gp0545

Who you actually sell to before you sell anything

The single most common failure in online grocery is defining the market as "people who eat." Every household buys groceries, which makes the total addressable market meaningless as a planning input and useless as a targeting instruction. The playbook begins by narrowing on two axes simultaneously: geography and household behavior. Geography comes first because grocery is a physical-goods business where the cost to serve is a function of drive density, and behavior comes second because within any given ZIP code the spread between your best and worst customer is enormous.

Start with the delivery zone, not the city. A viable launch zone in 2027 is typically a cluster of contiguous neighborhoods where you can reach a large share of addresses within a short drive of one fulfillment point. The operational reason is route batching: when several orders sit close together, a single driver run can serve them all, and the cost per delivery falls sharply. When orders are scattered across a metro, each one becomes its own trip and the economics collapse. Practitioners generally think in terms of orders per route rather than orders per day, because the former is what actually determines whether a delivery makes or loses money. A zone that produces four or five stops per route behaves like a different business than a zone that produces one.

Within that zone, segment households by frequency and basket composition rather than by demographics. The four segments that repeatedly show up in grocery data are worth naming explicitly. Stock-up households place large, planned, recurring orders — often weekly, often on the same day — and they are the most valuable customers you can acquire because their behavior is predictable and their baskets absorb delivery cost. Top-up shoppers place small, frequent, unplanned orders driven by a missing ingredient; they are high-frequency but often unprofitable unless you have a very low cost to serve or a fee structure that reflects the true cost. Convenience-first households — often dual-income families with young children — value the delivery window and the accuracy far more than the price, and they are the segment most willing to pay a real fee. Constraint-driven households include people with mobility limitations, no vehicle, dietary restrictions requiring specialty sourcing, or care responsibilities that make a store trip genuinely difficult; they exhibit unusually strong retention because the alternative is not "shop somewhere else," it is "have a materially worse week."

Your ideal customer profile for launch is almost always the intersection of stock-up and convenience-first inside your densest zone. That household orders often enough to matter, spends enough per order to cover the pick-and-drive cost, and evaluates you on reliability rather than on whether a competitor is running a promotion. Write the profile down concretely: order cadence, approximate basket size, the categories they buy, the times they want delivery, and the specific alternative they are switching away from. That last element is the one most teams skip, and it is the one that determines your messaging. Switching from a big-box chain's own delivery service is a completely different conversation than switching from a weekly in-person trip.

There is a related segment worth evaluating early because it changes the shape of the business: business and institutional buyers. Small offices, coworking spaces, daycares, gyms, churches, and short-term-rental operators all buy grocery items on a schedule, in larger quantities, with far more tolerance for a fixed delivery day. A handful of these accounts can anchor a route's density and smooth the demand curve that consumer orders make lumpy. The trade-off is that they negotiate on price and expect invoicing, so treat them as a margin-stabilizing supplement rather than the core motion. The adjacent lesson from restaurant and office-supply distribution applies directly: recurring scheduled accounts are what make an inconsistent consumer route financially survivable.

Finally, define who you are explicitly *not* serving at launch. Excluding the low-density fringe of your zone, excluding delivery windows you cannot reliably hit, and excluding categories you cannot source well are all acts of go-to-market discipline. Every exclusion tightens the operational promise you are making, and in grocery the promise is the product.

Choosing the fulfillment motion that matches the segment

Once the segment is defined, the motion follows from it almost mechanically. There are four viable fulfillment models in 2027, and mixing them without intent is how operators end up with the cost structure of one and the customer expectations of another.

Store-picked delivery uses existing retail shelves as the inventory. Pickers walk the aisles of a partner or owned store and fill orders alongside regular shoppers. The advantage is near-zero capital cost and instant assortment breadth — you inherit whatever the store already stocks. The disadvantage is pick efficiency: a human walking a store designed for browsing is slow, out-of-stocks are frequent because the shelf is shared with walk-in customers, and pick labor becomes the dominant cost line. This motion fits stock-up households in mid-density suburbs where breadth of assortment matters more than speed, and it is the fastest path to market when you are testing demand before committing capital.

Dark stores are retail-format facilities closed to the public and laid out for picking speed rather than browsing. Aisles are arranged by pick frequency, high-velocity items sit near the pack stations, and inventory accuracy is far higher because nobody is putting products back in the wrong place. The trade-off is fixed cost: you are carrying rent, inventory, and staffing whether or not the orders come. Dark stores only make sense once a zone demonstrates sustained order volume, and the decision to open one should be triggered by a density threshold you set in advance, not by optimism.

Micro-fulfillment, whether automated or manual, compresses a limited high-velocity assortment into a small footprint close to demand. The assortment is deliberately narrow — the few thousand items that account for the overwhelming majority of orders — and the pick is fast. This motion fits top-up and convenience-first shoppers who want speed and are shopping from a mental list of staples rather than browsing. The risk is assortment disappointment: a customer who cannot find their specific brand once will often not check again.

Pickup, whether curbside or locker-based, removes the most expensive line item in the entire model — the drive — while retaining the digital ordering relationship. It is chronically underrated in go-to-market planning because it is less exciting than delivery, but it converts a loss-making price-sensitive household into a profitable one, and it gives you a fallback offer when a delivery slot is full. Any serious 2027 playbook offers pickup as a first-class option, not as an afterthought buried in checkout.

The sequencing matters more than the choice. Almost every durable operator started store-picked or pickup-only, proved that a specific zone generated repeatable weekly demand, and only then committed capital to a dedicated facility. Running the sequence backwards — building fulfillment infrastructure and then hunting for demand to fill it — is the single most expensive mistake available in this market, because the fixed costs start accruing on day one while the demand curve takes quarters to establish.

There is a useful parallel in adjacent last-mile categories. Pharmacy delivery, pet supply, and prepared-meal services all faced the same fork, and the ones that survived generally rented capability before owning it: third-party couriers before an owned fleet, shared warehouse space before a dedicated one, a marketplace storefront before a proprietary app. The capability you rent is reversible; the capability you build is not.

Unit economics: the numbers that decide whether this works

Grocery retail runs on famously thin margins — low single digits at the net line for traditional operators is normal, and gross margins on food are modest by retail standards. Layering a pick-and-deliver operation on top of that structure adds cost to a business that had very little cushion to begin with. This is the central financial fact of the category, and any playbook that does not confront it directly is a marketing document rather than a plan.

Build the model bottom-up, per order, not top-down from market size. The line items are consistent across operators:

Gross margin on the basket. Start with the actual product margin on a representative basket, not a blended category average. Fresh produce, prepared foods, and private label carry meaningfully better margin than staples like milk, eggs, and paper goods. Basket mix is therefore a margin lever, not just a merchandising detail.

Pick cost. Measure it as picker minutes per order, then multiply by fully loaded labor cost. This is the number that separates the fulfillment motions. A store pick of a large mixed basket takes substantially longer than a dark-store pick of the same basket, and that difference flows straight to the bottom line on every single order, forever.

Delivery cost. The relevant unit is cost per *stop*, which is driver cost per hour divided by stops per hour. Stops per hour is a function of route density and drop time. This is why zone selection is a financial decision disguised as a marketing one — a dense zone can produce several times the stops per hour of a sparse one with identical drivers and identical vehicles.

Packaging and cold chain. Insulated bags, ice packs, and temperature-controlled handling are a real per-order cost that early models routinely omit and that scales linearly with volume.

Fees and subsidies. Whatever the customer actually pays in delivery fee, service fee, price markup, or membership amortization. Note the word *actually* — promotional waivers and first-order-free offers must be modeled at their real redemption rate, not the list price.

Shrink and remakes. Spoilage, damaged goods, and the cost of making a customer whole after a bad order. Budget for it explicitly; teams that leave it out are consistently surprised by the gap between modeled and actual margin.

The output of this stack is contribution margin per order, and the discipline is to track it weekly by zone rather than as a company-wide average. A blended average hides the fact that one zone is subsidizing another, and it lets a structurally unprofitable expansion continue for months.

Then layer in the acquisition math. Customer acquisition cost divided by contribution margin per order tells you how many orders are required to pay back the acquisition — the single most useful number in this business. If payback takes more orders than a typical customer places before churning, growth actively destroys value. This is precisely the trap that consumed enormous amounts of capital in the rapid-delivery wave: acquisition was cheap to buy and expensive to keep, and the per-order economics never crossed over.

Retention deserves a sharper treatment than the standard cohort chart. In grocery, watch three specific signals. Second-order rate within the first few weeks is the strongest early predictor of a cohort's eventual value; a customer who does not reorder promptly usually never does. Order frequency stability — whether a household settles into a rhythm — matters more than any single month's spend, because a predictable household is a forecastable one. Basket trajectory tells you whether trust is growing; households that begin with shelf-stable goods and later add fresh produce, meat, and dairy are demonstrating confidence in your quality, and that expansion is where margin actually lives.

Two more economic levers are worth naming because they materially change the model. The first is slot-based pricing. Delivery cost varies enormously by time of day, so pricing that varies with it — cheaper off-peak windows, premium for narrow or immediate delivery — pushes demand toward the hours you can serve efficiently. This is standard practice across logistics-constrained services and it works because a meaningful share of households are genuinely time-flexible if given a reason. The second is basket-size thresholds. A minimum order value, or a fee that steps down as basket size rises, converts unprofitable top-up orders into profitable consolidated ones. Both levers change customer behavior rather than merely charging more for it, which is why they beat a flat fee increase.

Finally, model the retail media and partnership layer separately from the core operation, and only after the core is at or near breakeven. Selling sponsored placement to brands against first-party purchase data is a genuinely high-margin revenue stream that can meaningfully improve blended economics. But it is a supplement, not a rescue. An operator losing money on every delivery does not fix that by adding advertising; they simply add a second business to an unprofitable first one. Get contribution margin to at least neutral, then monetize the audience you have earned.

Where operators consistently get this wrong

Certain failures recur so reliably in this market that they function as a checklist. Reading them as a list of things to avoid is more useful than reading another list of things to do.

Launching too wide. The instinct to open a whole metro at once is nearly universal and nearly always wrong. Wide launches destroy route density, which destroys delivery economics, which forces price increases or service cuts that damage the brand in every zone at once. The disciplined alternative is to saturate one zone until orders-per-route hits your threshold, then clone the playbook into an adjacent zone that shares fulfillment infrastructure.

Buying growth with fee waivers. Free-delivery promotions reliably acquire customers who are shopping the promotion rather than the service. When the waiver ends, they leave. Worse, the promotional cohort corrupts your retention data, making the underlying business look healthier than it is and delaying the moment you discover the real numbers. If you use promotional acquisition at all, track promo and non-promo cohorts separately from day one.

Treating substitutions as an edge case. Out-of-stock handling is not an exception path; it is a routine, high-frequency event that defines the customer's perception of your competence. A household that receives a thoughtless substitution — the wrong size, a different dietary category, a product they specifically avoid — reads it as evidence that nobody is paying attention. Give customers real control: preference settings, approve-or-reject before finalization, and a clearly communicated policy. This single workflow drives an outsized share of churn.

Promising delivery windows the operation cannot hit. Narrow windows are a competitive weapon only when you actually hit them. A wide window reliably honored builds more trust than a narrow window frequently missed, because grocery delivery is planned around — someone is home, or a meal is being cooked, or a child is being picked up. Missing the window does not just annoy; it breaks a household's day.

Under-investing in the fresh categories. Produce, meat, and dairy are the hardest to execute and the most decisive for retention. Customers who do not trust your fresh handling will use you for paper towels and canned goods, which is the worst possible basket mix — low margin, low emotional attachment, easy to substitute with any competitor. Getting fresh right is expensive and it is the whole game.

Confusing app features with the product. Recommendation engines, voice ordering, and slick interfaces are worth building, but they do not compensate for a bruised peach or a missed window. In this category the product is a box of correct food arriving when promised. Feature roadmaps that outpace operational maturity are a common way to spend engineering budget on the wrong bottleneck.

Ignoring the incumbent's structural advantage. Large chains already own the stores, the buying power, the private label, and the customer relationships. Competing on price and assortment breadth against that is a losing position. The defensible angles are local sourcing the incumbents cannot replicate at scale, service quality in a specific geography, specialty assortment for underserved dietary needs, and category focus — the same logic that lets a specialty pet or wellness retailer coexist with a mass merchant.

Skipping the pickup option. Operators who position themselves as delivery-only leave the highest-margin fulfillment path on the table and lose every price-sensitive household to a competitor who offers it.

Scaling the team ahead of the density. Headcount added on the expectation of volume, rather than in response to it, turns a variable-cost operation into a fixed-cost one at exactly the wrong moment.

The operating cadence that keeps the playbook honest

A go-to-market playbook is only as good as the review rhythm that enforces it. Grocery moves daily, so the operating model needs loops at several time scales rather than a single monthly business review.

Daily is the operational loop. Review fill rate, substitution rate, on-time delivery percentage, and any orders that required a remedy. The purpose is not analysis but intervention: a supplier that missed a delivery, a picker who needs coaching, a route that ran long. Anything trending badly for two consecutive days gets an owner the same day.

Weekly is the commercial loop. Look at orders per route by zone, contribution margin per order, second-order rate for the newest cohort, and the assortment gaps that generated the most substitutions. This is where fee structure, slot pricing, and minimum-basket thresholds get adjusted. Weekly is also the right cadence for reviewing which acquisition channels produced customers who actually reordered, as opposed to customers who merely signed up.

Monthly is the cohort and expansion loop. Retention curves by cohort, lifetime contribution against acquisition cost, and the explicit go/no-go on the next zone. The expansion gate should be a pre-committed number — a density and margin threshold agreed before the emotional pressure to grow arrives — because expansion decisions made under pressure are how operators end up serving zones that lose money on every order.

Quarterly is the structural loop. Revisit the fulfillment motion itself: is store-picking still right, or has volume justified a dark store? Reassess supplier terms, private-label opportunity, partnership pipeline, and whether the retail media layer is ready to be built. Quarterly is also when you kill things — underperforming zones, categories that generate more complaints than margin, partnerships that neither lowered cost per delivery nor raised revenue per customer.

Two organizational details make this cadence work. First, one person owns each zone's contribution margin end to end — merchandising, fulfillment, and marketing decisions for that geography roll up to a single accountable owner, because splitting them across functions is how a zone loses money for a quarter while everyone reports green. Second, the customer-service queue feeds directly into the weekly review rather than sitting in a separate support silo. Complaints are the highest-resolution data you have about where the operation is failing, and routing them into the commercial loop turns support from a cost center into the earliest warning system you own.

The broader point is that in a business this thin, the operating cadence *is* the strategy. Marketing spend, assortment breadth, and technology investment are all downstream of whether each delivery makes money and whether the household orders again. Everything in the playbook exists to protect those two facts.

Related questions

How long should a single zone run before expanding?

Long enough to hit a pre-set orders-per-route and contribution-margin threshold and hold it for several consecutive weeks — not a fixed calendar period. Expansion triggered by time rather than by density is the most common cause of margin collapse across a multi-zone footprint.

Is delivery-only viable, or is pickup mandatory?

Delivery-only is viable for genuinely convenience-first, fee-tolerant segments in dense geographies. For everyone else, pickup is the option that makes price-sensitive households profitable by removing the drive, and omitting it hands that entire segment to a competitor.

Should you build an app before proving demand?

No. Prove demand with the lightest ordering surface you can stand up — a mobile web storefront or an existing marketplace listing. Apps improve retention for households already ordering; they rarely create the first order, and they consume budget that belongs in fulfillment quality.

What is the earliest reliable signal a cohort will retain?

The second-order rate within the first few weeks. Households that reorder quickly are establishing a habit; those that do not almost never return, regardless of how much re-engagement marketing you spend against them.

How do local and specialty grocers compete with national chains?

Not on price or assortment breadth. They compete on local sourcing the incumbents cannot replicate, on service quality within a tight geography, and on specialty or dietary assortment that mass merchants under-serve because the volume does not justify shelf space.

FAQ

What is the single most important metric in online grocery go-to-market?

Contribution margin per order, tracked by zone rather than blended across the business. Everything else — acquisition cost, retention, basket size — is an input to that number. A company-wide average will hide a structurally unprofitable zone for months while it quietly consumes the margin generated elsewhere.

Do subscriptions and memberships actually improve the economics?

They can, but only when the perks push behavior toward profitability rather than simply discounting the current behavior. Bundling free delivery into a membership works if it increases order frequency and consolidates baskets. It destroys margin if it mainly subsidizes small, frequent, low-value top-up orders that were already marginal.

How much should you invest in technology at launch?

Enough for reliable ordering, accurate real-time inventory, and route planning — and no more. Inventory accuracy is the highest-leverage technical investment because it directly reduces substitutions, which directly reduces churn. Recommendation engines and personalization are valuable later, once the operation reliably delivers what it promised.

Is retail media a realistic revenue stream for a smaller operator?

Eventually, and only with sufficient order volume and clean first-party purchase data. It is a high-margin supplement that can meaningfully improve blended economics, but it does not fix a core operation that loses money per delivery. Get contribution margin to at least neutral before building the monetization layer on top.

What is the right response when an order goes wrong?

Resolve it immediately and generously, then feed the root cause into the weekly operating review. The cost of a refund or replacement is almost always lower than the cost of reacquiring the household. Arguing over a bruised item to protect a few dollars of margin is the most expensive discount in the category.

How do you decide between third-party couriers and an owned delivery fleet?

Rent the capability until volume makes owning it clearly cheaper per stop. Third-party couriers convert a fixed cost into a variable one, which is exactly what you want while demand is still uncertain. Once a zone consistently produces enough stops per route, an owned fleet lowers cost per delivery and improves control over the customer-facing part of the experience.

Sources

• https://www.mckinsey.com/industries/retail/our-insights
• https://hbr.org/topic/subject/operations-and-supply-chain-management
• https://www.statista.com/markets/423/topic/540/e-commerce/
• https://www.fmi.org/our-research
• https://nrf.com/research
• https://www.bain.com/insights/topics/retail/
• https://www.census.gov/retail/index.html
• https://www.usda.gov/topics/food-and-nutrition
• https://www.deloitte.com/us/en/insights/industry/retail-distribution.html

Related on PULSE

• [Inbound demand-capture GTM playbook in 2027](/knowledge/gp0511)
• [Sales-assisted PLG for mid-market in 2027](/knowledge/gp0510)
• [Reseller and VAR channel GTM playbook in 2027](/knowledge/gp0509)
• [International and geo-expansion GTM playbook in 2027](/knowledge/gp0508)
• [Vertical SaaS go-to-market playbook for healthcare in 2027](/knowledge/gp0507)