Australian Superannuation Member Experience Benchmark 2026

The steepest single drop occurs entering Task 3: success falls from 53.7% to 15.7%, a 38.0-point decline, larger than the drop from Task 1 to Task 2 (13.2 points) and larger than the recovery from Task 3 to Task 5 (which actually rises 5.0 points). Insurance is where the category’s member experience breaks down hardest, and it breaks down for a specific, recurring reason: cost information sits behind a personalised quote flow, a dense PDF, or both, rather than on the page that states cover types.

Decision confidence. Ease and confidence both fall from Task 2 to Task 3 (ease 3.55 → 2.76 industry-wide; confidence 3.80 → 2.86). It is the only task in the study where both self-reported measures drop together with the success rate rather than diverging from it. On every other task, confidence sits meaningfully above what the success rate would predict: Task 1 ease 3.40 vs. confidence 3.98; Task 5 ease 3.18 vs. confidence 3.47.

Success ranges from 28.6% to 95.0%, a 66.4-point spread, the widest of any task in the study. Hostplus’s collapse traces to a specific, fixable cause: its worked example sits inside a collapsed accordion nine and a half page views deep on average, the highest page-view count recorded for any task on any fund.

Themes. Participants consistently confuse a labelled component (an admin fee, an investment fee) with the full annual total, across Rest, Hostplus, Aware Super and UniSuper alike. Where a fund pre-calculates the answer for a $50,000 balance, the task becomes fast and confident; where it requires assembling the figure from a rate table, errors multiply. Several participants expected a standalone calculator in a tools section rather than an estimate embedded in fee-and-cost content. ART’s own worked example depends on which investment option is selected, and several participants never realised this and simply reported whatever total they landed on.

Rest illustrates the category’s core problem in its purest form: a lostness score of 0.07 (the single lowest recorded anywhere in this study, meaning participants moved almost directly to the right page) paired with a success rate of just 50%. This is not a findability failure at all. It is a comprehension failure once the right page is reached, and the same shape recurs on ART, Aware Super and UniSuper, each pointing to a different visual or labelling cause.

Themes. Multiple funds surface more than one number that can plausibly be read as “the answer” on the same page: a High Growth headline, a cumulative-growth chart, a differently-branded “Balanced” product. Participants pick confidently and are frequently wrong. “Default” is rarely stated as a fact in the interface; participants inferred it, often incorrectly. ART offers two products called “Balanced” (Balanced Pool, inside its Lifecycle default, and a standalone Diversified “Balanced” option) and reports them at different dates on different pages, so participants found four different 10-year figures (8.42%, 8.5%, 8.61%, 8.69%) for what they took to be the same thing. Any of the four was accepted as correct.

This is the task where completion and success diverge most severely on every single fund, and where the category-wide gap is largest. UniSuper’s lostness of 1.04 is the highest score recorded anywhere in this study. UniSuper participants averaged 11.6 page views on this task, working through a multi-step calculator, server errors, and a search that looped back to general content.

Themes. Cover types (Death, TPD, Income Protection) are reliably findable on every fund; a specific cost almost never is. The two routes to an actual number (a personalised quote calculator requiring contact details, or a lengthy PDF disclosure statement) are both disliked, and several participants abandoned the calculator over privacy concerns rather than complete it. At least three funds show participants concluding automatic cover is free or has “no additional cost,” when premiums exist and are simply not disclosed anywhere participants could find. Quotes obtained by different participants on the same fund are frequently not comparable to each other, since they depend on age, salary and occupation rather than a published baseline.

Industry-wide, 46% of participants said personal details were requested earlier than expected, and 86 of 121 sessions (71%) reached only “thin” exploration depth before the gate. Participants are being asked to commit before they have looked at much beyond the join CTA itself.

Themes. No fund offers a short “before you start” overview (what’s needed, how long it takes, what happens after) ahead of the personal-details request. ART shows the highest rate of “requested earlier than expected” (60%) alongside one of the two highest deep-exploration rates (40%): participants who explored furthest were also most likely to feel ambushed by the timing. Ease and confidence both sit above 3.3 on every fund despite the complaint; participants aren’t confused about how to join, only about when they’re being asked to commit. Technical errors interrupted the joining flow on at least two funds, which is materially worse than a labelling problem because it can block a genuine sign-up outright.

This is the task with the widest category-wide gap between how efficiently members navigate (lostness averaging 0.40, “efficient” on the reference scale) and how rarely they arrive at the right answer (20.7%). The pattern points to a conceptual mismatch rather than a navigation problem: a still-working 62-year-old doesn’t think of their question as “retirement” (they aren’t retiring) or “early access” (they’re over 60), so they land confidently on a page about a related-but-wrong condition and stop.

Themes. The most common wrong answer across every fund is a flat “no access before 65 unless retired”: correct as a general rule, incomplete as an answer to this scenario. A second failure mode recurs on at least three funds: participants land on early-release or hardship content and mistake it for the answer, rating the task easy and confident despite being on the wrong page entirely. Where the transition-to-retirement route is stated in the same breath as the general age condition, rather than on a separate, differently-labelled page, success rises sharply. Hostplus’s 47.6% traces directly to this.

Task 3 carries the highest average lostness of any task industry-wide (0.74, “high inefficiency”), while Task 5 carries one of the lowest (0.40, “efficient”), the exact inverse of what their success rates would predict if lostness explained the outcome. Insurance is hard to reach as well as hard to answer; access-at-62 is easy to reach and hard to answer. These are different problems requiring different fixes, and treating them identically would be a mistake.

What predicts what. Task success and SUS move together at the extremes but the middle of the table breaks the pattern. SUS does not reliably predict NPS: Hostplus and ART post nearly identical SUS scores (60 and 61) with NPS 35 points apart (−67 vs. −32). Lostness predicts neither. The clearest reading: task success and SUS partially predict each other, but neither reliably predicts advocacy. Something not captured by these measures is driving Hostplus’s NPS and Rest’s comparative goodwill.