The channel divide was always artificial
No customer has ever once thought of themselves as an online customer or an in-store customer. They research on a phone in bed, check stock on the train, decide in an aisle, and return the thing through whichever door involves the least argument. From where they sit there is no channel — there is a shop that is sometimes glass and sometimes glowing. The divide was never a description of customer behaviour. It was an accident of org charts, budgets, incentives and the software each team happened to buy in the year they were formed. It persists because it is load-bearing internally: someone owns digital, someone owns retail, and they have different targets that can be hit at each other's expense. Customers stopped respecting that boundary well over a decade ago. Most retail stacks still enforce it rigorously, which is why the experience of buying online and returning in store so often feels like being handed between two companies who have heard of each other. Every one of those seams is a place a competitor can win, and none of them exist for a reason the customer would accept.
The connected store
A connected store is not a futuristic concept. It is simply a store that knows what the rest of the business already knows. It recognises that the person standing at the shelf is the one who abandoned a basket last night. It knows the size they want is in the stockroom rather than letting them leave believing it is gone. It knows they have shopped here for three years, and it does not ask them to prove it. Notice that none of this requires invention. The abandoned basket is recorded. The stock position is in a system. The three-year history is sitting in the loyalty database. The data almost always exists — it simply terminates somewhere the shop floor cannot see, usually behind a nightly export and an integration nobody owns. That is a plumbing failure, and plumbing failures are unglamorous, which is exactly why they get reframed as strategy problems and put on a roadmap instead of being fixed. The gap between a normal store and a connected one is rarely a new capability. It is a join that was never built because it was nobody's job.
Identity is the connective tissue
Nothing connects without a thread that survives the trip between channels, and there are only two ways to obtain one. You can infer it — stitch together devices, locations, faces and probabilities into a guess about who this person might be — or you can be given it. The inferred version is seductive because it requires nothing from the customer, and it is a trap. It is brittle, breaking the moment someone changes phone or stands in an unusual place. It is faintly hostile, in that its entire premise is knowing something the person did not agree to tell you. And it is steadily becoming illegal in more places every year. The given version is slower to build and enormously more durable: a loyalty identity, handed over deliberately in exchange for something obviously worth having, becomes the join key for the entire system. It survives device changes. It survives regulation. It survives the customer asking what you know about them, which is the test that ultimately matters. Ask and earn, and the connection holds for years. Infer, and you are building on something that can be taken away by a browser update.
What to build now
The practical order almost never changes, and most failures are a result of attempting it out of sequence. First, instrument the store, because behaviour you cannot observe cannot be connected to anything. Second, give customers a reason to identify themselves that they would choose freely if they read the terms — not a dark pattern, not a mandatory account, something they would take. Third, join those two to the transaction record, which is the boring integration work that everybody defers and which is, in fact, the entire product. Only then automate. Automation applied to a disconnected stack does not produce intelligence; it produces the wrong decision faster and at scale, with more confidence and less human oversight than the manual process it replaced. The sequence is unforgiving in that respect: every step depends on the one before it, and skipping to the interesting part is the most common way to spend a large budget on a system that recommends nonsense politely.
The next five years
The prediction worth making is not about technology. Nothing described here needs a breakthrough — every component exists today and most of it is affordable. The change will be organisational, and it will look like the quiet disappearance of a job title. The Head of Digital and the Head of Retail will stop being separate people with competing targets, because the boundary they were hired to manage stopped describing anything real long before the technology caught up. Retailers who complete that sequence will not have an online strategy and an offline strategy. They will have a customer strategy, and channels will become what they always should have been: delivery mechanisms, chosen per customer, per moment, without ceremony. The ones who do not will keep running two businesses that share a logo, a warehouse and a growing suspicion that their customers prefer someone else's.
- The online/offline split describes org charts, not customer behaviour.
- The data usually exists — it just never reaches the shop floor.
- Identity must be offered by the customer, not inferred about them.