顾客留存、服务与复购

A new customer and an existing customer do not cost the same to reach. Every seller has heard a version of this claim, often with a dramatic ratio attached. The exact gap depends on category, channel, margin and service quality, so this lesson treats any single number as an illustration rather than a statistic. The useful question is not 'how much cheaper is retention' but 'what makes the existing customer come back'. The answer is usually unglamorous: the parcel arrived when promised, the size was right, the product page told the truth, and when something went wrong, someone replied.

§3.1.2 defines retention as building strong relationships and raising loyalty or stickiness to a brand, supported by personalised advertising and customisation. Operationally, three measures matter. Repurchase rate asks what share of buyers bought again within a window. A cohort retention curve follows a group of first-time buyers over time. Service response measures how quickly and completely a problem is handled. Together they describe a relationship, not a transaction. A coupon can raise one month's repurchase and still leave the relationship unchanged.

Read a retention curve carefully. Take 1,000 first-time buyers in January. If 180 buy again in month 1, period retention is 18%; if 90 more buy in month 2, cumulative retention is 27%; if 40 more buy in month 3, cumulative retention is 31%. The period rate falls while cumulative retention rises, and both are correct. The curve usually declines and then flattens; the flat part is the loyal core. Two warnings: the window must match the natural repurchase cycle, and the cohort must be fixed. Recomputing a rate on a changing mix of old and new customers can make retention look better while every cohort is getting worse.

Service is where retention is won or lost. A failure is a moment when the customer tests whether the seller is competent and honest. Service recovery has three parts. Apologise specifically, naming what went wrong rather than saying 'sorry for the inconvenience'. Compensate appropriately - a replacement, a partial refund, a return label - within a range that front-line staff can authorise. Then remove the root cause: if the wrong size keeps shipping, fix the size chart or the picking process, not only the complaint. Speed matters because uncertainty is worse than a known delay. A slow reply followed by a generic coupon is a double deviation: the customer is disappointed twice, and the second disappointment is about how the seller handled the first.

Email and messaging lists are the textbook's clearest example of a channel with a role. §3.3.4 says email is sent to an opt-in audience, click-through is typically 3%-4%, sending is almost free, and the main cost is the list at about 5-20 US cents a name, compared with 75-80 cents for a direct-mail postcard. The same section concludes that email works well for maintaining customer relationships but poorly for acquiring new customers. That conclusion should decide where the channel sits in the plan. Use it for post-purchase guidance, replenishment, service follow-up and member events; do not expect it to fill the top of the funnel. Consent is not a formality: a purchased list without permission is both ineffective and improper, and spam filters will remove much of it.

Affiliate marketing sits on the other side. §3.3.5 describes pay-for-performance marketing with four players: the advertiser (merchant or brand), the network, the publisher (the affiliate) and the customer. The advertiser pays a commission, typically 4%-20%; Amazon Associates pays up to about 10%, and eBay's partner network about 1%-6%. The publisher places a tracked link, the network records the sale and handles payment, and the advertiser bears the return risk. This model buys acquisition, not retention: the audience and the relationship belong to the publisher. It also has a margin arithmetic that students often skip. A 20% commission on a product with a 30% gross margin leaves 10% before logistics, returns and overhead; on a 15% margin it is negative. Check the contribution, not only the sales.

Private traffic is often sold as the answer to rising acquisition costs, and it does lower the media cost of repeat contact. §3.5.1 defines private traffic as owned by the firm, with repeated use adding no cost, and describes conversion through activities, promotions and one-to-one chats. Those activities are work. A community needs moderators, content, service replies and someone to handle complaints at the weekend. The media bill falls, the payroll does not. A community can also decay: members mute the group, the same ten people reply, and the reach that looked owned becomes a small, tired audience. Private traffic is therefore a cost shift and a management task, not a free asset.

Three clarifications close the common confusions. First, retention is not a coupon. A discount can buy a transaction, but it also teaches customers to wait for the next discount, and it lowers the margin on orders that would have happened anyway. Second, a good review is not retention. Reviews measure a moment of satisfaction, are biased toward extreme experiences, and may be influenced by incentives; repeat purchase is observed behaviour. A store can have excellent ratings and a one-time customer base. Third, service recovery is not unlimited compensation. Compensation has to be proportionate and capped, or it attracts abuse and rewards the loudest complaint rather than the most serious failure.

Measure retention with cohorts and money. Group customers by acquisition month, track 30-, 60- and 90-day repurchase, and compare the curve before and after a change. Segment by first product, channel and whether the first order was discounted, because a discount-acquired cohort often retains worse. Then compare the contribution from repeat orders with the cost of the retention programme - email tools, service staff, community moderation and compensation. A programme that raises repurchase by two points but costs more than the incremental margin is not a retention strategy. Lifetime value can help, but only with stated assumptions; a precise LTV built on three months of data is false precision.

The lesson is not that advertising is bad or that service is everything. Acquisition and retention are different jobs with different tools. Affiliates and paid reach bring the first order; email, personalised contact and service recovery protect the second. The funnel's last counter, repurchase, is therefore not a marketing afterthought. It is the score of whether fulfilment, product information and service kept the promise the acquisition campaign made. Lesson 15 turns to the product side of that promise: assortment and the long tail.