Retail portfolio optimization: deciding across the whole fleet

Single-site analysis answers a narrow question. Taken on its own, is this address a good place to operate. Portfolio optimization works one level up. Given everything you already run and everything you might build, which set of moves leaves the whole network better off. The unit of decision changes from the candidate to the fleet, and a site that scores well in isolation can still be the wrong call once you account for the stores around it. The job is to improve the system, which sometimes means passing on a strong-looking site because it would feed off units you already own.

Every unit you open pulls part of its sales from demand that no store was serving and part from stores you already run. At the single-site level you can estimate that overlap against a few nearby units and move on. Across a fleet the effect compounds. One new store in a dense market can draw from three or four existing locations at once, and a relocation can ease pressure on one trade area while adding it somewhere else. Reading this at the portfolio level means netting the new demand against the demand that simply transfers between your own stores. The number that matters is the fleet total, because sales shuffled between your own registers add nothing to it.

• Add a unit where a trade area holds unmet demand that your existing stores do not reach. The test is net-new demand, after subtracting whatever the new store would pull from siblings.
• Relocate a unit that sits in the wrong spot within a market it should own, capturing demand it currently misses while keeping the surrounding area covered.
• Close a unit whose trade area is now blanketed by nearby siblings, where most of the demand it serves would transfer to them rather than walk away.
• Hold a unit that performs and sits in a market already well covered, where another opening would mostly cannibalize what you have.

White space and saturation are the same map read from two directions. White space is demand your fleet does not yet cover, the markets and trade areas where a well-placed unit would find customers who currently have to go elsewhere. Saturation is the opposite edge, the markets where you have added enough units that the next one mostly reshuffles existing demand. At the portfolio level you want to see both at once, ranked, so the markets worth entering and the markets worth pausing sit on the same screen. A single-site view can flag a promising gap, but it cannot tell you whether that gap is one of forty like it across your footprint or the last good one left.

A portfolio plan is also a schedule. The order in which you open units changes their economics, because each opening reshapes the trade areas around it before the next one lands. Open the anchor location in a market first, and a later infill unit gets scored against a network that already includes it. Open them in the wrong order, and you may green-light a site that looked clear only because its eventual neighbor was not on the map yet. Pacing matters for the same reasons capital and operations do, and the spatial piece is the part most plans miss. Sequencing each unit so it is evaluated against the fleet as it will actually exist keeps the plan honest as it rolls out.

Most portfolio questions come down to comparing options, such as opening in one market versus holding for a stronger one, or entering a new metro with two units instead of three. Scenarios let you build each version of the plan and read the fleet-level outcome side by side, with cannibalization and net-new already accounted for in every variant. The team compares whole plans on the same footing, and the trade-offs between coverage, overlap, and pacing become something you can see and defend rather than argue from memory.

Geod scores single sites and rolls those scores up to the network. Its cannibalization model is network-gravity based, so a candidate forecast is split into demand transferred from your own stores, demand captured from competitors, and genuinely net-new demand, with the worst-affected existing unit named. That same decomposition reads at the portfolio level, where overlap and net-new are summed across the fleet rather than judged one pair of stores at a time. Scenarios let you compare add, relocate, close, and hold options and see the fleet outcome of each. Every score stays explainable, tracing back to named components and dated sources, so a portfolio plan can be defended in the same committee that signs off on individual sites.