The ecommerce CRM playbook.

Ecommerce is the one business in this guide where most transactions involve no conversation at all, and that shapes everything.

A store can take thousands of orders without anybody speaking to a customer, which is the point of it. The consequence is that when a conversation does happen, it is unusual, and it is almost always at one of two moments: somebody is stuck before buying, or something went wrong after. Both are expensive to handle badly and neither fits a sales pipeline.

The second difference is the shape of the value. A single order is often worth very little and a customer who orders regularly is worth a great deal, so the entire economics sit in the repeat rather than the first purchase. A system built to close a deal once has optimised the least valuable event in the business.

The third is that the useful signals are behavioural rather than declared. Nobody fills in a form saying they are considering a purchase. They put something in a cart and leave, or they look at a product page repeatedly, or they stop ordering after two years of ordering monthly. Those are the moments worth acting on and none of them is a lead in the traditional sense.

So what an ecommerce CRM has to do is watch behaviour, know the difference between a first-time buyer and a regular, and intervene in the small number of moments where a conversation changes the outcome, without inserting one into the thousands where it would only be friction.

Abandoned carts are the most written-about problem in ecommerce and the most thoughtlessly handled.

The standard response is a sequence of emails, which works at low cost and low rates and is worth having. What it does not do is distinguish between somebody who was idly browsing and somebody who filled in a shipping address for a large order and stopped at the payment step. Those are not the same event and treating them identically is why the whole channel gets written off as low value.

The second one is a genuine buying intention that hit an obstacle, and the obstacle is usually specific and solvable: the shipping cost surprised them, the delivery date did not work, they were not sure the size or fit was right, the payment method they wanted was not there, or they wanted to ask something and could not find how.

For a small order, an email is the right weight of response. For a high-value cart, a real conversation is not intrusive, it is service, and stores that never try it are leaving the most recoverable orders on the floor because the tooling only knows how to send email.

The judgement is the product here. A CRM that calls everybody who abandons a cart will annoy a great many people and deserve to. One that can tell a large, nearly complete, first-time order from a browse, and treat only the first as worth a conversation, is doing something no email sequence can.

The other conversation worth having is the one that happens before the order, and most stores answer it far too slowly.

Somebody is looking at a product and needs one thing confirmed before they will commit. Will it fit, will it arrive by a date, does it work with the thing they already own, what happens if they need to return it. That question is the last obstacle between them and a purchase, and it is usually asked in the evening.

A contact form answered the next working day does not solve it. By then they have either bought elsewhere or lost interest, and the store will never know the order existed, because nothing was abandoned and nothing was measured. This is the quietest leak in ecommerce and it does not appear in any report.

An agent answering immediately, by phone, chat or text, closes that gap. The questions are usually simple and answerable from the product data and the shipping rules. The value is entirely in the timing.

The same capability absorbs the post-purchase volume that eats a small team's day: where is my order, can I change the address, how do I return this. Those are not sales conversations and they should not be treated as opportunities to sell, but they are most of the contact a store receives, and handling them quickly is the difference between a customer who orders again and one who does not.

The rule is the same in both directions: answer the question, do not turn it into a pitch.

An ecommerce customer did not ask for a relationship, they asked for a product, and every line of this respects that.

Every call opens by saying what it is. The agent identifies itself as a digital assistant contacting them on behalf of the store and says the call is recorded. Every call, every state.

On a pre-purchase question it answers the question. What is in stock, what the delivery window is, how the sizing works, what the return policy actually says, whether two things are compatible. It answers from the store's own data rather than improvising, and when it does not know, it says so and gets a person instead of guessing, because a made-up delivery date is a complaint with a delay on it.

On a high-value abandoned cart it is light and it offers a reason to be calling. We noticed the order did not complete, was there something in the way. Then it either solves the thing or it does not, and either way it stops.

On a service contact it resolves rather than sells. Nobody chasing a late parcel wants to hear about a related product.

What it does not do: it does not invent stock, dates or policy. It does not offer a discount nobody authorised. It does not call somebody repeatedly about an abandoned cart, because two calls about a basket is harassment rather than service. And it honours contact preferences as a system rule, immediately, which in a business with this much volume has to be automatic.

Most stores spend almost everything they have on the first order and almost nothing on the second, which is the wrong way round.

A customer who bought once has told you more than any advert can: what they wanted, what they paid, when they bought and whether anything went wrong. If what they bought is consumable, replaced periodically or part of a set, the next order has a date attached to it that you can work out from the first one.

That is the difference between a promotion and a reason. A store sending the same offer to everybody is guessing. A store contacting somebody about the thing they actually bought, around the time they would plausibly need it again, is being useful, and it only needs the order history it already has.

The other side is the customer who used to order and stopped. In a subscription business that is visible and usually handled. In an ordinary store it is invisible: nobody cancels, they just quietly stop, and the report that would show it is a report nobody runs.

Wholesale and trade accounts deserve separate mention. A store with any trade side is dealing with businesses that reorder on a rhythm, place larger orders and are worth contacting by phone in a way individual consumers are not. Those accounts are usually managed by memory, they fade for the same reason every other relationship in this guide fades, and they are the highest-value list an ecommerce business has.

Use this on us as hard as on anybody else.

Ask whether it can tell a browse from a serious cart, and what it does differently for each. If the answer is that it treats them the same, it will annoy your customers and you will blame the channel rather than the tool.

Ask whether it reads your store's real data: stock, shipping rules, order status, return policy. An agent answering product questions from anything else is guessing on your behalf, and the guess arrives as a complaint later.

Ask what it does when it does not know. It should say so and fetch a person.

Ask how many times it will contact somebody about one cart. If there is no cap, that is your answer.

Ask whether it handles service contacts without turning them into sales conversations, and listen to a real recording of one.

Ask how contact preferences are enforced and how fast they take effect. At ecommerce volumes, overnight is not good enough.

Ask whether it can act on order history as a trigger rather than only on a date.

Ask whether it treats a wholesale account differently from a consumer.

Ask what it stores, where it lives and how you export it. And get in writing who answers for it when a call goes wrong on an order somebody has already paid for.