Turning CRM automation into a dynamic lifecycle strategy

Automated journeys were already part of the CRM programme, but they had largely been built as individual flows rather than one connected lifecycle strategy.

Most relied on simple triggers: a customer joined, abandoned a basket, purchased, or reached a fixed number of days since their last order. The automation worked, but it wasn't adapting enough to differences in customer behaviour, value, product interest or likelihood to purchase again.

Automation should respond to the individual customer, not force every customer through the same predetermined journey.

That meant connecting onboarding, retention, lapse prevention and reactivation into one lifecycle ecosystem, supported by behavioural signals, stronger segmentation, AI-led decisioning and personalised channel selection.

As the strategy matured, the automation stopped being a supporting layer and became the majority engine of CRM performance.

The biggest opportunity wasn't simply building more flows. It was making the automation ecosystem more intelligent.

Journeys were often triggered by fixed time periods, customers with very different values could receive the same message, email remained the default channel, and product recommendations weren't always closely connected to customer behaviour.

Broadcast and automated activity could also overlap, while lapse was often addressed after it had happened rather than predicted beforehand.

I treated automation as a customer decisioning system rather than a collection of triggered emails.

For me, a strong journey needed five things.

• A clear customer or commercial objective
• Meaningful behavioural or transactional triggers
• Dynamic progression based on customer actions
• Personalised content, timing and channel
• Measurement based on incremental behaviour, not just engagement

The strategy also needed to work as an ecosystem. A welcome journey didn't simply end after its final email — customers moved into first-purchase conversion, post-purchase onboarding, retention, lapse prevention or reactivation depending on what they did next.

The customer lifecycle became the organising principle, rather than individual CRM campaigns.

Four connected stages, each handing the customer to the next rather than running in isolation.

Onboarding

Onboarding was designed around moving customers from unknown, to known, to first purchase, to second purchase.

Welcome and preference capture adapted content to acquisition source, browsing behaviour, market and stated preferences. Customers who hadn't yet purchased were progressively personalised using product and category interest, while browse and basket recovery was coordinated with onboarding so customers didn't receive conflicting messages. After the first order, messaging shifted towards product education, complementary categories, loyalty and reasons to return.

The objective wasn't simply to welcome customers. It was to identify intent quickly and accelerate the behaviours associated with stronger lifetime value.

Retention

Once customers had purchased, the focus shifted towards frequency, relevance and customer value.

Recommendations used purchase and browsing affinity rather than generic bestsellers, while next-purchase activity reflected each customer's normal buying cycle. High-value customers could also receive different messaging, benefits and contact pressure.

The goal was to create relevant reasons to return rather than relying on blanket discounting.

Lapse prevention

One of the biggest changes was moving from reacting to lapse to predicting it.

Instead of defining every customer as lapsed after the same number of days, I introduced signals based on individual purchase patterns and engagement trends. Someone who normally purchased every 30 days but hadn't returned for 50 represented a different risk from someone who typically purchased twice a year.

This allowed CRM to intervene before the customer had fully disengaged, using relevant product or content triggers before moving towards stronger incentives.

Reactivation

Reactivation focused on customers who had genuinely moved beyond their expected buying pattern. I separated audiences by previous value, purchase frequency, product affinity, engagement and discount sensitivity.

Not every customer received a discount. Customers still showing intent could be reactivated through product relevance or newness, while stronger incentives were reserved for audiences where historic behaviour suggested they could materially change conversion.

Three things sat underneath every journey: how customers were segmented, what signals the automation listened to, and how AI turned those signals into decisions.

The segmentation model

Segmentation moved beyond demographics and simple "days since purchase" rules. The model combined ten dimensions rather than one.

This meant a high-value customer whose engagement had recently declined was treated differently from a one-time promotional buyer, even if both technically sat within a "lapsed" segment.

The signal architecture

The automation was powered by a combination of customer and commercial signals.

Customer signals

Commercial signals

The journey responded to both customer intent and business context — there was little point recommending a product the business couldn't profitably sell or ship.

How AI was used

Predictive models and scoring were used to identify churn risk, purchase propensity, product affinity, expected next purchase and customer value. Those outputs influenced journey entry, customer priority, recommendations and timing.

The biggest value of AI wasn't simply generating copy. It helped answer four questions.

Who should we contact, when should we contact them, what are they most likely to want next, and what action is most likely to change their behaviour?

The biggest shift was moving away from static rules towards progression driven by what the customer actually did.

Declining engagement could trigger lapse prevention earlier, and a returning customer could be pulled forward rather than waiting for the next scheduled step. Timing became customer-led rather than calendar-led.

Channel strategy

Channel selection worked the same way. Email supported richer content, push was useful for immediate behavioural moments, SMS was reserved for higher-intent opportunities, and onsite personalisation continued the journey when customers returned.

The aim wasn't to use every channel, but to use the channel most appropriate for that customer and that moment.

Paid media as part of lifecycle CRM

I also extended lifecycle thinking beyond traditional CRM channels. CRM audiences were activated across Meta and Google to support retention and reactivation.

High-value lapsed customers could be prioritised, recent purchasers suppressed from unnecessary acquisition activity, and affinity audiences used to support second-purchase or category strategies.

This created one customer strategy across owned and paid channels, rather than two teams talking to the same person with different intentions.

Automated journeys were evaluated on whether they actually changed customer behaviour.

Opens and clicks remained useful diagnostic metrics, but the more important question became a harder one.

Did the journey cause something that wouldn't otherwise have happened?

Where possible, control groups and holdouts were used alongside the measures that reflect real behaviour change.

This helped distinguish genuinely effective automation from journeys simply capturing customers who were already likely to buy.