Ground

Wave 2

Ground.

Ground is what everything else grows on: a product catalog rich enough to answer the questions buyers actually decide on, inventory numbers you'd bet on, one definition of what a customer is, and a stack where the pieces genuinely talk to each other. It won't get you found, win the sale, or keep the customer, those are later waves. It answers a narrower question, and it has to answer it cold: when a person, a system, or a model asks your business something, does it get one true answer or three conflicting ones?

It comes right after Truth because Truth usually finds it broken. The baseline read turns up a catalog missing the attributes buyers filter on, an inventory count that changes depending on which system you ask, and a "customer" who's an email address in one tool, a subscriber in another, and a household in a third. Every wave above this one consumes that data, and consumes it faster than a person can sanity-check it. When personalization recommends an out-of-stock bestseller, or an email flow fires on a segment that no longer exists, or a product page confidently describes last season's formula, the tool usually isn't the problem. The ground is.

None of this demos well, and all of it decides whether the work above it moves the business or just looks like it does.

And the wave got more important, because machines now read your product data before your customers do. The answer engines, the shopping surfaces, and the comparison layers assemble their whole picture of your brand from feeds, attributes, and reviews. A thin catalog used to cost you a filter ranking. Now it costs you the shortlist.

Ground versus the waves next to it

What good looks like.

Ground ships five things. None of them is glamorous, and skipping any one of them shows up two waves later.

1. A catalog that answers real questions

Not just titles and prices, but the things a buyer actually decides on: size and fit, ingredients, materials, compatibility, care, what it pairs with, what it replaces. Structured, complete across the range, and consistent, because a machine assembling an answer treats a missing attribute as a missing product. The test is simple. Take the ten questions your CX team hears most and see whether the catalog can answer them without a human translating. If it can't, neither can anything built on top of it.

2. One inventory number everyone trusts

One inventory position per SKU that the storefront, the ops team, and the finance sheet all agree on, updated fast enough to sell against. Everything downstream leans on it: the promise at checkout, the routing decision, the replenishment email that shouldn't fire for a product you can't ship. A brand that can't trust its inventory number ends up padding everything with safety margin, and the padding becomes its own leak.

3. One customer, one record

Identity stitched across the storefront, the email platform, the subscription tool, and the helpdesk, so "customer" means the same person everywhere. It's the difference between lifecycle work that feels attentive and lifecycle work that sends a cancellation offer to someone who never subscribed. It's also the quiet precondition for every retention model in Grow, because a churn signal computed on a fractured identity is noise.

4. Definitions, written down and agreed on

What counts as revenue: booked, shipped, or net of returns? What's an active customer, and over what window? When exactly does a subscriber count as churned? One answer per question, one owner per definition, served everywhere the question gets asked. The gaps between competing definitions are where confident, wrong conclusions come from, and the fix isn't a data warehouse project. It's a decision about which source wins, made once per definition, plus the small plumbing to make that source the one everyone reads.

5. A stack that's wired, not a stack that's new

Ground is almost never a replatform. The commerce platform, the email tool, the subscription system, the order management, the helpdesk: the pieces are usually fine, it's the connections that are missing. So we connect what exists, retire what's redundant, and recommend something new only where the current piece is genuinely the constraint. Most growth-stage brands do well with a solid engine and a flexible front rather than a full re-architecture, and recommending less than you expected to buy is part of the job.

How it runs. Back to Marlowe, the hypothetical brand from the Truth wave. The baseline found the catalog thin: the bestsellers carried full attributes, the long tail carried a title and a photo, and inventory disagreed with itself by warehouse. Ground filled the attributes against the buyer-question list, wired one inventory position through to the storefront promise, and gave "churned" a single definition with a single owner. Nothing about the work would make a headline, and every wave after it moved faster because of it.

The threads, in this wave.

Woven in.

A model running on a dirty catalog produces confident nonsense, at volume, faster than anyone can catch it. That's the whole discipline of this wave: ground first, tools second. The tools do help here, quietly, filling and normalizing attributes across a long tail, deduplicating identities, flagging the definitions that disagree. But they're staff on the project, not the point of it.

Your team.

Every domain gets a named owner: someone who owns the catalog's completeness, someone who owns the inventory number, someone who owns the customer record. Each gets a review cadence they can actually keep, because a canonical source that quietly goes stale is worse than none, it just lends confidence to the wrong number. The owners are your people. We set the cadence with them, then hand them the keys.

The coda.

Everything downstream stands on this. Build the ground once, keep it true, and every wave above it compounds.