How to Improve Customer Loyalty: Choose Actions and Test What Changes
Published:
Last Updated:
Category: Marketing Glossary, CRM, LTV & Customer Management
Authors: Shusaku Yosa
Improving customer loyalty starts with understanding why customers return and what makes continued use difficult. Choose a customer problem, make a focused change, and examine behavior, preference, and cost separately. More messages or bigger rewards do not automatically create a stronger relationship.
1. Separate repeat behavior from preference
Customer loyalty concerns a customer's willingness to keep choosing a business or service. Purchase history provides evidence of behavior, but it does not explain the reason. Someone who values the service, someone who stays because switching is difficult, and someone waiting for a discount can have similar transaction records.
Combine purchases, renewals, or completed uses with questions about choice and unresolved problems. Recommendation intent is another piece of evidence. When using Bain's explanation of NPS, remember that a stated willingness to recommend is different from a recorded referral. Neither measure stands for the entire relationship.
2. Define the customers and observation window
Start with a specific situation. Customers who never complete their first use need a different investigation from established customers approaching renewal. Choose an observation period that fits the purchase or usage cycle. A person who has not yet had a reasonable opportunity to buy again should not automatically be labeled lost.
Write down eligibility, the starting event, the observation window, and the definition of completion. For example, observe every eligible first-time user for 30 days after completing their first use. Include only people with a complete follow-up window in that comparison. Use support records and short conversations to understand reasons, keeping unanswered questions separate from satisfied responses.
3. Match the action to the identified problem
The following table is a practical selection aid, not a ranking of proven winners. Select an action because it addresses evidence about a particular obstacle.
Situation | Possible action | Change to examine | Burden or risk |
|---|---|---|---|
Customers stop during initial setup | Explain the essential steps and a completed example | Eligible people completing setup | Excessive reminders |
Customers lose sight of their use case | Ask about the task and suggest a relevant use | Resumption of useful activity | Promoting unnecessary features |
A support problem remains open | Assign an owner, deadline, and completion criterion | Problems resolved and confirmed | Counting replies instead of resolution |
Purchases occur only during discounts | Investigate reasons beyond price | Understanding of choice and ordinary-period behavior | Reward cost and margin |
Low activity alone is not a reason to send more email. Customers may not understand the benefit, may be waiting for a suitable occasion, or may need a product problem fixed. If the reason is unknown, a small discovery exercise can be a better first action than a large campaign.
4. Remove obstacles before adding rewards
A confusing process, uncertainty about delivery, or an unanswered complaint is not repaired by a reward. Address the underlying problem and explain the change to affected customers. Ask whether the proposed resolution actually meets their need.
If you introduce a benefit, specify eligibility and its end date. Separate collecting a reward from receiving value from the service. Referral programs can involve rewards, production, administration, and support costs. A rise in referrals does not by itself demonstrate that advertising is unnecessary or that profit has increased.
5. Read a worked example with costs
This fictional before-and-after example follows two groups of 200 first-time users for 30 days after initial completion. A returning person is counted once if they use the service again during that window.
Measure | Before | After | Interpretation |
|---|---|---|---|
Eligible people observed | 200 | 200 | The same selection rule |
People returning | 120 | 130 | People, not visits |
Return rate | 60% | 65% | A difference of 5 percentage points |
Additional support cost | ¥0 | ¥40,000 | Fictional incremental operating cost |
Other changes | None recorded | A price change | The intervention's effect is not isolated |
The relative difference is 5 divided by 60, or approximately 8.3%. That is different from a five-percentage-point difference. The ten-person increase cannot automatically be attributed to the intervention: price and customer composition may also matter. Dividing ¥40,000 by ten would therefore not establish a causal cost per additional returning customer.
Examine choice reasons, unresolved problems, and opt-outs separately. A higher return rate does not establish stronger affection, greater satisfaction, or improved profitability.
6. Decide what would justify continuing
Before launching, define a useful decision rule. You might continue if first-use completion improves without increasing customer burden, revise the action if conversations reveal a different obstacle, or stop an unwanted contact. Compare equivalent observation windows and customer conditions.
With small samples, report the actual counts and eligible population alongside rates. Avoid declaring success from a short fluctuation. Repeatedly investigating reasons, removing obstacles, and checking the result makes the next improvement more specific and defensible.
Related reading
