This audit is for the Concorso SSM candidate using SynapSSM who wants to know whether its statistics can be trusted as a readiness signal. In short: SynapSSM is a large, low-cost, mobile-first question app with spaced repetition and weakness analytics, and it is a strong option for high-volume retrieval. Its principal limitation is analytical: its marketing invokes AI and adaptive learning, but the metrics it substantiates are basic — so its home-screen average is not, on its own, a ranking-ready signal.
What SynapSSM offers for the Concorso SSM right now
Every figure below is vendor-reported from synapssm.com and was last checked on 20 July 2026. Verify the live count and price on the product page, and note that the "AI algorithms" language appears in site metadata rather than as a documented, adaptive-difficulty feature.
| Attribute | Vendor-reported position (synapssm.com, 20 July 2026) |
|---|---|
| Question pool | 50,000+ quiz, described as continuously updated |
| Free plan | 45 quiz/day; 1 simulation/day; a competitive simulation |
| Pro plan | Unlimited quiz and simulations; cancel anytime |
| Price | €4.15/month (a 60% discount from €9.99/month) |
| Retrieval method | Spaced repetition; "learn from your errors" (impara dagli errori) |
| Analytics | Basic statistics (free); advanced statistics (Pro); weakness analysis; progress monitoring |
| Adaptive / AI | "AI algorithms" and a "data-based method" claimed in metadata; not documented as adaptive difficulty |
| Not confirmed on site | Percentile, predicted score, item difficulty levels, time-per-item analytics |
| Platform | Mobile-optimised |
The honest read is that SynapSSM confirms spaced repetition, weakness analysis and basic-versus-advanced statistics — but not the richer analytics (percentile, difficulty, predicted score, time-per-item) that a full readiness dashboard would show. Treat the missing metrics as "verify," not "present."
The exam these analytics must serve
The Concorso SSM is a single computer-based test of 140 multiple-choice questions, five options with one correct answer, in 210 minutes, scored +1 for correct, −0.25 for wrong and 0 for blank. The written maximum of 140 points plus up to 7 for titoli gives a ceiling of 147, and a single national ranking (graduatoria nazionale) run by the Ministero dell'Università e della Ricerca decides placement — rank against the field, not a fixed pass mark. The 2026 sitting is on 21 July 2026.
For an analytics audit, two facts are decisive. There is no official weighted syllabus — the bando names the whole Corso di Laurea Magistrale and the specialty disciplines without publishing a domain weighting — so any "coverage" chart is measured against past-paper subject mixes, not an official blueprint. And the Ministry publishes no pre-exam question database; the exam is unseen every year. That is why a bank's internal percentage, however large the pool, is only ever a proxy for performance on questions you have not met.
Every metric, defined — and which SynapSSM confirms
An analytics layer is only as good as the metric definitions behind it. Here is what each metric means and its status on SynapSSM.
| Metric | What it means | Status on SynapSSM (20 July 2026) |
|---|---|---|
| First-attempt accuracy | Correct on the first exposure to an item | The number to trust; derivable but not clearly isolated in-app |
| Repeat accuracy | Correct on items you have already seen | Inflated by memory; drives the spaced-repetition queue |
| Coverage | Share of each domain you have attempted | Weakness analysis implies it; verify it is per-domain |
| Difficulty | Item-level difficulty calibration | Not documented on the site — do not assume it |
| Percentile / ranking | Your position versus other users | A "competitive simulation" exists; a true percentile is not documented |
| Predicted score | A forecast of exam or rank outcome | Not offered — and it should not be trusted if it were |
| Time per item | Seconds per question vs the ~90s exam pace | Not documented — track it manually against the clock |
The single most important distinction is first-attempt versus repeat accuracy. A rising home-screen average driven by repeat accuracy on a spaced-repetition queue can look like progress while first-attempt performance on new items stands still.
Selection bias — why the feed flatters you
Any error-driven or spaced-repetition feed deliberately re-serves items you got wrong or found hard. That is good for learning and misleading for measurement: your visible accuracy is computed over a set the algorithm has skewed towards material you have already been coached on. The number therefore rises partly because you are re-seeing familiar items, not only because you are getting better. This is precisely why a bank percentage is not comparable with a mixed, unseen block — a point set out in full in the percentage article. SynapSSM's "learn from errors" design makes this bias a feature, so read its headline average as a study aid, never as a ranking-ready score. No proprietary-algorithm claims are needed to see the effect — it follows from re-serving missed items.
Blueprint audit — attempted distribution versus the field
Do not trust the home-screen average until you have audited what it averages over. Export or note your attempted-question distribution by subject and compare it with the subject mix of recent official papers rather than the app's default feed. Because there is no published ministerial weighting, this comparison uses past-paper proportions as the reference. The common finding is that an error-driven feed over-samples two or three weak domains and under-samples everything you are already reasonable at — so your average is dominated by a narrow slice of the curriculum. Build a simple coverage matrix (subject on rows, attempted count and first-attempt accuracy in columns) to see the gaps the headline number hides; the completion-is-not-coverage method sets out how.
Readiness test — the conditions for a credible signal
A number is only a readiness signal if it is generated under exam-like conditions. Require all five: the items are unseen (not from your spaced-repetition queue), the block is timed at roughly 90 seconds an item, the content is mixed across the whole curriculum rather than a single weak domain, you take it with no assistance (no pausing to look things up), and the sample is large enough — a full 140-item simulation, not a 15-item burst — to be stable. SynapSSM's daily simulation can meet the timed and mixed conditions; the unseen condition is the hard one, because much of what the app serves you has been curated by your own error history. That is the specific job an independent bank does well.
Algorithm override rules
Because an error-driven feed optimises for your recent mistakes, it will not reliably surface everything the exam samples. Override it deliberately. Force in the low-frequency, high-consequence material the feed under-serves: statistics and epidemiology, medical ethics and legal medicine, pharmacology calculations, and image and lab-data interpretation. Force periodic full-curriculum mixed blocks even while a domain still feels weak, so you rehearse switching between topics under time. And force unseen items on a fixed cadence, so the readiness signal is never computed purely over material the algorithm has already coached. The rule of thumb: let the feed drive learning, but you drive coverage and measurement.
Worked dashboard example — turning analytics into quotas
Take a hypothetical week of SynapSSM analytics and convert it into next week's quotas — without inventing a pass prediction.
| Domain | Attempted | First-attempt accuracy | Read | Next-week quota |
|---|---|---|---|---|
| Internal medicine | 380 | 74% | Well covered, solid | 40 (maintenance) |
| Surgery | 250 | 69% | Covered, mid | 60 (mixed) |
| Pharmacology | 90 | 58% | Under-covered, weak | 80 + calculations |
| Statistics / epidemiology | 25 | 55% | Neglected, weak | 60 (forced) |
| Ethics / legal medicine | 15 | — | Barely touched | 40 (forced) |
The dashboard does not say "you will place in the graduatoria"; it says where first-attempt accuracy is low and where coverage is thin, and it converts that into a defensible quota. Note the two forced low-volume domains: the feed would rarely surface them on its own. No spurious pass prediction is made or implied.
Worked example — a seven-day plan
Give SynapSSM one job — run the feed and weakness analysis to generate quotas and daily volume — and give an independent bank the readiness-measurement job.
| Day | SynapSSM job (~60–90 min) | Unseen measurement (~30–40 min) |
|---|---|---|
| Mon | 60 items following the quotas above; review misses | — |
| Tue | 60 items, forced statistics + ethics blocks | — |
| Wed | Spaced-repetition review of the week's errors | 20 unseen items in iatroX, timed, mixed |
| Thu | 60 items, pharmacology + calculations focus | — |
| Fri | 45 items, image/data-interpretation set | — |
| Sat | One full 140-item simulation at 210-minute pace | — |
| Sun | Audit the coverage matrix; rebuild quotas | 40 unseen items in iatroX; compare first-attempt accuracy |
If your unseen iatroX first-attempt accuracy tracks your SynapSSM first-attempt accuracy, the app's signal is calibrated. A persistent gap means the SynapSSM number is inflated by its own feed, and the fix is more unseen volume, not more repeats.
Decision checklist — continue, supplement, switch or stop
Continue if your first-attempt accuracy on new items is rising and your coverage matrix is filling evenly. Supplement if your headline average looks strong but your unseen first-attempt accuracy lags, or if low-frequency domains stay thin — add forced blocks and an unseen bank. Switch the tool driving measurement if you cannot separate first-attempt from repeat accuracy well enough to trust the number. Stop leaning on the home-screen percentage the moment it is dominated by repeat accuracy on a curated queue; at that point it is measuring memory, not readiness. Decide on measurable gaps, not on novelty or on sunk cost.
Bottom line
SynapSSM is a large, cheap, mobile-first Concorso SSM bank with genuinely useful spaced repetition and weakness analytics, and it is a strong high-volume retrieval tool. Its analytics, though, substantiate less than the AI-and-adaptive marketing implies, and any error-driven feed inflates its headline average through selection bias. Use it to drive daily volume and to surface weak areas; audit coverage yourself, force the low-frequency domains, and confirm readiness on unseen, timed, mixed blocks before you trust any number.
Frequently asked questions
Is SynapSSM enough for Concorso SSM on its own? For high-volume retrieval, SynapSSM can carry a large share of preparation, because its pool (50,000+ quiz, vendor-reported, 20 July 2026), spaced repetition and low price suit daily practice at scale. It is less sufficient as your sole readiness gauge, because its confirmed analytics are basic and its error-driven feed biases the headline average, so it cannot on its own tell you whether your score will hold on unseen items. Treat it as strong for practice volume and pair it with independent, unseen measurement.
Which Concorso SSM component does SynapSSM not reproduce well? It does not reproduce a trustworthy, unbiased readiness signal or the national-ranking dynamic. The app measures performance over a set skewed by your own error history, and the graduatoria nazionale ranks you against the whole field plus your titoli — something a "competitive simulation" against other app users only loosely approximates. Do not read its statistics as a placement forecast; read them as a study guide.
How many SynapSSM questions should I complete per day for Concorso SSM? Anchor to the exam's own pace of 140 questions in 210 minutes rather than to the free plan's 45-a-day cap or the Pro plan's unlimited access (both vendor-reported). A sustainable pattern is 45–80 items a day with full review of every miss and deliberate weighting towards weak, low-frequency domains. Volume alone inflates repeat accuracy while first-attempt accuracy — the number that matters — stagnates, so cap raw volume in favour of reviewed, mixed, first-exposure practice.
When should I stop using SynapSSM and move to mixed mocks? Shift emphasis when your first-attempt accuracy plateaus and your daily sets are increasingly familiar items from the spaced-repetition queue — usually the final six to eight weeks. That is when full-format, unseen, timed mocks give more signal than more feed-driven drilling. Keep SynapSSM for targeted top-ups of weak domains, but let large unseen mixed blocks set your readiness judgement.
How should I combine SynapSSM with iatroX without duplicating practice? Give each a separate job: SynapSSM to drive daily volume, spaced repetition and weakness surfacing, and iatroX to provide unseen, timed, mixed items that measure whether that practice transfers. Never re-run a SynapSSM item inside iatroX to check yourself — that measures recognition, not readiness. Under the two-Q-bank rule, one bank builds volume and the other measures on fresh items, and you compare first-attempt accuracy across the two rather than pooling the totals.
Editorial notes and references
Written by Dr Kolawole Tytler, NHS GP and founder of iatroX. Last checked 20 July 2026. All SynapSSM figures — question counts, plan limits, analytics features and prices — are vendor-reported from synapssm.com on that date; the "AI algorithms" and "data-based method" language appears in site metadata and should not be read as a documented adaptive-difficulty engine, so verify the specific metrics on the product before relying on them. Disclosure: iatroX operates a competing Concorso SSM question bank, so this audit confines iatroX's role to the job SynapSSM's analytics do not reliably do — provide an unbiased, unseen readiness signal — and makes no proprietary-algorithm claims for either platform. Corrections are welcome via the feedback route on iatrox.com. References: Ministero dell'Università e della Ricerca (bando Concorso SSM 2026, mur.gov.it) for the official format and graduatoria; synapssm.com for the platform claims; and, on iatroX, the Concorso SSM bank, why a Q-bank percentage is not your exam score and the completion-is-not-coverage matrix method.