SMS marketing for ecommerce: the complete guide

Some catalogues suit this channel and some do not, so answer that question before you buy anything. The stores where text earns its place have at least one of four things.

The first is urgency. Products that sell out, drops with a start time, restocks people are waiting for, and delivery windows that a customer needs to act on. A text is the right tool when the message stops being useful in a few hours.

The second is a short reorder cycle. If people rebuy from you every few weeks, a well timed reminder is genuinely helpful and the message has a clear job. If they buy once every three years, there is nothing to say between purchases that could not wait for an email.

The third is a young or mobile heavy audience who already communicate this way. You can see this in your analytics before you assume it.

The fourth is a high enough order value that the cost per message is comfortably covered. Every text costs money to send, unlike email where the marginal cost is close to nothing, so a store with a small typical order needs a much sharper reason to send.

If none of those apply, put the effort into email first and revisit later. A store with weak email flows and a new text programme has usually built the expensive channel before the cheap one, and the results reflect that.

And if some of them apply, be honest about which. A store where the only real use is order and delivery updates should run those and skip promotional texting entirely. That is a perfectly good outcome, and the delivery notifications will be some of the most appreciated messages you ever send.

Consent for text is a separate permission from consent for email, and it has to be given explicitly for text. Somebody who ticked a box about your newsletter has not agreed to receive messages on their phone, and a phone number captured for delivery updates is not permission to send promotions.

The rules vary by country and change, so check the current requirements for every market you send to and take advice where the stakes are real. The principles are stable even when the detail moves. Consent must be knowing and specific, the person must be told what they are signing up for and roughly how often, the cost position must be clear, and opting out must be easy and honoured immediately.

Build the sign up so it can be evidenced. Record what the person saw when they agreed, when they agreed, and through which form. If a complaint ever arrives, that record is the only thing that matters, and reconstructing it afterwards is not possible.

Keep the permissions separate in your data. Email consent, marketing text consent and transactional text consent are three different flags. Merging them because it is convenient is how stores end up sending promotions to people who only ever wanted to know where their parcel is.

Honour opt outs instantly and across every list. When someone replies with a stop word, they are out, everywhere, including any flow currently running. Continuing to send after an opt out is the fastest way to turn a customer into a complaint.

Finally, treat consent as something that can go stale. Someone who signed up two years ago and has never engaged is a risk rather than an asset. Removing them protects your ability to reach the people who do want to hear from you.

Carriers decide whether your messages arrive, and they increasingly require the sender to be registered before traffic will flow reliably. This is administrative work with a lead time, so start it early rather than discovering it on the morning of a launch.

Registration generally means identifying your business, describing the kind of messages you intend to send, and showing how you collect consent, including the exact wording on your sign up form. Have your business details, website and consent language ready before you begin, and expect the review to take days rather than minutes.

Choose the number type deliberately. A local number reads as personal and works well for conversational messages and smaller volumes. A toll free number handles higher volume and is treated differently by carriers. A short code is expensive and suits high volume programmes. In markets that use alphabetic sender identifiers, those come with their own registration and their own rules about replies.

Whatever you choose, use the same number consistently. Recipients recognise a number they have saved, and switching numbers means every message arrives from a stranger.

The deliverability rules here are stricter than in email. Messages that look like a template used by many senders, links from shortening services shared with other businesses, and content in restricted categories can all be filtered before anyone sees them. Use a link domain that belongs to you, keep the message specific to your brand, and understand which categories your carrier treats cautiously.

Build in a check that messages are arriving. Have a phone on each major network in your market and send yourself a test with every campaign format. Filtering can be silent, and reporting will happily show a message as sent when it was never delivered.

The best sign up moments are the ones where the person is already typing their details for a reason they understand. Bolting a phone field onto a form that had none is what makes signups feel like a demand.

Checkout is the strongest moment. The customer is already entering a number for delivery, and a clearly worded, separate opt in for marketing messages sits naturally beside it. Keep it unticked and keep the wording plain about what they will receive.

The second moment is a sign up form on the site offering something the channel is uniquely good at. Not a generic discount, which people will happily take by email, but the thing text does better: first access to a drop, a low stock alert, a restock notification for a specific product. Ask for the number in exchange for speed, and the people who join are the people who will act on the message.

The third is the moment someone wants a notification anyway. Back in stock, waitlist and delivery updates are all requests to be interrupted, which makes them the cleanest permission you can hold.

Avoid two things. Do not make the phone number a required field for something unrelated, because you will collect fake numbers and damage your own list quality. And do not assume that a number collected for delivery is a marketing subscriber, however tempting the volume looks.

Set expectations at sign up. Tell people roughly how often they will hear from you and what about, then keep to it. A subscriber who knows they will get a few messages a month about restocks does not resent the fourth one. A subscriber who was told nothing resents the second.

Finally, welcome new subscribers immediately with a short message that confirms what they signed up for and how to leave. It sets the tone and it catches wrong numbers early.

Apply one test to every message: would this person be glad you interrupted them for it. Most of what stores put in text fails that test and would have been perfectly acceptable as email.

The messages that pass are usually time bound or personal. A product they asked about is back. Something in their cart is running low in stock. Their order has shipped or is arriving today. The drop they signed up for opens in an hour. A reminder that the thing they buy every six weeks is probably running out. Each of these is either about their order or about something they asked to be told.

The messages that fail are the ones with no deadline and no personal connection. General brand news, a blog post, a range that has not changed, a discount with no end date. None of those are bad messages. They are email.

Order and delivery notifications deserve their own note, because they are the most valued messages in the whole channel and many stores never set them up properly. Confirmation, dispatch, out for delivery and delivered, each with tracking, reduce support contacts and make people feel looked after. They also keep your number familiar so that when you do send a marketing message it arrives from a sender they recognise.

Be careful with urgency you have manufactured. Real scarcity, a genuine deadline and an actual restock all work. Repeatedly claiming a last chance that turns out not to be one teaches people to ignore you, and this is the channel where that lesson gets learned fastest.

A useful rule of thumb is that if you would hesitate to send it to your ten best customers individually, do not send it to five thousand people at once.

Timing carries more weight here than in any other channel, because a text arrives immediately and cannot be deferred by the recipient the way an email can. Get the hour wrong and a good message becomes an annoyance.

Use local time, always. A send scheduled in your own timezone will reach somebody in the middle of the night, and in some markets sending outside defined hours is not merely rude but restricted. Set the quiet hours in your platform and respect them for everything except genuinely urgent delivery updates.

Avoid the edges of the day even inside permitted hours. Very early morning and late evening messages get read but they get resented. Mid morning and early evening tend to sit better, though your own data will tell you more than any general guidance once you have a few months of sends.

Frequency is where most programmes fail. This channel has a much lower tolerance than email, and the right number for most stores is measured in a few messages a month rather than a few a week. Increase gradually, and watch opt out rate per send rather than in aggregate, since the message that caused the damage is usually identifiable.

Count every message the person receives, including automated ones. A subscriber in a cart recovery sequence who also gets the weekly promotion and a delivery update has heard from you three times, and none of the three teams sending them was looking at the total.

Finally, tie timing to the customer's behaviour wherever you can rather than to your calendar. A replenishment reminder timed from their own last order will always outperform a scheduled send to everybody, and it costs the same to deliver.

The first few words decide everything, because that is what shows in the notification. Front load the point, then the detail, then the link.

Say who you are immediately. Recipients do not have your number saved, so a message that opens with a mysterious offer reads as spam. Leading with the brand name solves this in three words and costs nothing.

Keep it to one idea and one link. There is no space for a second thought, and two links in a text produce fewer clicks on both. Send the person to the specific product or collection they need, not to the home page, since the message promised something specific.

Write the way a person writes. Short sentences, ordinary words, no marketing register. Text is a conversational medium and copy that sounds like a billboard sits badly on a screen otherwise full of messages from friends. That said, do not fake familiarity either. Pretending a bulk message is a personal note is transparent and people find it irritating.

Mind the length. Longer messages get split into multiple parts, which can cost more to send and can arrive out of order on some networks. Emoji and certain characters change how the message is encoded and can cut the available length sharply, so check the count in your platform rather than in a text editor.

Include the way out. Depending on your market, opt out wording may be required, and it is good practice regardless. Put it at the end where it does not compete with the message.

Finally, read the message aloud before sending. Anything that sounds strange spoken sounds worse on a phone, and this thirty second check catches most of what goes wrong.

Run the two channels from one plan, otherwise the same customer receives the same thing twice within an hour and both channels suffer. The split should be by message type, not by who is available to write it.

A workable division is that text takes anything time critical or transactional, and email takes anything that benefits from length, images or browsing. A restock alert is text. The story behind the restock is email. A dispatch notification is text. The care instructions for the product are email.

Inside automated sequences, alternate rather than duplicate. A cart recovery sequence might send an email first, then a text a few hours later if there is still no order, then a final email. Each step should check whether the previous one worked before firing, and the whole sequence should stop the moment the person buys.

Suppress across channels. Somebody who has just clicked an email about the drop does not need the text about the drop. Somebody who replied to a text with a question should not receive the automated follow up while they are waiting for an answer.

Use text where email is failing, not everywhere email works. If a segment of your list has stopped opening email entirely, a small number of text messages can reach them. That is a good use of the channel. Adding a text to every email you already send is not.

Measure them together as well as separately. The question is whether adding text increased total revenue, not whether the text messages produced revenue, because some of what text earns would have come through email anyway. Comparing periods, or holding a portion of the audience back from the text, is how you find out.

People reply to texts, and that is an advantage rather than a problem to be engineered away. A customer who answers a message is in a conversation with your store, and those conversations convert at a rate that no broadcast can match.

So make sure replies go somewhere a human reads. Decide before launch who monitors the inbox, during which hours, and what the expected response time is. An unanswered reply is worse than never having invited one, because the customer now knows you are not listening.

Set up the automatic handling for the words that need it. Stop and its variants must opt the person out immediately. A help request should return your contact details. Beyond that, keep automation light, since a bot answering a real question about sizing does more harm than a slower human reply.

Read what comes back. The replies to a promotional text are a free stream of feedback about your products, your prices and your delivery, written in the customer's own words at the moment of decision. That is the same material that makes good product page copy and good ad copy, and almost nobody mines it.

Where volume allows, use the channel for genuine two way service. Answering a question about fit, confirming an address change, or sorting out a delivery problem over text is fast for the customer and cheap for you. Stores that do this well find the channel earns its place through service before it earns anything through promotion.

One caution. If you cannot staff replies at all, choose a number type and wording that make that clear, and point people to the channel where you can help. Silence with no explanation reads as indifference.

Unlike email, every message here has a real unit cost, so the honest measure is revenue per message sent minus what the send cost. A campaign that produced a healthy looking revenue figure can still have lost money once the cost of reaching everyone who ignored it is counted.

Track four numbers per send. Messages delivered, since sent and delivered are not the same and the gap tells you about filtering. Click rate, which tells you whether the message earned attention. Orders and revenue attributed. And opt out rate, which is the cost you pay in future reach.

Opt out rate deserves particular attention because it compounds. Losing subscribers on a send that also made money is still a loss if the losses accumulate faster than your sign ups. Track list size net of opt outs alongside revenue, and treat a send with an unusually high opt out rate as a signal to examine rather than a cost of doing business.

Separate automated from broadcast. Automated messages tied to behaviour usually carry the programme, and averaging them with promotional sends hides a weak promotional strategy behind good automation.

Be sceptical about attribution. Text platforms typically credit orders from anyone who received a message within a window, which will overlap heavily with email and ads. The clean test is a holdout: keep a portion of the eligible audience out of the send and compare their behaviour with those who received it. That number is smaller than the reported one and it is the number worth acting on.

Finally, review the programme as a whole every quarter. List size, cost, net revenue, opt out trend and what you actually sent. If the list is shrinking and the messages are getting more promotional, the channel is being spent rather than built.