What a Referral Customer Is Actually Worth

Standard customer lifetime value math asks what a customer will spend with you over the life of the relationship. For a home service business it produces a number that is not just incomplete but systematically wrong, because it ignores the customers your customer brings you.

Key Takeaways

• Conventional lifetime value counts direct spend only: which understates any customer who refers.
• Referral value is the larger term for low-frequency trades: the customer who never buys again still sends people who do.
• Roughly half of new referrals come from repeat referrers: so the first referral is an entry point, not an event.
• Referred customers refer at a higher rate: the chain does not decay as fast as intuition suggests.
• Fixing the math changes what you are willing to spend: on service, on retention, and on the program itself.

What the standard calculation misses

The usual formula multiplies average job value by expected purchases by gross margin. For a subscription business that captures most of the truth. For home services it leaves out the part that matters most.

Consider a moving customer. They spend once, at a good margin, and by conventional math the relationship is closed. Lifetime value equals one job.

Except that over the following years they mention you to three people who are moving, one of whom books. That booked job is worth as much as the original, and it cost you a bounty instead of a lead acquisition cost.

By conventional accounting, that second job is attributed to “referrals” as a channel, and the customer who generated it is still recorded as worth one job. The math is not slightly off. It has assigned the value to the wrong place, which means every decision downstream about how much to invest in that customer is made from a number that is missing the larger half.

A version that includes the chain

The corrected framing has three terms:

Direct value. What they spend with you. For recurring trades this is substantial. For once-a-decade trades it is essentially one job.

First-order referral value. The expected number of customers they refer, times the value of those customers, minus what you paid in bounties. Note that a referral costs you a bounty rather than an acquisition cost, so the margin on referred work is structurally better than on purchased leads.

Second-order referral value. The customers those customers refer. This term is usually dismissed as too speculative to count, and that is a mistake. It is where the compounding lives.

To make this concrete, take your own numbers: average job value, your margin, and the share of customers who refer at least once. Multiply out the three terms. Even with conservative assumptions on the referral rate, the second and third terms typically exceed the first for any low-frequency trade, which is the whole point of running the exercise. (Plug in your own figures; the arithmetic here is illustrative, not a benchmark.)

Why the chain does not decay the way you would expect

The instinct is that each generation of referrals should be dramatically weaker than the last, which would make the second-order term negligible.

Two findings from Snoball’s data cut against that.

First, roughly 50% of new referrals come from someone who has already referred at least once. A customer who refers is not a one-time event. They are a source. Having been through the experience and seen how easy it was, they start actively watching their own network for other people to send you. Your expected referrals per referring customer is therefore well above one, and the standard model that assumes a single referral per person understates it badly.

Second, the first referral is the expensive one. It takes three to four touchpoints on average to get there. Everything after is cheaper, which means the return on a referral relationship is back-loaded, and judging it at the first ask, which is when most companies judge it, is judging it at the least informative moment.

There is also a compositional effect worth noting: referred customers arrive already trusting you, which tends to make them better customers: easier to sell, less price-sensitive, more likely to be satisfied, and therefore more likely to refer in turn.

What changes when you fix the number

This is not an accounting exercise for its own sake. Three decisions move.

What you will spend to acquire a customer. If a customer is worth one job, your acceptable acquisition cost is small. If they are worth their job plus the chain behind them, you can outbid competitors still using the narrow number, and win the customer.

What you will spend to keep service quality high. The extra crew member, the better packing materials, the time spent making a difficult job right. Under narrow math those look like margin erosion on a closed transaction. Under corrected math they are investments in the referral chain, which is the largest term.

What a lost customer actually costs. A botched job does not cost you one job. It costs the referrals that customer would have sent, and their referrals. This is the calculation that makes service recovery obviously worth doing rather than a judgment call.

What a program is worth. Most companies evaluate referral programs against the referrals produced. The honest comparison is against the chain those referrals start, which is why programs that look marginal in month three often look obvious at month eighteen.

Do the arithmetic once

Pull three numbers: average job value, gross margin, and the share of last year’s customers who sent at least one referral. Most companies cannot produce the third, which is itself the finding. If you cannot measure who refers, you cannot value them, and you are almost certainly underinvesting in your best customers.

For more, see referral program ROI, going deep with your best referrers, and how referral networks compound.

Referral behavior figures from Snoball’s own data across 300+ moving companies. Worked arithmetic is illustrative. Substitute your own figures.