GE-McKinsey 9-box matrix for product strategy

The GE-McKinsey 9-box gives you a defensible way to allocate investment across a product portfolio when you have more bets than budget. It is not a feature-ranking tool and it is not the BCG matrix. Used correctly, it structures a director-level portfolio conversation in a way that survives interrogation. Used incorrectly, it looks like a flashcard answer.

Where it came from and why it differs from BCG

McKinsey built this for GE in the early 1970s. GE had roughly 150 business units it could no longer manage with simple cash-flow projections. The BCG growth-share matrix was already in circulation, but it reduced everything to two fixed inputs: relative market share and growth rate. That worked for simple portfolios. For 150 units across industries with very different competitive dynamics, it collapsed nuance that mattered.

The 9-box replaced two fixed inputs with two weighted composite scores. The Y axis is Industry Attractiveness: market size, long-run growth rate, industry profitability, competitive intensity, entry and exit barriers, cyclicality, and regulatory environment. The X axis is Competitive Strength: market share, share growth relative to competitors, brand equity, customer loyalty, product differentiation, VRIO capabilities, and margin relative to industry peers. You choose which factors apply to your context, assign weights that sum to 1.0, rate each unit 1-5 on each factor, compute the weighted score, and plot. Each unit is a circle; circle diameter represents revenue or investment size. That visual carries real information: a small circle in the invest zone is a different conversation than a large circle in the harvest zone.

The nine cells collapse to three strategic zones. Top-left (high attractiveness, high strength): invest and grow. The diagonal: selective investment, manage for earnings. Bottom-right (low attractiveness, low strength): harvest or exit.

Running it as a PM

First, define your units. The matrix applies to distinct product lines or strategic bets, not individual features. A feature is too small a unit of analysis. If you are a director at a mid-size SaaS company, your units might be: core platform, a vertical expansion, an AI assistant, and a marketplace. Four units you can actually score differently.

Second, pick factors that are true for your context and weight them explicitly. For industry attractiveness at a B2B SaaS company, market size and long-run growth matter, but so does switching cost structure in the buyer’s procurement cycle. A market with strong network effects and high switching costs scores higher than an equally large market that is easy to exit. For competitive strength, weight distribution moat and retention rate heavily. Market share alone is a weak signal if it was purchased with discounting.

Third, score each unit on each factor (1-5), multiply by weight, sum, and plot. Do this in a spreadsheet, not in your head. The point is to make the trade-offs legible, not to generate a precise answer.

Fourth, read the visual. Where a unit sits tells you the starting allocation. What you do next depends on trajectory, synergy, and lovability, none of which the matrix captures on its own.

Worked example: a developer tools company

Say you have four products: a core IDE plugin (high share, maturing market), a code review tool (growing market, third-place position), an AI agent for test generation (high growth, early position), and a documentation generator (crowded, commoditizing fast).

• IDE plugin: high competitive strength, moderate attractiveness. Manage for earnings, extract margin.
• Code review: high attractiveness, low-to-mid strength. Selective investment, find the wedge.
• AI agent for test generation: high attractiveness, low strength currently. Invest to build strength fast or exit before the window closes.
• Documentation generator: low attractiveness, low strength. Harvest or kill.

That four-unit read takes ten minutes to sketch. It surfaces the conversation: are we willing to fund the test generation bet aggressively, or do we let it die quietly? The matrix does not answer that. It makes the stakes explicit so the room can argue with evidence instead of opinion.

Use it, do not recite it

In an interview, the failure mode is narrating the framework: “So on the Y axis we have industry attractiveness, which includes market size and growth rate…” That is a flashcard. The strong move is to use it live: propose the relevant factors for the specific portfolio in the question, acknowledge that scoring is a judgment call not an objective measure, and call out what the matrix will not tell you.

The matrix ignores synergies. A product that scores in the harvest zone may scaffold a higher-strength product through shared data, shared distribution, or network effects. Platform companies in particular need to overlay the matrix with cross-product dependency maps before making harvest or exit calls. A small cohort of deeply engaged users on a bottom-right product might anchor the whole ecosystem.

The matrix is also a static snapshot. It says nothing about velocity: a unit moving rapidly rightward on competitive strength is a different investment than one that has been stuck in the same cell for three years.

The 2026 reframe

In an AI-era portfolio, the classic Competitive Strength axis needs recalibration. Technical capability used to be a durable moat: building the thing was hard. Now you can prototype almost anything in days. When everyone can build the same thing, technical capability is not worth scoring heavily as a strength factor.

The factors that compound now are: distribution and switching costs (can users get this elsewhere for free?), data network effects (does the product get measurably smarter from usage in a way competitors cannot replicate?), trust and safety track record (users and regulators now gatekeep AI products), and integration depth (is this embedded in the workflow or a tab people close?).

On the Industry Attractiveness axis, raw market growth is a weaker signal when nearly every market is being disrupted simultaneously. The dominant filter is viability: is there a clear willingness-to-pay signal, and is the market structurally large enough to cover AI inference costs plus labor at a sustainable margin? A product can sit in a “high growth” category and still be economically incoherent.

One addition worth making explicit: the matrix does not score lovability. A product in the invest zone that users merely tolerate will commoditize faster than the matrix predicts. Before finalizing any invest or harvest call, overlay retention curves and qualitative signal on whether users advocate for the product, forgive imperfection, and resist switching. That distinction separates a durable invest position from one that collapses under competition.

When not to use it

The 9-box requires actual scoring. If you do not have data on market size, competitive position, and relative margins for each unit, the scores are decoration. In that case, name the framework, explain the factors you would score, and say clearly that you are working from directional estimates. Claiming precision you do not have is worse than acknowledging the limits of your information.

Also: do not confuse this with the talent management 9-box (performance vs. potential for people). They share a name and a visual form. In a PM interview, be specific about which one you mean.

See also: BCG growth-share matrix for the simpler two-input version, Porter’s Five Forces for the industry attractiveness factors in more depth, and Ansoff matrix for directional growth strategy after you have your portfolio read.