendor with a heavy enterprise sales motion, yet it ships frequent public SKU changes (Fast mode, Managed Agents) — because its API list prices are public. Sales-led posture does not guarantee a frozen page when the product is also self-serve.
- Sourcegraph Cody · Jun 2026 — Now sales-led/enterprise — and it did change, retiring self-serve tiers into one Enterprise plan. Sales-led vendors are not static; their changes are rarer and structural (a posture shift), not iterative price tuning.
- Observe.AI · Jul 2026 — The counterexample this trend said would falsify it, and it arrived from the most gated possible source. Observe.AI carries price_transparency sales-only and a sales-led-only motion; its own observe.ai/pricing returns 404. Yet its AWS Marketplace listing for VoiceAI Agents publishes a fully metered prepaid contract on two dimensions — $4.80 per Minutes-per-Month unit and $12.00 per Interactions unit for 12 months ($9.60 and $24.00 for 24 months, exactly double, so no posted commitment discount) — with NO per-agent seat dimension, even though its direct motion is reported to sell per-agent seats. Actual buyer-side unit prices from a sales-only vendor, dated and public. The correction this forces is narrow but real: sales-led pricing is not invisible, it is just not on the pricing page. The channel is the hyperscaler marketplace listing. Source: changes/observe-ai-2026-07-22-marketplace-minute-metering.md.
- Abridge · Jul 2026 — Sales-only structure disclosed without a price — the partial form of the same leak. Abridge named an 'unlimited enterprise agreement' alongside its per-clinician license and consolidated its product lines from four to three. A buyer learns the CONTRACT SHAPE (there is an unlimited option) without learning any number, which is the characteristic sales-led disclosure: structure yes, price no. Inflection AI did the negative version on the same date, retiring its Enterprise offering entirely to a 404.
Trivia
-
Of the corpus's logged packaging / price-change / deprecation events with a known go-to-market motion, 41 came from PLG vendors and just 2 from purely sales-led ones — an ~20:1 gap in *visible* SKU churn. (Recounted 2026-07-30 on a larger window: 84 PLG vs 8 sales-led-only across 101 events, so the raw ratio is closer to 10:1, and normalized per company the gap is ~3.4x.)
-
The four serial repricers in the corpus — Perplexity (3 public ladder changes), HeyGen, WellSaid, and You.com (2 each) — are all PLG / self-serve. No purely sales-led vendor changed its public ladder more than once.
-
Anthropic is the exception that proves the visibility rule: it shipped two new meters (Fast mode and Managed Agents) on 2026-06-15 — its churn shows up precisely because, as a frontier API, its list prices are public; an enterprise-contract peer making the same move would leave no trace.
-
Mercor's 2026-06-30 change is the corpus's cleanest demonstration that sales-led repricing is invisible: the only number it moved publicly was a *talent-side* headline (average rate $141→$80/hr) — its actual buyer-side take-rate has never appeared on any surface across the entire record. A sales-gated vendor's most visible act in a repricing-heavy week was relabeling a marketing stat, while ten PLG peers rewrote their actual prices in the open.
-
Across three straight capture batches (2026-06-24, 06-30, 07-06) every single logged public price or packaging change traced to a public-pricing GTM motion — 0 came from a genuinely sales-gated vendor's actual price. The only 07-06 sales-led event was Flexprice pushing a self-serve tier INTO the 'Contact Us' gate: the moment a price disappears from the feed is itself a PLG-to-sales boundary crossing, not a sales-led vendor becoming newly visible. (The streak ended at three: on 2026-07-22 the sales-only Observe.AI published real buyer-side unit prices — on AWS Marketplace, not its own page.)
-
The fourth window broke that three-batch streak, and the vendor that broke it is the most gated kind there is. Observe.AI carries price_transparency sales-only and its own observe.ai/pricing returns 404 — yet on 2026-07-22 its AWS Marketplace listing published $4.80 per Minutes-per-Month unit and $12.00 per Interactions unit on a 12-month contract, with no seat dimension at all. A sales-only vendor's actual buyer-side unit prices became readable; they simply were not on its website. The channel through which sales-led pricing leaks is the hyperscaler marketplace, not the pricing page.
-
Stated as a raw ratio the gap looks like 10.5:1 (84 PLG events vs 8 sales-led-only across 101 customer-facing events in 15 days). Normalized per company it is far smaller: PLG-tagged companies are 63.5% of the corpus and produced 83.2% of events (a 1.31x over-index), while sales-led-only companies are 20.7% of the corpus and produced 7.9% (a 2.6x under-index) — so the real difference in event rate per company is about 3.4x, not 20x. The original 20:1 figure was a corpus-composition artifact as much as a behavioural one.
-
Mercor remains the trend's purest specimen across two more events. On 2026-07-21 it renamed three enterprise product lines site-wide (Business / Enterprise evals / Data partnerships to Enterprise agents / Human data / Data monetization) and grew its data-integration count from 34+ to 50+; on 2026-07-29 it disclosed a $100K company-referral bonus cap on its Data page. Both are public, dated, and neither reveals a single thing about what a buyer pays — its buyer-side take-rate has never appeared on any surface in the entire record.
For buyers
Don't read a PLG tool's visibly churning pricing page as 'unstable' next to an enterprise vendor's 'stable' one — the enterprise price is just as movable, you negotiate it instead of reading it. If you buy PLG, expect your plan to be restructured within ~12 months and ask for grandfathering. If you buy sales-led, the page tells you little; your leverage is at renewal, so lock rate and ramp terms into the contract.
For vendors
If you're PLG, your pricing is your most public, most-iterated surface — instrument it, but change it legibly, because every visible move is watched and can spark backlash. If you're sales-led, you have cover to reprice privately, but that same opacity is a buyer objection: be ready to justify a rate that no public anchor supports.
Outlook — what to watch
Logged new in June 2026. The split holds as a cross-sectional fact today; what would sharpen it is buyer-reported renewal data showing sales-led vendors reprice as often as PLG vendors once you can see inside the contract — confirming the gap is mostly visibility. It weakens if the PLG concentration turns out to be a sampling effect of a corpus that over-indexes on product-led companies.
Bottom line
Of corpus SKU changes with a known motion, 41 came from PLG vendors and 2 from purely sales-led ones — public pricing churn is a PLG phenomenon. Sales-led vendors reprice too; they just do it invisibly, inside contracts.
FAQ
Do product-led companies change prices more often than sales-led ones?
More visibly, yes — in the corpus, 41 of ~49 attributable SKU changes came from PLG vendors and only 2 from purely sales-led ones. But sales-led vendors reprice too; they do it one contract at a time at renewal, so it never shows up on a public page.
Should I avoid a tool whose pricing page keeps changing?
No. A churning PLG page isn't less stable than a static enterprise one — the enterprise price is equally movable, you just negotiate it rather than read it. Focus on contract terms: grandfathering and rate-lock matter more than how often the page changes.
Why can't I see sales-led pricing changes?
Because they happen inside individual contracts under NDA, at renewal. There's no public rate card to update, so a sales-led vendor can reprice continuously without any visible trace — which is exactly why a public change feed shows almost none of it.