{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-07-14T10:23:43.470Z"},"content":[{"type":"documentation","id":"297708cf-ca9d-4a0e-be61-609fcb0bd8b7","slug":"importance-performance-analysis-guide","title":"Importance-Performance Analysis (IPA): The Priority Matrix Guide (2026)","url":"https://www.koji.so/docs/importance-performance-analysis-guide","summary":"Importance-Performance Analysis (IPA) plots how important each attribute is to customers against how well you perform on it, producing a four-quadrant matrix (Concentrate Here, Keep Up the Good Work, Low Priority, Possible Overkill) that turns satisfaction data into a clear prioritization decision.","content":"Importance-Performance Analysis (IPA) is a prioritization technique that plots how important each attribute is to your customers against how well you actually perform on it, then reads the result as a four-quadrant priority map. Rather than reporting a flat satisfaction average, IPA answers the question every team actually needs answered: of everything we could improve, which few things will move the needle most? Attributes that are highly important but where performance lags fall into the \"Concentrate Here\" quadrant, your fix-first list. Attributes that customers barely care about but where you pour in effort fall into \"Possible Overkill,\" resources you could redeploy.\n\nThis guide explains where IPA comes from, how to read all four quadrants, the crucial difference between stated and derived importance, a six-step process to run one, the pitfalls that quietly ruin the analysis, and how AI-moderated research fills the matrix in faster and with the reasoning attached.\n\n## Why a Satisfaction Average Is Not Enough\n\nMost feedback programs stop at a number: a CSAT of 4.2, an NPS of 38, a support rating of 8.1. Those numbers tell you the temperature but not the treatment. They cannot tell you whether to fix onboarding, speed up support, or add a feature, because they collapse every attribute into one figure. IPA exists to reopen that average and rank the drivers behind it.\n\nThe gap it targets is real and expensive. Bain and Company famously found that 80 percent of companies believed they delivered a superior experience, while only 8 percent of their customers agreed, a delivery gap that persists precisely because teams improve the wrong things. And the payoff for fixing the right things is large: the classic Harvard Business Review analysis by Reichheld and Sasser (1990) showed that a 5 percent increase in customer retention can raise profits by 25 to 95 percent. IPA is a simple way to point your limited improvement budget at the attributes most likely to protect that retention.\n\n## Where IPA Came From\n\nImportance-Performance Analysis was introduced by John Martilla and John James in a 1977 article in the Journal of Marketing, based on a study of automobile dealer service. Their argument was practical rather than theoretical. Measuring performance alone, they noted, \"leaves a problem in translating the results of research into marketing action.\" By adding an importance dimension and crossing the two, they gave managers a picture that mapped directly onto decisions. Nearly fifty years later, IPA remains one of the most widely applied prioritization tools in marketing, and it has become a staple in service, hospitality, tourism, healthcare, and product research.\n\n## The Four Quadrants\n\nIPA plots every attribute as a point on a grid. The vertical axis is importance; the horizontal axis is performance. The axes cross at the average importance and average performance across all attributes, dividing the space into four quadrants.\n\n- **Concentrate Here (high importance, low performance).** This is the headline of the analysis. Customers care about these attributes and you are underdelivering. These are your fix-first priorities, where improvement will most improve overall satisfaction and loyalty.\n- **Keep Up the Good Work (high importance, high performance).** These are your genuine strengths. The instruction in the label is literal: protect them, resource them, and use them in positioning. Losing ground here is the fastest way to erode loyalty.\n- **Low Priority (low importance, low performance).** Weak performance, but customers do not care much. Do not spend scarce effort here just because the score is low; the return is small.\n- **Possible Overkill (low importance, high performance).** You are excellent at something customers barely value. This quadrant is the one teams overlook, and it is where you find resources: effort you can move toward the Concentrate Here quadrant.\n\nThe discipline of IPA is that low performance alone never justifies action. A low score only matters when it sits against high importance.\n\n## Stated vs Derived Importance\n\nThere are two ways to get the importance axis, and the choice shapes the whole analysis.\n\n**Stated importance** asks customers directly: how important is response time, on a scale of one to five? It is easy to collect but flawed. Respondents tend to rate almost everything as important, compressing the axis, and they are genuinely poor at introspecting on what drives their own behavior.\n\n**Derived importance** infers importance statistically from how strongly each attribute correlates with an overall outcome such as satisfaction, loyalty, or repurchase, typically through a regression-based key driver analysis. It reflects what actually moves the outcome rather than what customers claim.\n\nThe two often disagree, and when they do, derived importance is usually the more trustworthy signal. The strongest practice is to capture both: a direct importance rating and a derived importance score on the same attributes. Where they diverge, you learn something, for example, that customers say price is paramount but their behavior is driven by reliability. See the companion [key driver analysis guide](/docs/key-driver-analysis-guide) for the regression mechanics behind derived importance.\n\n## How to Run an IPA in Six Steps\n\n1. **Define the attribute list.** Draw attributes from prior qualitative research so they reflect the language customers actually use. Keep each one concrete and single-barreled: \"support resolves my issue quickly\" rather than \"helpful and fast support.\"\n2. **Measure performance.** Ask customers to rate how well you deliver on each attribute, using a consistent scale (a 5- or 7-point Likert scale is standard). Consistency across attributes is what makes the axis comparable.\n3. **Measure importance.** Collect stated importance on the same scale, derive it from a driver analysis, or ideally do both.\n4. **Compute the means.** Calculate the average performance and average importance across all attributes. These two values become your crosshairs.\n5. **Plot the quadrants.** Place each attribute at its importance and performance coordinates, and draw the axes at the two means. Data-centered crosshairs, not the scale midpoint, spread the attributes out and make the quadrants meaningful.\n6. **Act by quadrant.** Build the roadmap from Concentrate Here, defend Keep Up the Good Work, park Low Priority, and harvest resources from Possible Overkill.\n\n## Common Pitfalls\n\n- **Scale-centered crosshairs.** Crossing the axes at the fixed scale midpoint (3 on a 5-point scale) instead of the data mean is the most common mistake. Satisfaction ratings cluster high, so scale-centering can dump every attribute into a single quadrant and destroy the analysis. Use the means.\n- **Treating stated importance as truth.** Direct importance ratings compress and mislead. Validate them against derived importance whenever you can.\n- **Vague or double-barreled attributes.** If an attribute bundles two ideas, you cannot act on the result. Split them.\n- **Ignoring collinearity.** Importance and performance are sometimes correlated across attributes, which can distort interpretation. Read the quadrants as directional priorities, not precise coordinates.\n- **A frozen snapshot.** Priorities shift. Re-run IPA on a cadence so the matrix reflects the current experience rather than last year's.\n\n## The Modern Approach: Fill the Matrix With AI\n\nTraditional IPA depends on a long rating-scale survey, and long scale surveys produce exactly the compressed, everything-is-a-five data that makes the matrix hard to read. AI-native research fixes both the data and the interpretation.\n\nKoji collects both axes of the matrix inside a single AI-moderated conversation. Its structured questions include a dedicated **scale** type that captures clean, consistent performance and importance ratings, one of six structured question types (open_ended, scale, single_choice, multiple_choice, ranking, and yes_no) you can mix into one study. The difference is what happens next: when a customer rates an important attribute poorly, Koji's AI moderator immediately asks an open-ended follow-up and captures why. So you do not just learn that \"resolution speed\" landed in Concentrate Here; you learn the three recurring reasons it did.\n\nBecause Koji records a scale rating and the reasoning on the same attribute, you can compare stated importance with derived importance without running a separate study, and its real-time reporting rebuilds the matrix as responses arrive. Where a traditional survey platform such as SurveyMonkey or Qualtrics hands you a spreadsheet of averages to chart by hand, an AI-native platform hands you a quadrant map with the customer's own explanation attached, in hours rather than weeks. And Koji's built-in quality scoring (a 1 to 5 scale that only counts high-quality conversations) keeps speeders and straight-liners from flattening the importance axis. You do not need a statistics background to read the result: the priorities, and the reasons behind them, are written in plain language.\n\n## A Worked Example\n\nImagine a B2B software team surveys 400 customers on eight attributes: onboarding, support speed, reliability, reporting, price, integrations, ease of use, and account management. Support speed scores low on performance (2.9 on a 5-point scale) but high on importance (4.6), landing firmly in Concentrate Here, so it becomes the top roadmap item. Reporting also scores low on performance (3.0) but low on importance (2.4), landing in Low Priority, so the team resists the temptation to rebuild it. Meanwhile the polished account-management program scores high on performance (4.5) but low on importance (2.7): Possible Overkill, and a candidate to trim. Without IPA, the low reporting and low support scores look equally urgent; the matrix shows only one of them is worth the sprint. That single reallocation, from a low-importance rebuild to a high-importance fix, is the entire return on the method.\n\n## When to Use IPA (and When Not To)\n\nReach for IPA when you have a defined list of attributes and need to sequence improvements: post-purchase experience audits, feature-set reviews, service quality studies, and win-loss debriefs are all natural fits. It is less useful when you do not yet know the attributes (do exploratory interviews first) or when you need to model trade-offs between bundled options, where conjoint analysis is the better tool. IPA is a prioritization lens, not a discovery method, and it works best downstream of qualitative research that has already named the attributes worth measuring.\n\n## Related Resources\n\n- [Structured Questions Guide](/docs/structured-questions-guide) — the six question types, including the scale question that powers IPA ratings\n- [Key Driver Analysis Guide](/docs/key-driver-analysis-guide) — the regression method behind derived importance\n- [CSAT vs NPS vs CES](/docs/csat-vs-nps-vs-ces) — choosing the overall outcome metric your IPA improves\n- [Customer Satisfaction Survey Questions](/docs/customer-satisfaction-survey-questions) — writing the attribute items to plot\n- [Feature Prioritization Survey Guide](/docs/feature-prioritization-survey-guide) — narrowing a long attribute list before IPA\n- [Voice of Customer Metrics and KPIs](/docs/voice-of-customer-metrics-kpis) — tracking the outcomes IPA is meant to move","category":"Analysis & Synthesis","lastModified":"2026-07-14T03:18:14.874357+00:00","metaTitle":"Importance-Performance Analysis (IPA): The Priority Matrix Guide (2026)","metaDescription":"Learn importance-performance analysis (IPA) step by step: the four quadrants, stated vs derived importance, how to plot the matrix, common pitfalls, and how AI research fills it in faster.","keywords":["importance-performance analysis","IPA matrix","priority matrix","importance performance grid","concentrate here quadrant","attribute prioritization","customer satisfaction analysis","Martilla James","stated vs derived importance","quadrant analysis"],"aiSummary":"Importance-Performance Analysis (IPA) plots how important each attribute is to customers against how well you perform on it, producing a four-quadrant matrix (Concentrate Here, Keep Up the Good Work, Low Priority, Possible Overkill) that turns satisfaction data into a clear prioritization decision.","aiPrerequisites":["Familiarity with survey scales and averages","Access to attribute-level satisfaction or feedback data"],"aiLearningOutcomes":["Explain what importance-performance analysis measures and when to use it","Interpret all four quadrants of the IPA matrix correctly","Choose between stated and derived importance","Run a six-step IPA from attribute list to prioritized action","Avoid the crosshair, collinearity, and vague-attribute pitfalls","Collect consistent attribute data with structured scale questions"],"aiDifficulty":"intermediate","aiEstimatedTime":"11 min read"}],"pagination":{"total":1,"returned":1,"offset":0}}