Experience Optimization Solutions

Forrester scored eleven vendors in this category in December 2024 and eight in the edition it published this August. In between it published a map of the same market that names fifteen.

Both numbers are correct, and they describe different instruments. A Wave is a shortlist of the vendors Forrester thinks a buyer should consider. A Landscape is a map of everyone worth knowing about. The distance between eight and fifteen is not a contradiction. It is Forrester saying the market has more participants than it has candidates.

That gap is the most useful thing in the 2026 edition, and it is not on the chart.

The scored field shrank and the market did not

The Q4 2024 Wave, published on 10 December 2024, evaluated what Forrester called the top 11 players. The Q3 2026 edition scores eight. Three of the names the 2024 edition carried belong to companies that have since combined or been absorbed.

SiteSpect, which took maximum scores in two criteria in 2024, was acquired by Monetate in a deal announced at the end of June 2025. The domain that used to serve its product pages now serves Monetate's. VWO and AB Tasty, both on the 2024 chart, merged and relaunched under a single brand, Wingify, in September 2026, with the combined company reporting more than 100 million dollars in revenue and four thousand customers. Dynamic Yield, which the 2024 edition also scored, is a Mastercard product line; Mastercard bought it from McDonald's.

So the field of eight is not eleven vendors minus three weak performers. It is a field from which consolidation has already removed names the previous edition spent time scoring, and the 2026 edition scored the surviving entities rather than the products it scored two years ago.

There is a version of this in which the category is simply getting more concentrated around suites, which is what happened to most marketing software. There is another in which a buyer reading a Wave from either year is reading a scorecard of products that have since changed owners. The evidence for the second is that the mechanism is visible twice in one edition, and that both transactions were large enough to be announced as strategic.

Forrester changed what it says these products do

The 2024 definition described "solutions that enable the ongoing delivery of relevant, timely, and optimized digital experiences to meet evolving customer needs by leveraging cross-channel customer interactions."

The 2026 definition describes offerings that "enable organizations to design, deliver, measure, and continuously improve digital customer experiences."

Two things moved. The verb list went from delivery to design, delivery, measurement and continuous improvement. And the subject went from the experiences to the organizations producing them.

Read side by side, the category has claimed the whole loop. In 2024 it was a delivery mechanism for decisions made elsewhere. In 2026 it is the system that makes the decisions and checks whether they worked.

That is a larger claim than a testing tool can carry, and Forrester's own companion research says why. The Experience Optimization Solutions Landscape, Q1 2026 puts the market's vendors into two clusters: experience-led suites, which connect optimization to journey orchestration and channel delivery, and analytics- and product-led platforms, which embed experimentation in analytics and product workflows. Neither cluster is defined by the ability to design an experience. The definition describes a product that the market Forrester then maps is not organized around selling.

Inside The Forrester Wave: Experience Optimization Solutions, Q3 2026

Eight providers are scored. Two are Leaders.

Adobe says it took the highest scores among the eight in both the current offering and strategy categories, on Adobe Journey Optimizer and Adobe Target inside Adobe CX Enterprise. Optimizely was named a Leader and also a Customer Favorite, the mark Forrester gives to the vendor whose reference customers were most positive.

That is a change at the top of the chart. Optimizely topped the 2024 edition and took maximum scores in sixteen current offering criteria, including campaign design, techniques of online testing, techniques of personalization and collaboration. In 2026 it is a Leader and holds the customer feedback distinction, but the top scores in both categories belong to Adobe.

The criteria the two Leaders name are where the edition shows its age. Optimizely reports the highest possible scores in vision, innovation, roadmap and partner ecosystem, and in generative AI, agentic AI, web experimentation and feature experimentation. Adobe's list runs through data collection; data ingestion, export and zero-copy; identity resolution and privacy; customer profile; analytics; web personalization; 1:1 personalization; customer engagement; devices and channels; agentic AI; performance analysis; and organization type.

Set that against what the 2024 Leaders named. Optimizely's maximum scores were in campaign design, testing technique, personalization technique and collaboration. SiteSpect's were in implementation and deployment and in techniques of online testing.

The 2024 list is about testing craft. The 2026 list is about data architecture, identity and agency. Generative AI and agentic AI now appear as separate criteria, where the 2024 report treated generative AI as one theme among three, alongside data integration and strategy support.

Two caveats belong here rather than at the end. The criteria above are the ones the two Leaders say they topped, not the full set, and the full set is not published. And the counts of eight and eleven come from Adobe's account of the 2026 edition and Forrester's own summary of the 2024 one, because the 2026 report body is not published in a freely readable form and does not render to any of the retrieval methods this site uses.

The problem nobody puts in a demo

Here is the structural issue underneath this category, and it is the reason a maturity story and a mediocrity story can both be true.

An experiment detects an effect only if it has enough traffic to distinguish that effect from noise. The smaller the effect, the more traffic required, and the relationship is unforgiving. Detecting a large improvement takes modest volume. Detecting the two or three percent improvements that most real optimizations produce takes a great deal.

Most organizations do not have that traffic on most pages. They run a test on a checkout flow with a few thousand weekly visitors, watch the numbers for two weeks, see one variant ahead, and ship it. The result is frequently noise, and the program accumulates a portfolio of changes that were never validated at all.

Three specific failures follow, and none of them are the platform's fault.

Underpowered tests, as above, producing confident conclusions from insufficient data.

Peeking, where someone watches the dashboard daily and stops the test when it looks significant. That practice inflates false positive rates substantially, because a random walk will cross a threshold eventually if you keep checking. Sequential testing methods exist to handle this correctly, and most programs do not use them.

And multiple comparisons, where a team tests twenty variants or slices results across fifteen segments and reports the ones that reached significance. At conventional thresholds, a portion of those will be false by construction.

The uncomfortable summary is that a mature experimentation program with poor statistical discipline produces a documented history of decisions that feel evidence-based and are not. That is arguably worse than no program, because it carries the authority of measurement.

The 2026 edition moves the constraint rather than removing it. In 2024 Forrester credited Optimizely with developing features that generate variations of experiences, and the argument worth making about it then was that removing the cost of producing variants also removes the discipline that scarcity imposed. When each variant cost design and development time, somebody had to justify testing it. That justification was a crude form of hypothesis-setting, and it kept the number of comparisons low.

In 2026 the same vendor's headline capability is Opal AI, which Forrester's summary describes as an agentic orchestration layer that lets users create experiments without coding, and which Optimizely's customers told Forrester was a game changer. Orchestration moves the constraint again, and further. Generating variants multiplied the number of comparisons. A layer that also proposes which experiments to run, configures them and reads the results multiplies them again, and it removes the pause where a hypothesis used to be written down by a person.

Multi-armed bandit allocation addresses part of this by sending traffic toward better-performing options, and most serious platforms offer it. Bandits optimize for outcome rather than for learning, which is the right trade when the goal is a better page and the wrong one when the goal is knowing why something worked so it can be applied elsewhere.

The practical consequence is that generation and orchestration raise the requirement for statistical sophistication rather than lowering it. A team that ran six tests a quarter could get away with informal practice. A team whose platform proposes and runs six hundred cannot, and the same platform makes the requirement harder to see, because the interface reports significance rather than power.

Personalization is a different kind of claim

Worth separating clearly, because the two capabilities sit in the same products and are now covered by the same definition.

An experiment answers a causal question with a controlled comparison. Version B outperformed version A for this population, and randomization means the difference is attributable to the change.

Personalization answers a predictive question. Given what is known about this individual, which version will they respond to best. That is a model, and its accuracy depends on data quality, on the stability of behavior over time, and on the assumption that patterns learned from past visitors apply to this one.

Both are legitimate. They fail differently. A bad experiment gives a wrong answer that replication can detect. A bad personalization model quietly serves worse experiences to some segments while the aggregate metric holds, because the improvements to well-modeled segments mask the degradation to poorly-modeled ones.

Forrester's own note that Optimizely integrates with its data platform so customers can power segments points at the input side of this. Behavioral data describes what someone did on a site. Contextual and declared data describes who they are, and models built on the first alone are working from a narrow window.

Why the mediocrity call can still be right

In December 2024 Forrester opened its summary of the Wave with a prediction that 2025 would be another year of customer experience mediocrity, and reported in the same piece that adding or improving digital experience was the single most common action organizations were taking to improve customer experience. The most popular remedy was digital experience improvement. The tooling for it was mature. The expected outcome was continued mediocrity.

Two years on, the tooling is better, the category has claimed more of the loop, and agents are being handed work that people used to do. None of that touches the original explanation, which is that optimization operates on a fixed surface.

Testing a checkout page makes that checkout page better. It cannot tell you that the checkout page should not exist, that the product is confusing, or that customers are irritated by something that happens after purchase and outside the digital experience entirely. Forrester's broader position on customer experience has consistently been that the differences that matter are organizational and cross-functional rather than interface-level. Optimization tooling is exceptionally good at local improvement and structurally incapable of the other kind.

The 2026 definition sharpens the point rather than softening it. A category that claims only delivery can be judged on delivery, and on that measure these products work. A category that now claims design, delivery, measurement and continuous improvement has set a bar that a program with a documented cumulative conversion lift and a flat experience score does not clear.

That is not an argument against the category, which does what it claims efficiently. It is an argument for reading its definition carefully. A product that promises to improve a defined surface is useful. A product that promises to design the surface is making a claim its customers have to be organized to support, and most of them are not.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.

A close neighbor in Forrester's own coverage is Customer Experience Platforms For Healthcare, where consumer comfort with a general AI chatbot and with their own health insurer's AI tool differ by two percentage points. That indifference is the strategic problem this category exists to address.

The related category on this site is Personalization Engines. Only one vendor on Gartner's crowded seven-Leader quadrant scored first across all three use cases at once, marketing, commerce, and service, proof the category has quietly outgrown the marketing department that used to own it alone.