# GE-McKinsey Nine-Box Matrix The GE-McKinsey nine-box matrix rates each business on industry attractiveness and competitive strength, then sorts it into invest, hold, or harvest. **The question.** Which of our businesses deserve investment, which should be held for earnings, and which should be harvested or sold? Also known as: Also called the nine-box matrix, the industry attractiveness–business strength matrix, or GE multifactorial analysis. Hold is the middle band; McKinsey's own charts label it selectivity or earnings. ## When to use it - A diversified group has to split a limited capital budget across businesses that do not share a market, and share-and-growth data would mislead. - Unit heads each bring a forecast that makes their own business look investable, and the board wants an outside scoring of industry and position. - Share is a poor stand-in for cost or profit, so the BCG matrix's horizontal axis measures the wrong thing. - The portfolio mix includes slow but structurally attractive industries and fast but structurally poor ones, and a growth-only sort would invert the ranking. ## When to avoid it - There are only two or three similar units and the trade-off is already clear. A conversation and a cash-flow comparison will do the same job. - Market boundaries are so contested that attractiveness scores will be an argument about definitions. Settle the served market first. - Units share plants, brands, or customers so heavily that treating each as a standalone investment object is false. Score the shared system, or do not plot. - The decision is a valuation or a sale price. The box is a posture, not a present value. ## Overview The nine-box places each business by the attractiveness of its industry and by its competitive strength inside that industry, then sorts the portfolio into invest, hold, and harvest. Each axis is a weighted score of several factors, so the tool can handle cases where share and growth are poor stand-ins for position and opportunity. The decision it supports is where a diversified group's scarce capital should go. The scores are judgments. If units choose the factors or rate themselves, every business climbs above the diagonal and the grid is a negotiation. Attractiveness is about the industry's structure, not last year's profit, and a harvest still has to be checked against the plants, brands, and customers other units share. Re-score when those facts move. ## Two axes and three postures ### Industry attractiveness The external axis. Typical factors include market size and growth, industry profitability, cyclicality, competitive intensity, barriers to entry, customer power, technology change, and regulation. Weight the factors for this portfolio and write the weights down before anyone scores. Attractiveness is a property of the industry, not of the unit's current year. ### Competitive strength The internal axis. Typical factors include relative share, cost position, brand or spec control, quality, access to distribution, technology, and management depth. Score the unit against the rival that matters in its served market. A unit can be strong in a poor industry, which is a different problem from a weak unit in a rich one. ### Invest High attractiveness with high or medium strength, and medium attractiveness with high strength. The posture is to fund growth, capacity, or share-building at a level that can change the unit's position. McKinsey's charts put these cells above the diagonal. The test is whether added capital can earn more here than in the rest of the portfolio. ### Selective (hold) The diagonal. High attractiveness with low strength, medium with medium, and low attractiveness with high strength. The posture is earn, fix a specific gap, or wait for a defined event, not a blank check. A strong unit in a weak industry is not the same as a weak unit in an attractive one, and the actions differ. ### Harvest Medium attractiveness with low strength, and low attractiveness with medium or low strength. The posture is to take cash out, sell, or close, and to put little new capital in. Check dependencies on other units before acting, because a harvest that looks clean on the grid can remove volume or coverage the invest units use. ## What it is not - A more accurate BCG matrix. It replaces two single measures with judged scores. If the scores are political, it is less accurate, not more. - A ranking of current profit. A highly profitable unit in a structurally poor industry can still be a harvest, and a modest earner in a rich industry can still be an invest. - A rule that every box must be filled. A group of invest and hold units with no harvest is not incomplete. ## The steps 1. **Define the units and their industries.** Write the product and geographic boundary of each unit, and the industry against which attractiveness will be scored. Shared costs and customers get listed here, so they are not discovered after a harvest decision. 2. **Agree the factors and weights.** Choose a short list of attractiveness factors and a short list of strength factors, and set weights that add to 100 on each axis. Do this centrally, before scores exist, or every unit will pick the factors it wins on. 3. **Score with evidence, not the unit's forecast.** Score each factor from industry data, customer and competitor evidence, and cost comparisons. The unit's own plan is an input, not the score. Keep a one-line reason next to each factor. 4. **Plot and separate the three bands.** Place each unit on the nine-box. Size bubbles by capital employed or revenue. Mark the invest, selective, and harvest bands, and write why any unit sitting on a line belongs on one side. 5. **Assign capital and a review trigger.** Give each unit a posture and a capital amount that adds up to the cash the group actually has. For selective units, name the gap to close and the date the bet is retired. For harvest units, list what other units lose if this one is sold or starved. ## In practice A $1.6 billion building-products group came to the capital review with six units and $180 million of requests against $95 million of free cash. The consultant scored each unit on a shared factor list instead of on last year's EBIT. Roofing membranes sat in a $2.1 billion market growing 6 percent, with industry returns around 14 percent, and the unit held 28 percent share at the lowest delivered cost, so it scored high/high. Decorative laminates led a slow market and scored medium/medium. Vinyl siding earned $8 million on $110 million of sales, held 8 percent share, and sat in a market shrinking 3 percent a year, so it scored low/low. The plan put $55 million into a second coating line for membranes, held laminates at $12 million of maintenance capital, and stopped growth spending on siding, which is expected to throw off about $18 million of cash over three years. An adhesives unit at medium attractiveness and high strength received $28 million against a twelve-month earn-out the board could reverse. Two remaining units stayed on the diagonal with earnings targets and no expansion money. ## What you leave with - A written industry definition and factor list, with weights, used for every unit. - A scored grid with the evidence behind each axis, not only the cell label. - A capital allocation that sums to available cash, split by invest, hold, and harvest. - For each selective unit, the specific gap, the money, and the date the decision is revisited. - The shared plants, brands, and customers that constrain any harvest or sale. ## Failure points ### Growth standing in for attractiveness The team uses market growth as the whole vertical axis, which is BCG with extra boxes. A fast, structurally poor industry gets investment, and a slow, structurally rich one gets starved. ### Units scoring themselves Each division picks factors and scores its own strength. Every unit lands above the diagonal. The plot records lobbying, not position. ### One posture for the whole diagonal A strong cash generator in a poor industry is treated the same as a weak challenger in a rich one. The first should be milked with care; the second is a turnaround or an exit. The diagonal is a band of different problems. ### Harvesting a unit the others use Siding is sold on its own returns, and the membranes business then loses mixed loads, a shared sales force, or a plant's overhead absorption. Group cash falls after a harvest that looked clean on the grid. ### A plot made once Industry returns and relative cost move. Last year's invest unit is this year's hold, and the capital plan still treats it as a growth engine. Re-score on the planning cycle, and when a factor used in the weights changes. The failures above are usually [Framing](https://usefulconsultingframeworks.com/your-edge/#framing): the wrong problem was chosen before this framework was applied. ## Origin McKinsey built the nine-box for General Electric's corporate planning staff in the early 1970s, when GE ran dozens of businesses and needed a way to allocate capital that did not rest on each unit's own forecast. The two axes are industry attractiveness and the unit's competitive strength inside that industry. Each axis is a weighted score of several factors, not a single measure. McKinsey's later account, in the 2008 Enduring Ideas series, presents it as a systematic way for a multibusiness company to decide where cash should go. Units above the diagonal are candidates for investment and growth, those on the diagonal for selective investment or earnings, and those below for sale, liquidation, or running for cash. GE's name and McKinsey's name travel with the tool because that is where it was built. McKinsey is a trademark of McKinsey & Company, and GE is a trademark of GE Aerospace. The method itself is widely taught and is not licensed. McKinsey's presentations and text are copyrighted, so this note states the logic in its own words. Later McKinsey work, including MACS, added whether the parent is the best owner, which the original nine-box does not ask. ## Further reading - [Enduring Ideas: The GE–McKinsey nine-box matrix](https://www.mckinsey.com/capabilities/strategy-and-corporate-finance/our-insights/enduring-ideas-the-ge-and-mckinsey-nine-box-matrix). McKinsey, 2008. The firm's own short account of the tool, including the three bands and the warning that a strong unit in a weak industry is not the same case as the reverse. Free. - [GE multifactorial analysis](https://en.wikipedia.org/wiki/GE_multifactorial_analysis). Wikipedia's summary of the nine-box and its factor lists. Secondary, and some of the factor counts are textbook conventions rather than McKinsey's original. Use it for orientation and check the sources. ## Related frameworks - [BCG Growth-Share Matrix](https://usefulconsultingframeworks.com/frameworks/bcg-matrix/): Uses two single measures, growth and relative share, and is faster when those are fair stand-ins. Pick BCG when the data already exist; pick the nine-box when they would mislead. - [PESTLE and Its Extensions](https://usefulconsultingframeworks.com/frameworks/pestle/): A scan of outside forces that can feed the attractiveness axis. Run it when attractiveness is being scored from growth and nothing else. - [Porter's Five Forces](https://usefulconsultingframeworks.com/frameworks/porters-five-forces/): Explains how much profit an industry can earn, which is the core of attractiveness. Use it to build that axis rather than averaging a long, unweighted list. - [SWOT Analysis](https://usefulconsultingframeworks.com/frameworks/swot/): A looser internal and external sort for one unit. Use the nine-box when the job is a portfolio capital decision, not a description. Last updated October 20, 2026.