The core answer
Before you cut prices or rewrite the contract to make leaving harder, find out which customers are leaving, at what stage of their time with you, and why. Those three answers usually show the problem sits in one place, for example one customer segment from one channel cancelling within three months because delivery was late. Blanket promotions then cost money without fixing anything.
Work in this order: measure correctly (customers, revenue and cohorts); find causes from your data and from talking to customers who left; fix the cause, starting with the lowest-effort actions; then adjust the contract. A contract keeps renewals and cancellations orderly, but it cannot make a customer satisfied.
This guide is written for businesses that sell on a recurring basis, such as software, monthly services, memberships and standing service work. Information as of 21 November 2026.
Who should read which section
| Your stage | Sections to read |
|---|---|
| Before deciding (nothing changed yet) | Sections 1 to 3 to measure correctly, and Section 7 to see which actions pay off first |
| In progress (you have data and are looking for causes) | Sections 4 to 6 to split causes by segment and time on service |
| When the numbers miss the plan (you fixed things and churn is still high) | Section 8 to check whether the contract and SLA match what you really deliver, and Section 10, the checklist |
1. What is churn, and how do customer and revenue churn differ?
Churn is the share of what was lost in a period compared with what you had at the start. You need two views at once.
| Measure | How to count | What it tells you | How it can mislead |
|---|---|---|---|
| Customer churn | Customers lost divided by customers at the start of the period | Product appeal and overall satisfaction | Many small customers leaving may hurt revenue less than one large customer |
| Gross revenue churn | Monthly revenue lost from cancellations and downgrades divided by monthly revenue at the start | Real damage to revenue | Ignores upsell, so it looks worse than reality when existing customers buy more |
| Net revenue churn | Lost revenue minus extra revenue from existing customers | Whether the existing base is growing or shrinking | Can hide heavy customer loss if a few big accounts expand |
Points to watch when calculating:
- Fix the period (monthly or quarterly) and use the same one throughout. Do not compare a monthly rate with an annual one.
- Leave customers who joined during the period out of the denominator; they had no chance to cancel yet.
- Separate "cancelled by the customer" from "suspended for non-payment". They are fixed differently: the second usually sits in billing and reminders, not the product.
- Customers who downgrade or stop using the service but still pay do not show in customer churn. Track them separately as a warning sign.
2. Why use cohorts instead of an average rate?
An average over the whole base mixes old and new customers, so it cannot tell you when customers leave. A cohort groups customers by the month they started and shows how many remain after 1, 3, 6 and 12 months.
- If most of a cohort leaves in the first 90 days, the problem is usually onboarding, delivery, or expectations set during the sale.
- If customers leave near the end of the term, the problem is usually perceived value, no contact before renewal, or the renewal price.
- If cohorts after a certain month are consistently worse, look at what changed that month: sales channel, promotion, delivery method or team.
A plain spreadsheet is enough. You only need the start date, end date, segment and acquisition channel for each customer.
3. If there is no monthly subscription, when is a customer "lost"?
Businesses where customers order case by case, such as job-based services, parts or wholesale, have no cancellation date. You must define the point yourself.
- Find the normal buying cycle for each customer group from past sales, for example one group orders about every 30 days and another every quarter.
- Set a rule, for example no order for more than twice the normal cycle means "at risk", and more than three times means "lost". These multiples are examples; adjust them to your business.
- Write the definition down and use the same one every period. Do not change the rule when the numbers look bad.
- Separate customers who left for natural reasons, such as closing down or a project ending, from those who left unhappy.
A customer on an annual service contract whose use has dropped sharply should be counted as "at risk" before the term ends, not only after cancellation.
4. How do you find the causes: existing data and talking to customers who left
Start with the data you already have before sending any survey.
| Source | What to look at | Example signal |
|---|---|---|
| Usage data | Frequency, features used, number of users in the customer's organisation | Usage falling steadily for 2 to 3 months before cancellation |
| Complaints and support tickets | Volume, response time, resolution time, repeated issues | The same problem reported repeatedly and not resolved |
| Delivery records | Contract date versus actual delivery or go-live date | Delivery later than agreed, or unusable in the first week |
| Billing history | Overdue invoices, disputed amounts, requests to change payment method | Invoices begin to go overdue before cancellation |
| Sales notes | What sales promised before signing | The customer expected features or service levels the product does not have |
Then speak to customers who cancelled (an exit interview).
- Have someone other than the account manager ask; customers speak more openly.
- Keep it to 3 to 5 questions: the main reason for leaving, whether an alternative was in use when they decided, and what would have kept them.
- Record answers in the same categories every time (price, quality, delivery, support, no longer needed, moved to a competitor) so you can total them.
- Be wary of polite answers such as "we don't need it now"; there is often a deeper cause. Always compare with usage data and complaints.
5. Splitting causes by customer segment, time on service and delivery problems
A high overall rate usually hides several small problems that need different fixes. Build a table on three axes.
| Axis | Example split | Question it answers |
|---|---|---|
| Customer segment | Company size, industry, acquisition channel, package bought | Which segment cancels unusually often, and should we have sold to it at all (see choosing B2B customer segments) |
| Time on service | 0 to 30 days, 31 to 90 days, 91 to 365 days, over a year | Do they leave early because they cannot start, or late because they see no value |
| Delivery problem | Late, incomplete, wrongly installed, frequent change of account owner | Which delivery problems are followed by cancellation |
Common findings:
- Small customers drawn in by a low-price promotion cancel faster than those who arrive by referral. The issue may be the channel or the price used to attract them, not the product.
- Early cancellations tend to relate to delivery and getting started, which welcome steps and realistic delivery dates can fix.
- Late cancellations tend to relate to value, price, or a product that has not kept up, which needs deeper conversations with customers.
6. Hypothetical example: 6 percent of monthly customers leave each month
This is a hypothetical example, not a real case or business. All figures are round and illustrative only.
Company A sells monthly task-management software to SMEs. At the start of the month it has 200 customers and 1,000,000 baht of monthly revenue. During the month 12 customers cancel (80,000 baht), others downgrade (20,000 baht), and existing customers add 30,000 baht.
- Customer churn = 12 ÷ 200 = 6 percent per month
- Gross revenue churn = (80,000 + 20,000) ÷ 1,000,000 = 10 percent
- Net revenue churn = (100,000 − 30,000) ÷ 1,000,000 = 7 percent
Splitting by cohort and segment shows that 9 of the 12 were small customers from a low-price advertising campaign, and 8 cancelled within the first 90 days. Delivery records show these customers waited about 14 days for account setup, while sales had told them it would take 3 days. Exit interviews agreed: "we waited so long the team went back to the old way".
The hypothetical conclusion is that the problem lies in onboarding and what sales promised, not price. The suitable response is faster account setup, realistic delivery dates from sales and a go-live SLA in the contract, rather than discounting every customer. Real results depend on each business and its data.
7. Retention actions, ranked by effort
Start with the low-effort actions that match the cause. Do not begin with discounts or lock-in.
| Rank | Action | Effort | Fits which cause |
|---|---|---|---|
| 1 | Align what sales promise with what you can deliver, and write it in the quotation | Low | Customers disappointed by inflated expectations |
| 2 | A welcome and onboarding process for the first 30 days with a named owner | Low to medium | Early cancellations within 90 days |
| 3 | Reminders before card charges or invoices, and a fix path when payment fails | Low | Suspensions from expired cards or forgotten payments |
| 4 | Contact customers whose usage is falling before they cancel | Medium | Customers not seeing value or not knowing how to use the product |
| 5 | Resolve repeated complaints at the root, not one ticket at a time | Medium to high | The same problem keeps recurring |
| 6 | Change the product or service for the most common cancellation reason | High | Product does not meet the core segment's needs |
| 7 | Adjust packages and pricing to the segment (see SaaS pricing models and subscription contracts) | High | Price does not match the value that segment receives |
| 8 | Adjust the contract: minimum term, renewal terms, SLA | Medium, but only alongside the items above | Orderly renewals and customer confidence |
Discounts are for cases where the data shows price really is the cause, and for segments you truly want to keep. Do not discount every customer who asks to cancel, because others will learn that cancelling gets a lower price.
8. What can a contract do to keep customers?
A contract orders the relationship but does not make customers want to stay. Points to review:
| Topic | What to make clear | What to watch |
|---|---|---|
| Minimum term | An initial term that matches real set-up cost, such as installation and training | An overlong term makes customers hesitate at purchase and does not guarantee renewal |
| Renewal | Auto-renewal or confirmation, length of each renewal, renewal price | Auto-renewals the customer did not notice cause disputes; see auto-renewal contracts and notice-date registers |
| Notice period | How far ahead, by which channel, to whom | A cumbersome notice route annoys customers but does not make them want to stay |
| Service levels (SLA) | Delivery time, response time, resolution time, and the remedy if missed | Promising more than the team can deliver creates an obligation you will breach |
| Early-termination fee | Explain how the amount was reached | A court may reduce an excessive penalty, and it is worthless if the customer simply stops paying |
| Data export on exit | In what format and within how many days the customer gets data back | Locking in customer data damages reputation and may lead to disputes |
What a contract cannot do:
- Make an unhappy customer genuinely keep using the service. They may stop paying, stop using it, or argue that you did not deliver as agreed.
- Offset poor delivery. If the company does not do what it wrote, terms that bind the customer may be challenged.
- Make every one-sided term in a standard-form contract enforceable. Terms that are excessively unfair may be limited under the law on unfair contract terms, and if the customer is a consumer, consumer-protection rules may add limits. Have a lawyer review the template before use.
Before each renewal date, run a performance review with the customer ahead of time; see reviewing performance before contract renewal. If a large customer has already sent a termination letter, see termination letter from a major customer.
9. Using customer data to analyse churn and win customers back: a PDPA caution
Churn analysis uses customers' personal data, such as contact names, usage history and complaints. If the customer is an individual, or people at a customer organisation can be identified, keep these points under the Personal Data Protection Act (PDPA) in mind.
- Purpose. Use data for the stated purpose, such as improving service quality. Do not reuse data collected for one purpose for another without a basis.
- Notice. Tell customers in your privacy notice what usage data you collect and analyse and for how long.
- Opt-out. Win-back messages that are marketing should give customers an easy way to refuse further messages, and you must respect a refusal.
- Only what is needed. Do not collect data unrelated to the analysis, and limit who can see it.
For official detail and guidance see the PDPC. If your company uses customer data with AI tools, see company AI policy and customer data.
10. Checklist before and during the work
- Write down the definitions of "lost customer" and "at risk" and use the same ones every period
- Calculate customer churn and revenue churn (gross and net), split by segment
- Build cohorts by start month and look at 1, 3, 6 and 12 months
- Gather usage data, complaints, delivery records, billing history and sales promises
- Run exit interviews by someone other than the account owner and record causes in fixed categories
- Rank actions as in Section 7, starting with low-effort actions that match the cause
- Check the contract: minimum term, renewal, notice period, SLA and data export against what the team can really deliver
- Check the privacy notice, purposes and the opt-out for marketing messages
- Review the numbers monthly and note what you changed and when, to see whether it worked
Common misunderstandings
| Myth | Correction |
|---|---|
| Cutting price will lower churn | If the real cause is delivery or care, customers still leave, and others learn that asking to cancel earns a discount |
| A contract that is hard to exit keeps customers | Unhappy customers may stop using or paying and may argue you did not deliver as agreed; an excessive penalty may also be reduced by a court |
| Low churn means customers are satisfied | They may simply still be bound by contract. Check usage and complaints too |
| A business without monthly subscriptions has no churn | A customer who stops ordering is a lost customer; you only have to define it yourself |
| Exit interview answers can be taken at face value | Compare them with usage data, complaints and delivery records before concluding |
| Once we hold customer data we can message any of them to win them back | Marketing needs a stated purpose and an opt-out under the PDPA |
Further reading
- Choosing B2B customer segments in Thailand
- SaaS pricing models and subscription contracts
- Auto-renewal contracts and notice-date registers
- Reviewing performance before contract renewal
- Company AI policy and customer data
High churn is fixed by finding which customers leave, at what stage and why, and fixing that cause first; the contract's job is to keep renewals and exits orderly, not to stand in for a good product and service.
We help companies with the legal side of their business plan, including revenue models and customer segments, and we prepare and review service contracts, renewal terms, notice periods and SLAs so they match what the business can really deliver. We are not accountants or marketing consultants, we do not analyse your company's figures for you, and we cannot promise that customers will stay. Fees depend on the stage of the matter and the work required. Send us the documents first, and we will give you the scope of work and fees in writing before we start. Call +66 92 254 2045 or send us the details. See also business plan and contracts and documents.
This guide is published by Suwanvara Law Firm — a Khon Kaen law firm established in 1986. General information only, not legal advice on a specific matter.

