Learn
The honest answer: from $0 to six figures a year — the spread has almost nothing to do with features. Two pricing cultures divide this market. One publishes prices, bills on usage, and lets you start without talking to anyone. The other hides prices behind quotes, bills on visitor counts, and sells through enterprise contracts. The same workload can cost $0 on one tool and a reported $36,000+ a year on another.
Most A/B testing tools bill you one of two ways — on events or on tracked visitors — and which one matters more than any feature list: it decides how your bill behaves as you grow.
Event-volume pricing charges for the data you actually send: every variant assignment (an exposure) plus every tracked goal or custom event. More experiments and traffic mean more events, and the invoice moves with them; a quiet month is a quiet bill. The vendors on this model are the exception — Statsig, ABsmartly, GrowthBook's managed tier, and ProveBetter — but it is spreading, because the vendor's cost of serving you aligns with what you pay: your bill grows only when your experimentation volume grows, and no tier ceiling forces a jump for one busy week.
Per-visitor pricing (MTU, or monthly tracked users — also MUU/MAU) charges for the unique visitors who enter your campaigns in a billing month, capped in tiers: 10K users, 50K users, 100K users. VWO is the clearest example, with Convert and Kameleoon on the same model. VWO counts a visitor once regardless of how many tests they enter, and there is no quota rollover — a traffic spike can push you into the next tier mid-month. The price is set by audience size, not usage: you can run very few experiments and still pay for every visitor you own. And the market's loudest documented complaint is exactly this — MTU bills that punish growth, with inflexible caps and surprise upgrade pressure.
A third factor compounds both models: feature gating — a published entry price is rarely the whole price. VWO prices its Insights, Personalize, and Rollouts modules separately and gates key features behind higher tiers; Optimizely gates warehouse-native analytics to paid plans; several tools hide per-module costs until after signup.
Representative published entry points, from market research dated mid-2026; third-party-reported figures are marked †.
| Tool | Billing unit | Entry price | Free tier |
|---|---|---|---|
| Optimizely | Not published (custom) | Quote-based, enterprise contracts; third-party analyses report ~$36,000/yr scaling past $200,000†; ~AUD $40,000/yr minimum cited on G2† | No — free plan discontinued |
| ABsmartly | Events | €60,000/yr published floor (50M events/mo) | No — 60-day proof of value, no trial |
| VWO | MTU (monthly tracked users) | ~$314/mo at 10K MTU; ~$665/mo at 100K MTU (annual, in-app figures)† | Starter (≤50K MTU) — being discontinued; new signups get a 30-day trial |
| Statsig | Events (metered) | $0 — Developer; Pro $150/mo (5M events) | Yes — 2M events/mo, unlimited seats, no credit card |
| GrowthBook | Seats + events | $0 self-hosted OSS; cloud Pro $40/user/mo | Yes — OSS (unlimited users); cloud Starter 3 users / 1M events |
| Eppo | Custom (MTU + flags + seats) | Quote-based; third-party reports put entry at ~$12–30K/yr† | None |
Two clusters stand out: published, self-serve, $0–$400/mo (Statsig and GrowthBook prove the model works), and quote-gated, enterprise, $36K+/yr (Optimizely, ABsmartly, Eppo — real products, priced around contracts, not checkout pages). The research behind this page sums it up: the same mid-market team pays $0–$150/mo on Statsig or GrowthBook, roughly $300–$1,300/mo on VWO, Convert, or Kameleoon, and is priced out of — or quoted $36K+/yr by — the enterprise vendors.
The price on the page is rarely the price in practice; four cost classes recur in the market research:
Sales calls and procurement. Quote-gated vendors require a conversation before you get a number — Optimizely's “call us” pricing page is a documented buyer complaint — and contracts are typically multi-year. ABsmartly has no self-serve path and requires a 60-day proof of value before you buy; VWO's Growth plan typically requires an annual contract. A cost in calendar time and leverage that never appears on an invoice.
Implementation projects. Enterprise tools often ship with onboarding rather than self-serve setup. ABsmartly's onboarding runs through training rather than a signup flow; third-party estimates put implementation fees for quote-based vendors in the $5–20K range on top of the contract — the most avoidable line item for a team that just wants a snippet installed.
Per-seat and per-module add-ons. Seat-based pricing (GrowthBook's Pro at $40/user/mo) scales with headcount, not usage. Module-based pricing (VWO's Insights, Personalize, and Rollouts billed separately) means the entry plan covers only the testing module — researchers note the add-ons balloon the total. Even self-serve vendors have quirks: on Statsig, each Pro subscription applies to one project.
Tier-gated features and hidden caps. VWO's key features are gated behind higher tiers — a documented complaint in its reviews — and its plans cap stored events per monthly active user (50 on Growth, 100 on Pro, 200 on Enterprise), with 1–3 months of Insights retention depending on tier. Optimizely's warehouse-native analytics are enterprise-gated; any feature above your tier is a future cost or migration.
The pattern: quote-based vendors carry the biggest hidden costs (calls, contracts, implementation); MTU vendors carry surprise-variable costs (caps, spikes, add-ons); event-based, self-serve vendors carry the fewest, because the published price is the whole price.
Traffic volume, not feature appetite, sets your rung.
$0 — a real program, with limits. Free tiers are scarce — Google's Optimize shut down in 2023, VWO's free Starter is being discontinued, Optimizely killed its free plan — but the ones that remain are real products, not trials. Statsig's Developer tier includes 2M events/mo with unlimited seats; GrowthBook is fully open-source to self-host; ProveBetter's Free tier includes 1M events/mo with no credit card. A site with modest traffic can run genuine tests here and upgrade when volume justifies it.
~$200/mo — a working experimentation program. At this rung you get volume for several concurrent tests with meaningful sample sizes: Statsig Pro at $150/mo (5M events), ProveBetter Small at $199/mo (5M events), and VWO's entry around $314/mo at 10K MTU (annual, testing module only) sits just above it. The difference between the models shows up here: at 5M events you can test a large fraction of your audience; at 10K MTU you pay for exactly 10K tracked users.
~$400/mo — scale and headroom. ProveBetter's Medium tier is $399/mo for 25M events; Convert Pro is $420/mo annual for 100K tested users; Kameleoon starts at $495/mo for 50K MTU; VWO's Growth plan lands around $665/mo annual at 100K MTU. This band covers most mid-market traffic without watching the meter — and, on event-based tools, without a ceiling that punishes a good month.
~$2,000/mo — high-volume, self-serve. ProveBetter's Enterprise tier is $1,999/mo for 100M events. Compare the quote-gated floor: third-party reports put Optimizely at ~$36,000/yr (~$3,000/mo) and rising; ABsmartly's published floor is €60,000/yr. At this rung you are buying ~100M events of self-serve capacity — the volume enterprise vendors reserve for contracts.
ProveBetter prices on events, with every analytic feature on every tier — nothing is gated behind a higher plan:
| Plan | Price | Events/month |
|---|---|---|
| Free | $0 | 1M |
| Small | $199/mo | 5M |
| Medium | $399/mo | 25M |
| Enterprise | $1,999/mo | 100M |
The published price is the whole price: no per-seat add-ons, no modules to buy separately, no features hidden behind a tier. The Free tier needs no credit card and includes everything — the same statistics engines, builder, and reporting — within a 1M-event monthly budget. CSV/JSON and API export are included on every tier; a scheduled warehouse sync is planned for Enterprise. One platform runs client-side tests (a ~1.4 KB boot snippet plus ~4.2 KB SDK) and server-side tests over plain HTTP — 2–5 variants, both a frequentist and a Bayesian engine on every result, live traffic adjustment, future scheduling. Full numbers live on the pricing page.
The honest limitations: ProveBetter has no visual editor — variant copy is authored in the builder — and no personalization or multivariate testing. If those are requirements, a quote-based vendor like Optimizely is a legitimate purchase; see the comparison for where that's the right call.
A/B testing costs range from $0 to six figures, and the deciding factor is the billing model, not the feature set. Event-based, published pricing (Statsig, GrowthBook, ProveBetter) starts at $0 and scales with real usage; per-visitor pricing (VWO, Convert, Kameleoon) charges for your audience; quote-based enterprise vendors (Optimizely, ABsmartly, Eppo) start around $36K+/yr with procurement attached. Compare the billed unit before the price: events tools charge for what you actually run, visitor tools for the audience you own. Start on the pricing page; if you want the published-price version, run a real test on the Free tier before you talk to anyone. See the product on the homepage, or read the server-side testing guide.