What Is Retention in Marketing: A Complete Guide to Customer Retention

Picture a bucket with a small hole in it. You can pour water in faster, but if it keeps leaking out, the water level never really rises. That’s what a business looks like when it keeps acquiring new customers but barely works with the ones it already has.

A customer places an order, signs up for a subscription, or registers on a platform. The sale happens. But the work doesn’t end there.

If that person never comes back, the company has to go find another customer all over again: run ads, drive traffic, handle inquiries, and repeat the entire path from first contact to a closed deal.

That’s exactly the problem retention marketing solves. A company keeps working with the customers it has already won: it tracks their purchase or usage history and picks the right moment for the next touchpoint.

Retention, in a marketing context, measures whether a customer keeps buying or using the product after that first deal. In this guide, we’ll cover how to calculate the core metrics, what sets Day N, Rolling, and Range Retention apart, which strategies actually help retain customers, and how to build a retention strategy from scratch.

• Retention marketing works with your existing customer base to drive repeat purchases, plan renewals, and other actions that matter to the business.
• Before you calculate retention, decide what counts as a “return” for your business and over what time frame.
• CRR and churn show retention and attrition, RPR reflects repeat purchases, and LTV and CAC help you understand customer economics.
• NPS and CSAT speak to customer experience, but they don’t replace behavioral data — a high score alone doesn’t guarantee a customer will buy again.
• Personalization, loyalty programs, onboarding, and triggered messages all work better when they respond to a customer’s specific situation, not a blanket blast to the whole list.
• Email, SMS, Viber, push, and CRM automation each solve different problems, and you can combine them in one system.

Put simply, retention marketing helps a company stay engaged with a customer after that first action, so they’re more likely to come back.

For an online store, that’s a second order. For a SaaS product, it’s regular usage or a subscription renewal. For an online course platform, it’s buying the next course.

The key thing here is to figure out which action genuinely signals an ongoing relationship.

Say a user signs up, logs into their account, and does nothing else. Technically, there was activity — but it means little for the business. If the product’s real value only kicks in once someone creates a project, connects an integration, or launches their first campaign, that’s the action worth tracking. Just logging in is too weak a signal.

The same logic applies to online stores. A visitor might come back to the site a few times after buying, but that’s not the same as placing another order. If the business goal is to grow repeat sales, what you should be analyzing is orders themselves and the time between them.

Start by defining what counts, for your specific business, as continued engagement or a new purchase. Different companies will land on different answers, and that’s fine.

Retention Rate, or the customer retention rate, shows the share of your customer base that stays active over a given period.

Churn Rate, or the attrition rate, shows the share of customers who, over that same period, stopped using the product by whatever rule the company sets.

Put simply:

• Retention tells you how many customers stayed.
• Churn tells you how many left.

What counts as “active” depends on the business model. For a subscription service, churn can be tied to a cancellation or the end of a paid period. For an online store, it depends on the purchase cycle: someone who buys a new phone once every year or two isn’t a lost customer just because they didn’t order again last month.

where:

• S: customers at the start of the period;
• E: customers at the end of the period;
• N: new customers acquired during the period.

For example: 1,000 customers at the start of the period, 1,150 at the end, 200 of them new.

CRR = (1,150 − 200) / 1,000 × 100% = 95%.

A simple version of Churn Rate:

Churn Rate = customers lost / customers at the start of the period × 100%.

If 50 out of 500 customers left during the period, churn is 10%.

The formulas themselves are simple. What actually matters is locking in your calculation rules ahead of time and not changing them month to month — otherwise, comparing periods becomes unreliable.

Retention problems rarely show up overnight. They usually build up gradually.

It’s worth digging into retention if:

• a large share of customers make only one purchase and never return;
• users lose interest in the product shortly after signing up;
• the share of customers renewing subscriptions or plans is lower than expected;
• your customer base is growing, but so is churn;
• acquiring new users keeps costing more;
• your CRM has piled up a large number of inactive contacts;
• everyone gets the same messages regardless of their history with the company.

In that situation, one more email blast won’t fix anything.

Say a store gets plenty of first orders but rarely sees a second one. It’s worth looking at how long it typically takes people to buy again, what they buy after that first order, and what messages they’re getting in between.

The job here is to figure out at what point customers stop buying and why. The cause could be the product itself, the service, poor timing on the next touchpoint, or simply that a repeat purchase isn’t something the customer needs that soon.

Day N Retention, also called Classic Retention, shows the share of users who completed the target action on a specific day after first using the product. It’s the same calculation under two different names. The denominator is the number of people who used the product for the first time.

Example: 100 people used a service for the first time on September 1. If 24 of them returned on September 2, Day 1 Retention is 24%. If 10 of them showed up on September 8, Day 7 Retention is 10%.

Someone who returns on day six or day eight, but skips day seven specifically, won’t count toward that number. That’s the whole point of the method.

Day N gives you a clear checkpoint. But that same strict cutoff can also be a limitation: a customer who returns on day 13 won’t count toward Day 14, even though they’re clearly still using the product.

Some analytics platforms, Amplitude among them, call this approach “Return On”: it counts users who took the target action on exactly the chosen day.

Day N works well for apps and services where you genuinely expect regular, daily-ish return visits. For an appliance store, a low Day 7 number doesn’t say much about churn on its own — a customer simply might not need another purchase a week later.

Classic Retention and Day N Retention are the same calculation.

In both cases, you’re checking whether the target action happened on a specific day after the first event.

That’s why, when comparing reports, it’s worth checking how the metric is actually defined rather than just going by its name.

Classic Retention is useful when a specific checkpoint date matters to the business. But for behavior where customers aren’t expected to return on an exact schedule, a different measurement approach may fit better.

Rolling Retention answers a different question:

Did the user return on the chosen day, or any day after?

For example, Rolling Day 7 counts someone who returned exactly on day seven, as well as someone who returned later — day ten, or even a month later — within the period being analyzed.

This calculation works well when customers aren’t expected to follow a strict schedule.

Say someone opens a trip-planning app, doesn’t touch it for a few days, and comes back only when it’s time to prepare the next leg of the trip. Here, the fact that they came back at all matters more than the exact date.

Range Retention looks at a window, not a single day.

Instead of asking “did the customer return on day seven?”, it asks whether they took the target action sometime during the second week after signing up.

Range Retention is typically calculated over a window of several days.

For an online store, this format can work better if repeat orders tend to cluster around a certain period rather than land on one specific date.

Beyond the method itself, you also need to decide how days are counted: by calendar date or by rolling 24-hour windows.

Pick your calculation method before you start analyzing. The two approaches can produce different numbers from the exact same data, so they shouldn’t be compared directly.

CRR shows the share of your original customer base that’s still active over a chosen period.

CRR = ((E − N) / S) × 100%

New customers get subtracted from the ending count, and the result is compared against the starting base.

For example: 500 customers at the start of the period, growing to 530 by the end, 80 of them new. That leaves 450 customers from the original base, putting CRR at 90%.

It’s worth calculating this beyond just the overall base. Breaking it down by customer group, plan, or acquisition channel can surface a problem that gets buried in the average.

Churn shows the share of customers who stopped using a product or service during the period.

Churn Rate = customers lost / customers at the start of the period × 100%

For subscription businesses, churn can be tied to a plan cancellation. For online stores, the period should account for the typical purchase cycle of that product category.

For B2B, it’s important to look beyond the number of lost customers to the revenue actually lost. Losing one large account can hurt the business more than losing ten small ones. That’s why it’s worth distinguishing customer churn from revenue churn.

LTV estimates a customer’s potential value over the entire relationship with a company.

A simplified formula for businesses with repeat purchases:

LTV = average purchase value × average number of purchases × customer lifespan

For example, if the average order is $100, a customer buys twice a year, and stays with the company for five years:

LTV = $100 × 2 × 5 = $1,000.

This isn’t profit — the formula doesn’t account for cost of goods, servicing costs, or other expenses. It’s an estimate of total revenue from one customer under a given model.

CAC shows how much it costs to acquire a new customer:

CAC = sales and marketing spend / number of new customers

Looking at LTV and CAC together is useful for weighing acquisition cost against the value a customer generates over time.

RPR shows the share of buyers who made a repeat purchase.

RPR = customers with 2+ purchases / total customers × 100%

For example, if 120 out of 400 buyers placed a second order:

RPR = 120 / 400 × 100% = 30%.

For an online store, this is a clear, practical metric — but it doesn’t explain why someone bought again. That’s why RPR is best viewed alongside the time between orders and data on what people actually buy on repeat.

NPS and CSAT speak to customer experience, not actual retention. That’s why they’re most useful alongside behavioral metrics.

NPS measures how likely a customer is to recommend the company. Customers respond on a 0–10 scale:

• 9–10: promoters;
• 7–8: passives;
• 0–6: detractors.

NPS = % promoters − % detractors

CSAT measures satisfaction with a specific interaction: a purchase, a delivery, a support request. On the common 1–5 scale, ratings of 4 and 5 count as positive:

CSAT = number of 4–5 ratings / total responses × 100%

The difference between these metrics is simple:

Retention shows whether a customer kept buying or using the product.

NPS — willingness to recommend the company.

CSAT — satisfaction with a specific interaction.

For example, if CSAT drops noticeably after a support interaction, it’s worth checking whether customer behavior changes afterward and whether churn rises among that group. A high score on its own doesn’t guarantee a customer will come back.

Personalization isn’t just a first name in the subject line. It’s far more useful to look at a customer’s history: what they bought, when, what they were interested in, and where they stalled.

A cosmetics brand might time a repurchase reminder to when a product typically runs out. A SaaS company might send a specific how-to guide to someone who signed up but never activated a key feature.

In both cases, the message shows up not just because “it’s time to send something,” but because the timing genuinely makes sense for that customer.

Email works well for instructions, recommendations, and curated content that needs context.

SMS is the right tool when you need to deliver a short message quickly or flag something time-sensitive. UniTalk’s SMS service supports targeted, transactional, and promotional messages with personalization, analytics, and CRM integration.

Points, cashback, tiered discounts, and special perks can all give a customer an extra reason to come back.

But a discount on its own doesn’t prove retention actually went up.

Take a clothing store. A customer was already going to place a second order, but got a 10% coupon anyway. The sale happened, and the company’s margin just shrank for no added benefit.

That’s why a loyalty program should be judged by its incremental effect: are customers buying more often, has the share of repeat orders gone up, has the time between purchases gotten shorter?

Framed that way, a loyalty program becomes one retention tool among others — not just a way to hand out points.

Before launching a program, it’s worth defining:

• what behavior you want to encourage;
• what the customer actually gets;
• when and how they can use the reward;
• how you’ll measure the result.

Sometimes people stop using a product simply because they never got to see its value.

A user signs up but can’t get the service set up. A customer subscribes but never finds the feature they needed. A buyer gets the product but doesn’t figure out how to use it.

Another promotional email isn’t going to fix that. The first step is removing whatever’s standing between the customer and the product’s value.

Onboarding has one job: get someone to a first, clear win as fast as possible.

For example, if a product is built for team collaboration and only one person signs up, it makes sense to immediately show them how, and why, to invite teammates.

Duolingo is a good example of building that kind of regularity: its daily streak feature helps people keep the habit going.

If a user finishes onboarding but still stops coming back, the problem may lie in the next stage of their journey. That’s the point to look again at which action actually reflects real product usage, and exactly where customers stop taking it.

Standard retention analysis shows you what’s already happened: a customer stopped buying, a user is logging in less, a subscription didn’t renew.

Predictive retention is about catching a behavior change before it turns into churn.

For example, a customer who used to order every month skips their usual purchase window. A SaaS user who used to run a core feature regularly barely opens it anymore.

A behavior change on its own doesn’t mean a customer is leaving. But it’s a reason to dig in: did they hit a problem, did their needs change, did they switch plans, or do they simply not need the product anymore?

AI can be genuinely useful once it’s too hard to spot these patterns manually across a large base — there’s too much data, and the signals of churn can be spread across several different actions.

Even so, you still need to find the actual cause. One customer needs a how-to guide, another needs help from a rep, a third just needs a nudge, and some shouldn’t get a message at all.

Before bringing AI into the picture, define what counts as active behavior and what signals churn. Otherwise, the model just learns from data where activity and churn were mislabeled from the start.

In practice, getting started doesn’t require a complex model. A few simple rules are enough — for example, flagging a customer as “at risk” if they haven’t logged in longer than their usual interval, or missed a typical purchase. Those threshold rules are easy to set manually, and it’s only worth moving to more sophisticated algorithms once you’ve accumulated enough real churn data.

Start with your customer base.

Split it into new, active, inactive, and reactivated (returned-after-a-pause) customers. You can also carve out your highest-value customers as a separate group.

Look at the date of the last order or action, purchase count, average order value, product, and acquisition source.

ABC analysis makes sense if it matters to split customers by their contribution to revenue. For a smaller base, this isn’t always necessary.

At this stage, the priority is pinning down a specific problem rather than working with vague generalizations.

For example: “Most customers place a first order, but the repeat-purchase rate is extremely low.”

Customer data is usually scattered: orders live in the store, calls are logged in your phone system, conversations happen in chat, and email reactions sit in a separate platform. Without connecting that data, you can’t track a customer’s full path before and after a purchase.

The journey itself (Customer Journey) looks different depending on the business:

• E-commerce: first purchase → delivery → new need forms → repeat order.
• SaaS: signup → first meaningful action (the “aha moment”) → regular usage → subscription renewal.

CRM or CDP platforms pull these fragments into a single profile. That gives you the full interaction history, lets you analyze behavior in depth, and makes personalized communication possible.

A CJM breaks a customer’s path into sequential steps: from seeing an ad and landing on the site, to the first purchase, using the product, and ordering again. For each stage, it’s worth defining four things: what the customer does, what they expect, what could go wrong, and how the company responds.

For example, if first orders go smoothly but customers rarely come back, it’s worth studying the post-purchase stage closely. You need to understand when a new need actually arises for the customer, whether they get a relevant reminder at that moment, and whether they have any real motivation to buy again. That kind of analysis makes it clear exactly where along the journey a company loses the most customers. For more details on how to properly organize this process, see our article “What Is a Customer Journey and How to Create a CJM.”

Start rolling out a few basic automations:

• a welcome message after a first purchase or signup;
• a reminder about an unfinished action;
• a heads-up before a subscription renews;
• reactivation for inactive users;
• a recommendation for a related product (cross-sell).

Match a tool to each scenario. Email works when context matters. SMS is enough for a quick reminder, and Viber is hard to beat when you need visual attention — an image, a link, a button.

After launching your baseline scenarios, look at the results and decide what needs adjusting: the copy, the send time, the channel, the offer itself, or the audience segment.

A/B testing is the best tool for this. For example, one part of the audience gets SMS, another gets email, and a third — the control group — gets no message at all. Comparing the share of repeat purchases across groups tells you whether your outreach actually made the difference, or whether customers would have ordered anyway, with no extra nudge.

If a scenario proves itself and brings in additional sales, scale it up with confidence. If it doesn’t work, don’t just send more messages — revisit the original hypothesis, the segment settings, or the mechanic itself.

• Reducing retention to discounts alone. Promo codes can nudge a one-off purchase, but they don’t fix a weak product or a poor customer experience.
• Ignoring base segmentation. A brand-new customer and someone who hasn’t bought anything in six months need completely different approaches. Blasting everyone with the same message just burns through your list.
• Chasing vanity metrics like open rate and CTR. Opens and clicks only show a reaction to the text or subject line. The real marker of success is an actual repeat purchase or renewal.
• Confusing retention with loyalty. A customer might stick around out of plain habit, inertia, or a lack of easy alternatives. That doesn’t mean they’re loyal or willing to recommend you.
• Ignoring the natural purchase cycle. Someone who buys a refrigerator isn’t coming back for another one next month. For durable goods, a low Day 30 retention number is completely normal and doesn’t signal churn.
• Bringing in AI before the basic analytics are in place. AI models should only come in once you’ve clearly defined what counts as a target action and what counts as churn. Otherwise, the algorithm just automates whatever mess and errors already exist in your data.
• Skipping a control group. Without one, there’s no way to prove a purchase happened because of your outreach rather than a customer’s own organic need, which might have led to the same purchase anyway.
• Setting up automations and forgetting about them. Customer habits change. Scenarios and email sequences that converted well a year ago need a regular audit.

Before launching any new retention mechanic, honestly answer two questions: what specific change in customer behavior are you going for, and which metric will tell you it actually happened?