Why GMAT Focus Two-Part Analysis trips up even strong Data Insights candidates

Two-Part Analysis is the shortest question type on the GMAT Focus Data Insights section, but it carries an unusual scoring weight: a candidate must select two correct answers from a small answer grid, and partial credit rules determine how the item is resolved on the official score report. For most candidates reading this for the first time, the question reads like a standard data interpretation prompt until they reach the answer grid, at which point the familiar A–E pattern disappears and a 3×2 or 4×2 matrix of options takes its place. The shape of the grid signals that two answers are required and that one wrong choice on either side of the matrix will drag the result downward in a way that single-answer data interpretation items do not.

The aim of this article is to give a working vocabulary for Two-Part Analysis (TPA), explain how the question type is constructed and scored inside the GMAT Focus edition, and walk through the patterns that recur across official material. By the end, the reader should be able to identify a TPA stem, anticipate whether the two requested values are independent or linked, and apply a triage routine that minimises the risk of a half-correct entry on the score report.

What Two-Part Analysis actually asks of the candidate

TPA items always present a short business or scientific scenario followed by an explicit instruction of the form: “Select the value that fills the first blank and the value that fills the second blank.” Unlike standard Data Sufficiency, which hides a single unknown behind two statements, TPA asks the candidate to commit to two numerical or categorical values simultaneously. In most official items the two requested values belong to the same equation or distribution, but a handful present two independent decision points within the same scenario. Recognising whether the two blanks are linked is the first decision the solver must make; in practice, around four out of five official TPA items are linked, meaning one equation governs both blanks.

The answer grid is structured so that each row corresponds to the first blank and each column to the second. A 3×2 grid, for example, offers three options for blank 1 and two for blank 2, and the solver must pick exactly one cell. The grid is purely a presentation device, but it changes the cognitive load. Instead of choosing from five isolated options, the candidate is asked to fix a pair, which doubles the surface area for error. A correct cell requires both coordinates to be right; a half-right pair is not a partial-credit row in the same way that an eliminated distractor is in other Data Insights items.

Another structural feature is the universal absence of an “either/or” trap. Every TPA item has exactly one cell that is fully correct. There is no equivalent of an all-of-the-above or a hidden symmetrical pair. This makes TPA more deterministic than many candidates initially assume: once the two values are derived, the answer is mechanical, and there is no benefit to re-reading the stem to look for a second viable pairing.

For a typical Data Insights section of 20 items, the test-taker will see approximately one TPA question. That small footprint is misleading. Because the item is constructed around two commitments, even a strong candidate can lose ground on it within a two-minute window, especially if the stem disguises a linked equation inside a verbal wrapper.

How Two-Part Analysis is scored in the GMAT Focus edition

The GMAT Focus replaced the old AWA and Integrated Reasoning scoring framework with a unified 205–805 scale across Quant, Verbal, and Data Insights, but the underlying reporting unit for individual item types is unchanged. A TPA item contributes a fixed number of scaled points to the Data Insights sub-score, and the contribution is binary at the cell level: the pair is either correct or it is not. There is no “one of the two blanks right” credit that the test-taker can bank.

This is the part most candidates misread. On a standard multiple-choice Data Interpretation item, eliminating three wrong options is itself a scoring safeguard; on TPA, eliminating half the grid still leaves a 50–50 risk between two viable cells. The penalty for a wrong second coordinate is identical to the penalty for a wrong first coordinate, because the scorer does not see the candidate’s reasoning — only the final cell. For most candidates I work with, this is where preparation drift happens: they prepare for TPA the way they prepare for single-answer items and forget that the second commitment is graded on the same scale as the first.

The GMAT Focus score report does, however, give partial-information signals. The enhanced score report breaks Data Insights performance down by question type, and TPA is reported separately from Table Analysis, Graphics Interpretation, Multi-Source Reasoning, and Data Sufficiency. If a candidate’s TPA accuracy sits below the rest of the section, the report will surface that gap with a percentage figure, which is useful for a mid-prep diagnostic. Candidates who skip TPA in their drills are typically the ones who see this percentage drop below the other Data Insights item types.

Time budgeting also interacts with scoring. Each TPA item should consume between 2.0 and 2.5 minutes for a candidate at the 75th percentile or higher on Data Insights. Going past 3 minutes on a single TPA item is a leading indicator that the candidate is re-deriving the same equation twice — once for each blank — rather than solving for both blanks in a single algebraic pass. The scoring benefit of speed is indirect, but real: faster TPA work frees up time for the longer Multi-Source Reasoning sets elsewhere in the section.

The six recurring patterns inside Two-Part Analysis stems

Although official TPA items vary in their surface stories, the underlying mathematical structures cluster into a small set. In my experience, a candidate who recognises the pattern has solved about 60 per cent of the item before reading the answer choices.

Pattern 1: Linked equation with two unknowns

The stem describes a relationship, often a percentage split, a cost decomposition, or a rate calculation, and the two blanks are two values that satisfy the same equation. The solver is expected to set up the equation once, solve for one unknown, and substitute to obtain the second. A typical example is a manufacturing question asking for both the per-unit cost of one component and the per-unit cost of another, given a total cost figure and the ratio between them. The answer grid then offers a few component-price pairs, and only one cell satisfies both the ratio and the total.

Pattern 2: Decision-and-consequence

The stem frames a manager’s choice between two options, and each blank corresponds to the outcome of choosing one branch. For instance: “If the company chooses Vendor A, the per-order cost is $X; if it chooses Vendor B, the break-even volume is Y orders.” The two requested values are not mathematically linked, but the candidate must read the scenario carefully enough to pair each option with the correct outcome. This is the only common pattern in which the two blanks are genuinely independent, and the main risk is mis-assigning the option letter to the wrong value.

Pattern 3: Constraint with two targets

Here the stem gives a single linear constraint (a budget cap, a time limit, a workforce availability) and asks for two quantities that together satisfy it. Common variants include a labour question in which hours allocated to two tasks must sum to a total, and the answer grid presents feasible hour combinations. The solver is essentially doing bounded integer arithmetic, sometimes with a profit maximisation overlay that determines which boundary is binding.

Pattern 4: Distribution split with one parameter

The stem describes a population, a sample, or a revenue pool that is split into two segments, and the two blanks are the segment sizes. A single proportion or percentage ties the two together. The risk in this pattern is the candidate computing only one segment and guessing the other, which is exactly the failure mode the two-blank structure is designed to catch.

Pattern 5: Threshold with crossover

Two quantities are equal at some boundary, and the stem asks for both the value of the boundary and the value of one of the quantities at that boundary. Algebraically the candidate is solving f(x) = g(x) and then substituting; the second blank is downstream of the first. Candidates who try to solve each blank separately usually waste time, because the second blank has no meaning without the first.

Pattern 6: Probability with two events

The stem asks for the probability of event A and the probability of event B, where the two events are not mutually exclusive and a joint probability is given. The candidate must apply the inclusion–exclusion rule or a conditional probability formula to recover both. This pattern is rarer on the official test but appears often in third-party practice banks, and it is the one pattern where candidates benefit from sketching a small Venn diagram before computing.

Setting up a decision tree before reading the answer choices

The single biggest tactical mistake I see in TPA work is jumping to the answer grid before the stem has been read twice. The grid is a magnet for early commitment. Once a candidate locks onto a row that “looks right,” the second blank gets treated as a confirmation step rather than an independent derivation. To prevent this, a disciplined solver builds a decision tree — a short, written-out logic chain — before any answer choice is even considered.

The tree has three nodes. Node 1 identifies the relationship between the two blanks: linked equation, decision-and-consequence, constraint, distribution split, threshold, or probability. Node 2 identifies the single equation or condition that governs both blanks. Node 3 identifies the first variable to solve for, which is usually the one that appears in only one equation. Once the tree is written, the solver can scan the answer grid with the same eye they would use for a Data Sufficiency stem: looking for the row and column whose intersection is consistent with the equation, and rejecting any cell where either coordinate contradicts the algebra.

This approach also defends against the “plausible half” trap. If the first blank is correctly identified but the second coordinate is not, the candidate has at least flagged the partial credit row mentally and can re-check the second derivation. Most unprepared test-takers, by contrast, lock in a half-correct cell and never re-examine the second coordinate, because the grid visually presents the pair as a single answer.

Worked walkthrough: a linked-equation TPA item

Consider a representative item. A retailer sells two product lines, A and B. The total revenue from both lines is $480,000. Line A revenue is three times line B revenue. The two blanks ask: “What is the revenue from line A? What is the revenue from line B?”

The decision tree opens with the linked-equation pattern. Let B = x; then A = 3x. The governing equation is x + 3x = 480,000, so 4x = 480,000 and x = 120,000. The first blank is therefore 120,000 (line B), and the second blank is 360,000 (line A). A 3×2 grid offers three possible values for A ($360,000, $400,000, $480,000) and two for B ($120,000, $160,000). The only consistent cell is the intersection of $360,000 and $120,000. Any other pairing fails the ratio or the total.

The trap in this item is the candidate who solves for B correctly but, seeing the grid, picks the row where A = $360,000 is correct and assumes B can be either value. The grid layout tempts the eye to treat the row as an answer, when the second coordinate is a graded commitment in its own right. A second trap is the candidate who, on spotting the ratio 3:1, guesses $360,000 and $120,000 without writing the equation. If the stem had said the total was $360,000 instead of $480,000, the guess would have been wrong but the cognitive pattern would have felt the same. The decision tree forces the algebra into the open, which makes the second trap visible.

Worked walkthrough: a decision-and-consequence TPA item

Now consider a contrasting item. A logistics manager must choose between two shipping contracts. Contract X charges a flat fee of $2,000 plus $50 per shipment. Contract Y charges a flat fee of $3,500 plus $20 per shipment. The two blanks ask: “If total shipping volume will be 30 shipments, the cost under the optimal contract is $________. The break-even shipment count between the two contracts is ________.”

The decision tree starts with the decision-and-consequence pattern, but the twist is that the two requested values are linked: the optimal contract at 30 shipments depends on which side of the break-even the volume lies. The governing condition is 2,000 + 50n = 3,500 + 20n, which gives 30n = 1,500 and n = 50. So the break-even count is 50 shipments. At 30 shipments, Contract X is cheaper: 2,000 + 50 × 30 = 3,500. The grid offers 3,500 and 4,100 as possible cost values, and 30 and 50 as possible break-even values. The only consistent cell pairs 3,500 (cost) with 50 (break-even). Note the trap: if the candidate pairs 3,500 (cost) with 30 (break-even), the second blank is wrong, even though 3,500 happens to match the cost computation at 30 shipments. The break-even and the chosen volume are different concepts, and the grid does not label them.

This is the pattern that punishes candidates who skim. Reading the stem for two minutes and writing the two variables down before touching the answer grid prevents the mis-pairing that the grid layout would otherwise encourage.

Common pitfalls and how to avoid them

Across several hundred TPA items in our practice bank, the same small set of errors accounts for the majority of wrong answers. The first is single-equation overconfidence, where the candidate derives the first blank, scans the grid for a row that contains the right value, and selects the row without re-checking the column. The remedy is to treat the second blank as a separate graded commitment and to verify it with the same algebraic care as the first.

The second is pattern misclassification, where the candidate assumes a linked equation when the stem is actually presenting a decision-and-consequence item, or vice versa. The remedy is the three-node decision tree described above; the cost of writing the tree is ten seconds, and the cost of misclassifying a pattern is often two minutes of re-work.

The third is unit confusion. Several TPA items mix dollars and thousands of dollars, or shipments and dozens of shipments, in the same stem. A candidate who reads quickly can place a value in the wrong blank because the magnitude is plausible for one but not the other. The remedy is to label units next to each blank before solving.

The fourth is grid misreading. A 4×2 grid has eight cells, not eight answer choices, and a candidate who treats each cell as a standalone option will systematically over-count. The remedy is to read the grid header carefully and confirm which axis corresponds to which blank before selecting.

The fifth is time over-investment. Because TPA items are short, candidates sometimes spend four minutes on a single item, hoping to convert uncertainty into certainty. The opportunity cost is real: an extra minute on TPA is a minute lost elsewhere in Data Insights where the scoring return is often higher. A practical rule is to cap TPA work at 2.5 minutes; if the second blank is still uncertain, the candidate should commit to the most consistent cell and move on.

Comparative reference: TPA versus the other Data Insights item types

TPA is best understood by contrast with the other item families in the section. The table below summarises the structural and scoring differences that matter for preparation planning.

The table makes one point clearly: TPA is unique in requiring two graded commitments from a single prompt. Every other Data Insights item type requires one commitment, and the candidate can rely on process-of-elimination as a partial safeguard. TPA requires the candidate to push past elimination and lock in a specific cell. Preparation that treats TPA as a faster version of the other item types will under-prepare for that moment of double commitment.

Building a TPA preparation plan inside a broader GMAT Focus schedule

For most candidates, TPA is best drilled in short, focused sets of 8 to 10 items, not in marathon blocks of 30. The reason is attentional. A 10-item TPA set fits inside 25 minutes, which is the natural attention window for a single item family, and it produces a clean signal about accuracy and pacing. Marathon sets blur the patterns and encourage guessing as a fatigue response.

Within a broader GMAT Focus schedule, TPA should be sequenced after the candidate is comfortable with the linked-equation pattern in standard Data Interpretation work. If the candidate cannot yet solve a two-variable linear system in under 90 seconds on a single-answer item, the TPA pattern will compound that weakness. Conversely, a candidate who can solve a two-variable system confidently will find TPA a natural extension, and the marginal preparation time required is small.

Diagnostic logic: a candidate whose TPA accuracy lags the other Data Insights item types by more than 15 percentage points should return to pattern-recognition drills rather than to timed mixed sets. Pattern recognition is the cheapest intervention; timing improvements follow once the patterns are second nature. A candidate whose TPA accuracy is within 5 percentage points of the rest of the section but whose pacing is slow should focus on collapsing the two derivations into one, by writing a single equation that solves for both blanks simultaneously.

Test-day positioning matters as well. Most candidates encounter TPA early or mid-section, and the item is rarely the last in the section. The practical consequence is that a stuck candidate cannot simply “save it for the end.” A reasonable rule is to spend 90 seconds on the stem, 30 seconds on the decision tree, 30 seconds on the algebra, and 30 seconds on grid verification. If the cell is not locked in by 2.5 minutes, the candidate should commit to the most consistent cell and move on; the section-level scoring reward for a clean guess is higher than the reward for burning 90 seconds on a single item.

The takeaway is straightforward. Two-Part Analysis is short, structured, and pattern-driven, and it rewards a candidate who reads the stem twice, writes a three-node decision tree, and treats each blank as an independent graded commitment. Candidates who prepare for it as a single-answer item will be tripped by the grid; candidates who prepare for it as a two-answer item will find the pattern recognition work transfers naturally from the rest of Data Insights. TestPrep İstanbul's diagnostic assessment is a natural starting point for candidates building a sharper Two-Part Analysis preparation plan.

Frequently asked questions