How Alma Insights actually finds your weak topics (without pretending to predict your score)

By Roupen Odabashian, MD, FRCPC · Published October 23, 2026 · 22 min read

How we write and check articles: Editorial Policy.

A student studies a wall of topic bubbles where a few weak areas are clearly highlighted.

Disclosure: Alma publishes this article and makes one of the study tools it discusses.

If you use Alma as a medical, nursing, NP or PA student, Alma Insights turns patterns in your own study activity into estimates of how you are doing on your own topics, not predictions of your exam score. Insights looks at how likely you are to remember material you have studied with spaced practice, how you have actually been answering questions, and how recently you have recalled the material.^[1]^[2] It then surfaces weak spots, fading topics and calibration problems so you can decide what to do next.

This article walks through how Alma Insights works, what “mastery” really means, how “Do this now”, “About to forget”, the confidence check and “What Alma knows about you” fit together, and what Insights is not. The goal is to help you judge when to trust the dashboard, when to be cautious, and how to use it alongside primary resources and your exam’s official guidance.

What is Alma Insights, and who should actually use it?

A student looks at a dashboard of topic circles and trends

Insights lives in the left navigation under “Insights”. At the top of the page, four buttons let you choose the period: 7 days, 30 days, 90 days or All (up to one year back). Most of the page follows this choice.

The first row shows five tiles:

Readiness – an average of your topic mastery scores when Insights has enough data.
Days to exam – based on your Study Schedule or exam date on your profile.
Topics – how many topics are Solid, Getting there or Need work.
Cards due – how many flashcards are waiting in your review queue.
Time studied – an estimate based on flashcard review time and finished Practice Patients stations where that feature is available.

Next to the tiles, a ring chart “Topics by status” shows what share of your topics fall into the four status groups: Solid, Getting there, Needs work, and Too early to judge. Clicking a group filters the topic map below so you can drill into, for example, only your weak topics.

Because Insights can only see what you do in Alma, it will underestimate topics you mostly study elsewhere. When a topic has very little Alma data behind it, the interface calls that out instead of pretending it is certain.

How does Alma decide a topic is Solid, Getting there or Needs work?

Alma labels each topic based on its “mastery” score, which is an estimate of how well you currently know that topic from Alma’s point of view. Mastery combines how well you seem to remember the topic’s material and how recently and reliably you have been recalling or answering questions on that topic. Topics with too little data stay “Too early to judge” until more evidence accumulates.

Topics themselves are created from your own material. Alma groups your flashcards and quiz questions into topics by meaning (for example “hyponatremia” or “heart failure”) based on how you titled cards and what appears in their content. Each learner’s topics are therefore different; Alma does not use a fixed syllabus. Overnight, topics are also sorted into specialties such as Internal Medicine or Psychiatry where they clearly fit, and anything that does not fit remains “Other”.

To apply a status label, Insights needs enough evidence. A topic is “Too early to judge” until it has a minimum amount of activity such as several flashcard reviews or quiz answers. Once there is enough data, Alma computes a mastery percentage and then places the topic into one of several bands from weak to strong.

Mastery score	Status label
Under 50%	Needs work
50% to under 75%	Getting there
75% or more	Solid
Very little data (1–2 answers)	Too early to judge

This “lower of memory and recall” rule matters. Spaced-repetition systems can make cards look well remembered even if you keep getting them wrong, because the schedule is projecting from past success.^[1]^[2] By tying mastery to the weaker of “can you still recall the material?” and “are you actually getting questions right lately?”, Alma avoids calling a topic Solid when you keep relearning and forgetting it.

Mastery for each topic also contributes to the Readiness tile at the top of Insights. Once Insights can judge at least three topics, Readiness becomes the weighted average of their mastery scores, where topics with more evidence (more reviews and answers) count more, up to a cap. If there are not yet three judged topics, the tile simply says how many more it needs.

What do “memory”, “understanding” and “recalled lately” actually mean?

Behind every mastery label, Alma tracks three ideas separately: how likely you are to remember material on each topic now, how often you get its questions right, and how reliably you have recalled that topic in recent practice. These come from spaced retrieval practice — actively recalling information at intervals rather than just rereading.^[1]^[2]

A brain with three dials representing different learning measures

How Alma defines memory for a topic

Alma’s Flashcards use a spaced-repetition memory model to schedule reviews: each time you grade a card (Again, Hard, Good, Easy), the system updates its estimate of how likely you are to remember that card and chooses the next interval accordingly. Spaced repetition is an application of the spacing effect: information is better retained when study is spread over time rather than crammed into one session.^[1]^[2]^[3]

How Alma defines understanding for a topic

Understanding is the probability that you would get a typical question on that topic right, estimated from your quiz answers. Alma looks at graded answers in Learning Hub quizzes whose questions are tagged to that topic, counts how many you got right, and turns that into a percentage. Topics with no quiz answers have no understanding figure.

Understanding is only part of the mastery story because you can sometimes answer a question correctly by guessing or using test-taking tricks, and sometimes miss a question even if the underlying concept is sound. Alma therefore blends this with the other signals rather than using understanding alone.

How Alma defines “recalled lately”

Recalled lately is about recency and reliability. It looks at your “recall tests” on a topic in the last 30 days and asks: when you had to recall this recently, how often did that go well?

Recall tests include:

Flashcard reviews after the first learning stage (cards you supposedly already know) contribute evidence about how well you are recalling that material, especially when you space those reviews over time.^[1]^[2]
Graded quiz answers on that topic add a second kind of evidence: how often you can successfully retrieve and use the information when tested.^[2]

A topic needs at least five recall tests of its own before Alma will compute its own recall figure. If there are fewer, Insights blends in the recall figure from the rest of that deck or quiz, and the topic panel calls this out so you know the estimate leans on context.

How does “Do this now” pick what you should study next?

A student stands where one study path is clearly highlighted

“Do this now” can pick a topic for a few different reasons, including:

Your memory of this topic is fading – its flashcard memory is dropping.
You have been missing questions on this topic – your recent quiz accuracy is poor.
This topic is one of your weaker areas – its mastery score is low relative to others.

The card also tells you enough about the suggested review to decide if it fits the chunk of time you have. Clicking “Review cards” jumps into Flashcards focused on that topic. “Ask Alma” opens the Alma chat already primed to help you with that weak area.

If nothing is clearly weak or slipping, the card simply says “Nothing stands out as weak right now.” That is not a claim that you are exam-ready; it only means the Insights rules do not see an obvious outlier among the topics with enough data.

What does “About to forget” do that normal flashcard due dates don’t?

Spacing and retrieval practice work by scheduling reviews after some forgetting has occurred and requiring you to actively recall the material, which strengthens long-term memory more than cramming or rereading.^[1]^[2] The spacing effect literature shows that study spread across sessions produces better long-term retention than the same amount of study compressed into one session.^[1]^[2] “About to forget” applies that idea to your topics by estimating how your ability to recall each one is likely to change over time without extra review.^[1]^[2]

This section:

Lists up to 12 topics whose memory is predicted to drop the most in the next seven days.
Shows memory now and in a week (for example, “80% now, 50% in a week”).
Displays a small strip of the next 30 days for each topic, all on a shared 0–100% memory scale.

By default, topics are ordered by biggest weekly drop. When Alma knows your exam date, a second option orders them by lowest projected memory on exam day, which is more useful as the exam approaches.

A topic only appears if it would lose at least three percentage points of memory in a week and has enough studied cards behind the estimate. Clicking a topic opens its topic panel, and “Review cards” sends you into Flashcards.

What is the “Is your studying working?” weekly chart actually telling you?

The “Is your studying working?” card on Insights shows a week-by-week trend of flashcard retention, quiz accuracy and overall study volume, so you can see whether your current routine is leading to better recall and more consistent practice over time rather than serving as any kind of exam prediction.^[1]^[2]

The chart covers 4, 6 or 13 weeks for the 7, 30 or 90-day views and up to a year for “All”. Each week shows:

Retention – for flashcards beyond the learning stage, the share of reviews not graded Again.
Quiz accuracy – the share of graded quiz answers marked correct.
Study events – bars for how many study actions (card reviews, quiz answers, game questions, practice stations, Teach Me sessions) you did.

Buttons let you toggle each line or bar on and off, and a “Show as table” option lists the same weeks and percentages. If your exam date falls within the chart’s timeframe, the exam week is marked.

For any given week, Insights only shows a retention or accuracy percentage if there are at least ten relevant reviews or answers for that week. Otherwise, it leaves a gap rather than overinterpreting tiny samples. This protects you from reading too much into a handful of questions.

Using this chart, you might notice patterns such as:

Retention creeping up while study volume stays steady – your repetitions are working.
Retention flat but quiz accuracy rising – you are getting better at applying knowledge even if card recall is unchanged.
Study volume dropping with retention sliding down a few weeks later – a sign your spacing is stretching too far for your retention goal.^[1]^[2]^[4]

Research on spacing and retrieval practice shows that spaced retrieval is one of the most effective ways to support long-term retention, especially when sessions are distributed over time.^[1]^[2]^[3]^[5] However, most of that research focuses on relative advantages over cramming, not on predicting performance on a specific high-stakes exam for an individual learner.^[3]^[5] Alma follows that same conservative line: Insights describes your trends; it does not turn those into a score forecast.

How does the “Does feeling sure match being right?” confidence check work?

A “Show as table” view listing counts and percentages.

To protect against noisy data, the chart:

Requires at least five answers at a given confidence level before showing a percentage.
Waits until you have at least ten answers tagged both Guess and Sure before it adds a summary sentence.

Once there is enough evidence, Insights may tell you:

You are overconfident – when you tapped Sure you were right under 70% of the time. For example, “You were right 60% of the time when you tapped Sure”.
You know more than you think – when you tapped Guess you were right 70% or more of the time.

Research on metacognition and effective learning strategies notes that students often misjudge how well they are learning, confusing the feeling of fluency with durable memory.^[2]^[5] Spacing and retrieval practice can help reset these intuitions, but only if you pay attention to the feedback. The confidence check is Alma’s way of surfacing that feedback so you can change behaviour: slowing down when you are overconfident, or trusting yourself more when you are underconfident.

What is the topic map and specialty view showing you?

The default view is a map (for 4–80 topics): one circle per topic, with size proportional to how much you have studied it (reviews and answers) and colour set by its status label. Topics that are still “Too early to judge” appear as dashed circles without percentages. Keyboard navigation is supported so you can move between circles with arrow keys and open one with Enter.

You can switch to:

List – rows of topics, weakest first, each with its status and a mastery bar. Topics still being judged show how many more answers or reviews they need.

Filter buttons (All, Solid, Getting there, Needs work, Too early to judge), with counts for each, let you home in on, say, only your weak topics. A search box “Find a topic” matches substrings in topic names.

Clicking any topic in the map, specialty view, list or related charts opens its topic panel so you can see exactly what cards, questions and trends lie behind its label.

What does the topic detail panel tell you when you click a topic?

The topic panel is the detail view that appears when you click a topic anywhere on Insights. It explains, in plain language, why a topic is weak or strong, which cards and questions are driving that, and how its mastery has changed. It also gives you one-click buttons to review cards or start an Alma chat focused on that topic.

The panel includes:

Overview – mastery, memory, recalled lately, last studied date and a short statement about how the topic moved over the last two weeks compared with the previous two (when there is enough data).
Composition – how many flashcards, quiz questions and Alma notes contribute.
Buttons – “Review cards” (opens Flashcards on that topic) and “Ask Alma” (starts a chat about it).

Tabs let you drill down further:

Cards – up to 50 flashcards, most forgotten first, never-studied last, each with its deck, “% likely remembered”, and notes like “forgotten N times” or “due now”. Each card has an “Open deck” button.
Questions – up to 50 quiz questions, missed first, with number of attempts and an “Open quiz” button that takes you to that notebook’s Quiz tab.
Recent – the latest 30 reviews and answers for that topic, with outcome, seconds taken and the confidence level you tapped.
History – up to 90 days of mastery history based on nightly snapshots, plus an option to “Show daily numbers”. This needs at least two days of data.
Alma noticed – short pattern notes where Alma has spotted a recurring confusion for that topic (only when enough consistent wrong answers make it likely).

These tabs only appear when there is data; an empty tab is hidden. For example, if you have never had a quiz question about that topic, there is no Questions tab.

The purpose of this panel is to make Insights explainable. Instead of just seeing “Needs work”, you can see that, for example, a clutch of flashcards you keep forgetting are dragging memory down, or that a series of missed quiz questions last week caused mastery to drop. This helps you choose targeted fixes: rewriting clumsy cards, re-reading a concept, or asking Alma to re-explain a subtopic.

What is “What Alma knows about you”, and how does it connect to Insights?

You can get there from the account menu at the bottom of the left rail: User Profile → “What Alma knows about you” → “View and manage”. The page is for every learner type Alma supports.

It has three main sections:

What Alma thinks you know – a list of your topics, weakest first, each with a label (Needs work, Getting there, Solid or Not enough yet to judge) and how many answers and reviews that estimate uses. This is explicitly described as an estimate based on flashcards, Learning Hub quizzes and recall, not a grade.
Practice stations – your last few marked Practice Patients stations (where that feature is available) (case presentation, marks, weakest marking area, and the checklist items where marks were lost) plus the average per marking area across your last ten stations, lowest first. Alma can use this when you ask about OSCE or clinical-skills weak spots.
Things Alma has noticed – short misconception notes like “Tends to mix up X and Y”. These only appear when at least two wrong quiz answers on the same topic show the same specific confusion. The model that writes them sees only graded quiz answers, never your chats.

You also get controls over your data:

Download my data – a JSON export with topics, notes, the learning-event log and chats.
Start over – resets the estimates and notes but keeps the underlying history; that past activity stops counting toward new estimates after the reset.
Delete my learning data – erases estimates, notes and the event log (you must type DELETE). Flashcards, notes, quiz answers and chats remain, though chats are removed from the chat history list.

When Alma answers questions in chat, it can draw on this page to say which topics look weak, point to specific cards behind them, mention “early signals” (topics with a wrong answer but not enough evidence to rank) and suggest what to study. If there is not enough activity, Alma is told to say it cannot tell yet, rather than guessing.

What Insights is not: how far you should trust the numbers

Insights can be a powerful guide, but it has clear limits. It is built only from your Alma activity, not your entire learning life, and it deliberately stops short of forecasting your exam score or offering guarantees. Knowing what it does not do helps you use it well.

Key boundaries:

Not a fixed syllabus – topics come from your cards and questions, not an official content outline. You should still read your exam blueprint and primary resources.
Not clinical or diagnostic judgement – “Alma noticed” misconception notes are tentative patterns drawn from quiz answers, not assessments of your clinical reasoning or cognitive profile.

In the wider learning-analytics world, there is active research on using learner data to predict test scores or classify outcomes like Distinction, Pass or Fail.^[6]^[7]^[8]^[9] Some systems build predictive models that try to flag at-risk students early.^[6]^[7]^[8] Alma Insights does not try to do this. It sticks to descriptive analytics (what you have done, how your topics look) and simple, interpretable estimates so that you and your supervisors remain in charge of judgements about readiness.

When you plan your study, use Insights alongside your syllabus, school guidance, trusted primary texts and your exam owner’s own materials. For example, our other guides on studying for medical school and on using AI safely for USMLE Step 1 both emphasise that AI support should be checked against primary resources, not treated as an authority. The same principle applies to Insights.

How should medical, nursing, NP and PA students use Insights in practice?

For most students, the right way to use Insights is as a compass and early-warning system: let it surface weak or fading topics and calibration blind spots, but keep final decisions anchored to your curriculum and exam requirements. Use its “Do this now”, topic map and About to forget sections to pick daily work, then cross-check against your own priorities.

Some practical patterns that tend to work:

Daily – Open Insights, glance at “Do this now” and “About to forget”, and choose one or two topics to review through Flashcards. Check the streak and Time studied tile for basic accountability.
Weekly – Look at “Is your studying working?” and “Improving and slipping”. If retention or accuracy is trending down, or a key topic is sliding, adjust your schedule or card quality.
Every few weeks – Review the topic map and specialty view. Are there big red (Needs work) circles in core systems (like cardiovascular or renal)? Are some specialties neglected? Consider adding new cards or planning focused blocks.
Before practice exams – Use Insights to pick weak-but-important topics to shore up, then verify with targeted practice questions from primary Qbanks and your exam’s own materials.
After changes – If you “Start over” in What Alma knows about you or switch study style, give Insights time to build new evidence before relying heavily on its labels.

Always remember: Insights evaluates only what you have done inside Alma. If a topic matters on your exam but is missing or underrepresented in Alma, add it to your flashcards or quizzes, or it will never show up accurately on the dashboard.

Where does AI fit here, and what should you double-check?

Alma uses AI inside Insights to group topics, generate “Alma noticed” misconception notes and power the “Ask Alma” chat. AI is strong at turning raw study traces into patterns, practice questions and explanations, but it is not a substitute for your exam’s official content or for clinical guidelines. You should treat any AI suggestion as a starting point and verify clinical or exam-relevant details against trusted primary resources.

When Insights nudges you to work on a topic, you can use Alma’s AI to:

Explain a concept from first principles.
Ask Alma to quiz you conversationally on the topic. A quiz inside the chat is not saved or graded into Insights, so use a Learning Hub quiz when you want results to count toward your mastery estimate.
Summarise your weak subtopics or compare similar diseases.

At the same time, when AI-generated content touches real patients, management decisions or high-stakes exam facts, compare it with your textbooks, question banks, guidelines and your faculty’s teaching. Insights itself is descriptive and rule-based, and the AI that writes misconception notes never sees your chats, but the explanations you get via “Ask Alma” are still AI answers, and the responsibility to cross-check them stays with you.

Frequently asked questions

Does Alma Insights predict my USMLE, NCLEX, PANCE or NP exam score?

No. Insights does not predict scores, pass chances or percentiles for any exam. The Readiness tile is simply the average mastery of your judged topics, based on your own flashcards and quizzes inside Alma. It is meant as a study estimate, not as a guarantee or forecast. You should always make exam decisions based on your school’s guidance, official practice materials and trusted primary resources.

How much do I need to study in Alma before the labels become trustworthy?

A topic is only judged once it has at least three flashcard reviews or quiz answers; with one or two data points it stays “Too early to judge”. “Recalled lately” needs at least five recall tests before using that topic’s own data, and the About to forget forecast requires enough actively studied cards to project memory. In practice, that means you need a few sessions of spaced retrieval per topic before the labels are stable. Until then, treat them as tentative.

Why does Insights say a topic is weak when I feel like I know it?

Several things can cause this mismatch. You may be overconfident – the confidence chart often shows that people feel sure but are wrong more often than they think.^[2]^[5] You may also be mostly studying that topic outside Alma, leaving little evidence in your Alma data. Or a few recent wrong quiz answers may have pulled “recalled lately” down. Use the topic panel to see which cards and questions are driving the low mastery, then decide whether the label reflects a real gap or just missing data.

How should I balance Insights with my question bank and textbooks?

Use Insights as a guide to where your Alma-based learning is strong or weak, then cross-check with primary resources. If Insights flags “heart failure” as Needs work and your Qbank results on heart failure questions are also poor, that is a strong signal to remediate. If Insights calls a topic weak but you rarely study it in Alma and do well on it in your main Qbank, consider adding more Alma cards if it is important, or de-prioritising it in Alma. The dashboard is most helpful when it reflects the material that truly matters for your exam.