# E-commerce marketing automation: the 3 scenarios that pay

> An online shop with no CRM team does not need twenty scenarios. It needs three, set up in the right order, on reliable customer data. Which ones, and how to measure what they bring in.

- Original page: https://dotsland.com/en/ecommerce-marketing-automation-3-scenarios/
- Language: en
- Published: 2023-09-21T07:56:31+02:00
- Updated: 2026-10-19T18:23:05+02:00
- Section: Digital Marketing
- Publisher: Dotsland, https://dotsland.com/en/

---

## In brief

- **Three scenarios are enough in the first year.** Abandoned basket, welcome and first purchase, reactivation of dormant customers, in that order: the order of return on effort.
- **Data comes before the tool.** Without reliable customer identification and a clean purchase event, automation sends good messages to the wrong people.
- **Measure revenue, not opens.** A control group of 10% of eligible contacts is enough to know what the scenario really brings in.
- **The overbuilt machine shows early.** When nobody can say who receives what and why, it is time to cut back and restart from three scenarios.

We regularly open automation accounts containing a dozen active scenarios, built over two or three years by people who have since left the company. Nobody can say which one brings in what.

In these accounts, two or three scenarios produce most of the attributable revenue. The others consume per-contact budget and sender reputation, without anyone daring to switch them off.

An online retailer with no CRM team has no reason to aim for that complexity. They need three scenarios that run, clean data, and an honest way of measuring what they bring in.

## The 3 scenarios to set up, in this order

The order follows return on effort: how fast the scenario reaches a purchase intent already expressed. A message sent to someone who has just filled their basket works on a warm intent; one aimed at a contact who signed up eight months ago must first recreate it. Our [marketing automation support](https://dotsland.com/en/expertise/automation-workflows/) almost always starts with these three scenarios.

*The three scenarios in order of return on effort, with their trigger settings.*

| Scenario | Trigger | Timing | Number of messages |
| --- | --- | --- | --- |
| **Abandoned basket** | Add-to-basket by an identified person (known contact or email entered at checkout), with no order within the set time. Disarms as soon as an order is recorded, including from another device. | First message between 1 and 4 hours, never less than 30 minutes or you will chase a customer who is still paying. Second at 24 hours. A third at 72 hours is justified for high-value baskets, not for a €40 average basket. | Two, three at most. Beyond that, you are no longer recovering a sale, you are generating unsubscribes. |
| **Welcome and first purchase** | Newsletter sign-up, account creation or first order. Three entry points that call for three different contents, not one generic message. | First message within minutes, not the next day. The following ones at D+3 and D+7 for a prospect. For a first-time buyer, time the second after the estimated delivery, not after the order. | Three. The first reassures, the second explains what you do better than others, the third offers a way in (entry product, buying guide, trial). |
| **Reactivation of dormant customers** | No order over a period calculated from your real repurchase cycle: median time between two orders from your regular customers, multiplied by 1.5. Often 6 to 9 months for clothing, 6 to 10 weeks for consumables. | Two messages 10 to 14 days apart, then stop. With no response to either, the contact goes into dormancy. | Two. The first recalls what they bought and what has changed since. The second asks a direct question, with the option to unsubscribe cleanly. |

### 1. The abandoned basket

**Classic mistake:** putting a promo code in the first message. You teach your best customers that abandoning their basket triggers a discount, and you pay a reduction on sales you already had. Keep the discount test for the last message.

### 2. Welcome and first purchase

**Classic mistake:** sending the catalogue. A new subscriber does not yet know why they would buy from you rather than elsewhere. For a first-time buyer, the opposite mistake is to drown them in offers when they are waiting for information about their order.

### 3. Reactivating dormant customers

**Classic mistake:** defining dormant by email inactivity instead of the absence of purchases. A customer who orders twice a year without ever opening a message is not dormant, they are independent.

> **The mistake not to make**  
> Sending the reactivation to the whole dormant database at once. In a single send you reach your coldest addresses, those that concentrate closed accounts and spam traps, and the deliverability of all your other scenarios drops, order confirmation emails included. Spread it out in cohorts.

## Data comes before the tool

An automation scenario is a mechanical rule applied to data. If the data is wrong, the rule still applies and produces errors at scale.

Two foundations need checking before you subscribe. Identification means recognising that a basket created on mobile and an order placed on a computer come from the same person. Events are the add-to-basket, order, amount and refund signals, reported reliably.

When this base is shaky, the symptoms are always the same: basket reminders sent to customers who have ordered, welcome emails received by loyal customers, revenue counted twice. Browsers blocking third-party cookies has made these gaps more frequent than in 2020, as detailed in our analysis of [server-side tracking and measurement reliability](https://dotsland.com/en/tracking-server-side-lying-in-silence/).

The fix is server-side collection of commercial events, independent of the browser. Datafirefly Server-Side, published by Datafirefly Limited (same group as Dotsland), offers [a free plan of 10,000 requests a month](https://server-side.datafirefly.com/) to test this without commitment.

One decision rule: as long as you cannot rebuild last week’s revenue from your events, to within a few per cent of your back office, do not add a scenario.

## Consent and deliverability: what cannot be undone

A purchased database, or one collected in a hurry, does not produce mediocre sales. It lastingly damages your sender reputation, which also affects your transactional emails.

The mechanism is simple: the lists sold contain abandoned addresses that mailbox providers have recycled into spam traps. A single campaign is enough to tip your domain into filtering, and it takes months of clean volume to get out.

Since February 2024, Gmail and Yahoo have applied inbox access rules, gradually rather than all at once. The baseline applies to everyone: SPF or DKIM authentication. At Gmail, above 5,000 messages a day, DMARC is added, along with one-click unsubscribe honoured within two days and a complaint rate kept below 0.10% without ever reaching 0.30%. Yahoo applies the same principles but publishes no volume threshold and uses a single complaint threshold of 0.3%. Microsoft introduced a 5,000-messages-a-day threshold in May 2025 for authentication only. Since November 2025, Gmail has been tightening enforcement: temporary rejections, then permanent ones. We detail these rules in our [email marketing approach](https://dotsland.com/en/expertise/email-marketing-gdpr/).

On compliance, three points cover the essentials: a consent box that is not pre-ticked and is separate from the terms of sale, timestamped proof, and unsubscribes that take effect within 48 hours. We detail the framework in our articles on [managing personal data under the GDPR](https://dotsland.com/en/gdpr-day-to-day-marketing/) and on [the role of clear consent in digital marketing](https://dotsland.com/en/cookie-banner-design-and-wording/).

Add a dormancy policy: any contact who has not opened, clicked or bought in twelve months leaves mass sends. You lose displayed volume, you gain deliverability.

## Choosing a tool without falling into a trap

Demos all look alike. The differences appear after eighteen months, on four points.

*The four points on which automation tools part ways after eighteen months.*

| Point of divergence | What to check before signing | What happens if you do not |
| --- | --- | --- |
| **Cost per contact and how it rises** | Simulate the bill with three times your current database, then ask whether unsubscribed and inactive contacts are charged. | You pay to store contacts you no longer reach. |
| **Reversibility** | The list of what you get back when you leave: contacts, but also send history, events, consent dates and statistics per scenario. | A tool that only returns addresses makes you start measurement from scratch. |
| **Real integration with the CMS** | What the PrestaShop, WooCommerce or Shopify connector really syncs: orders, refunds, catalogue with stock, variants, frequency. | An integration that pushes contacts but not refunds has you chasing refunded customers. |
| **Getting data out day to day** | A documented API and webhooks, not just a monthly CSV export. | You cannot cross-reference your scenarios with your analytics tool when you decide to. |

One last reflex: test on your real data. A real database contains duplicates and malformed addresses that no demo dataset shows.

## What to measure, and how not to fool yourself

The open rate is no longer a performance measure. Since September 2021, Mail Privacy Protection (which the recipient turns on themselves, on iPhone, iPad, Mac or iCloud.com) routes remote content through two relays operated by separate entities, which loads the images and counts an open without anyone having read anything. Depending on the share of your recipients who have turned it on, a share you do not know, your rates are inflated by unknown proportions.

The only measure that holds is attributable revenue, over a fixed window announced in advance: seven days for most baskets, thirty days for considered purchases. That leaves the point almost nobody deals with: part of that revenue would have happened without the scenario.

> **What to remember**  
> The honest method fits in one sentence: keep a control group. Randomly exclude 10% of eligible contacts, send them nothing, and compare the revenue of the two groups adjusted to the same size. The gap is your real gain.

Two precautions make the calculation credible. The draw must be random and stable over time, not redone at every send. And you need to get above the noise: a few hundred contacts per group, over several weeks. The same reasoning applies to your journey tests, described in our article on [conversion rate optimisation](https://dotsland.com/en/cro-low-traffic-method/).

An order of magnitude observed in the accounts we audit, to be taken as such and not as a market statistic: abandoned basket most often concentrates the majority of attributable revenue, and reactivation the least. That is the reverse of the order in which they are built.

## The signs you have built an overcomplicated machine

The tipping point is not seen in the number of scenarios, but in the loss of control.

- **Nobody can say who receives what.** You have to open the tool and follow the branches to answer a simple question.
- **Some scenarios have not been touched for six months** and nobody dares cut them, for lack of knowing what they bring in.
- **A customer can receive two contradictory messages on the same day**, for lack of a priority rule between scenarios.
- **Time spent maintaining branches exceeds time spent writing messages.**
- **Decisions are made on open rates**, revenue per scenario no longer being calculable.

Going back is simpler than it seems. Pause everything except abandoned basket and welcome, wait three weeks, then look at total revenue: in most cases, it does not move detectably.

Then only reactivate what you can justify with a figure. Each scenario switched back on must have a named owner, a quantified objective and a review date. The rest stays off.

## Your first 60 days

The plan below works with one part-time person, on a shop that is already running.

### Days 1 to 15: make the data reliable before automating

1. Compare last week’s revenue in your back office and in your measurement tool. A gap above 10% comes before any scenario.
2. Check that add-to-basket and order events are reported with a stable customer identifier and an amount.
3. Audit consent collection: box not pre-ticked, separate from the terms of sale, timestamped proof.
4. Check SPF, DKIM and DMARC on the sending domain.

### Days 16 to 30: a single scenario live

1. Build the abandoned basket in two messages, at 1 hour and 24 hours, with no promo code.
2. Randomly exclude 10% of eligible contacts to form your control group.
3. Have five colleagues go through the full journey, on mobile and on a computer.
4. Set the attribution window (seven days) and write it down in a shared document.

### Days 31 to 60: measure, then add

1. At D+45, calculate the revenue gap between the exposed group and the control group.
2. If the gap is clear, build the welcome in three messages, with a separate entry for first-time buyers.
3. If it is not, add nothing: rework the content and timing of the abandoned basket.
4. Calculate your median repurchase time, for next quarter’s reactivation threshold.

Reactivating dormant customers comes in the third month: it requires a clean database and a measured repurchase cycle, which you do not yet have at day 60.

## Frequently asked questions

### Which marketing automation scenario should an online shop start with?

The abandoned basket. It reaches a purchase intent already expressed, can be set up in a day, and its effect is quick to measure. Two messages are enough: the first between one and four hours after abandonment, the second at twenty-four hours. Add the welcome next, then reactivation of dormant customers.

### Do you need a paid tool to start with marketing automation?

No, not for the first three scenarios. Several solutions run an abandoned basket and a welcome on a few thousand contacts without commitment. The real cost is not the licence, it is the time spent making the data reliable upstream. Choose on reversibility and on how the per-contact price rises.

### How long before you see results from an automated scenario?

The first attributable purchases appear in the days after an abandoned basket goes live. A serious judgement takes longer: four to six weeks and several hundred contacts per group for the comparison with a control group to rise above statistical noise.

### Can you send automated emails to a purchased database?

No. Beyond the legal consent problem, a purchased database contains addresses recycled into spam traps by mailbox providers. A single campaign can tip your domain into filtering, which also damages your order confirmation emails.

### How do you know whether your scenario really brings in revenue?

By randomly excluding a share of eligible contacts, around 10%, to whom you send nothing. Then compare the revenue of the two groups over an identical window, adjusted to the same size. The gap is the real gain. Without a control group, you attribute to the scenario sales that would have happened anyway.

## What remains to decide this week

Marketing automation does not reward those who automate the most, but those who know what each automation brings in. Three measured scenarios are worth more than twelve inherited scenarios that nobody dares switch off.

Without a CRM team, that is your advantage: you have no debt to carry. Check your purchase data, build the abandoned basket with a control group, and keep the other eleven ideas for next year.

**Unsure about your e-commerce automation scenarios?** We start by comparing the revenue in your back office with the revenue in your tools, before looking at a single scenario: that is where the messages sent to the wrong people are hiding. [Let’s talk](https://dotsland.com/en/contact-us/).