Customer Referral Analysis SOP

A customer referral analysis measures how many of your new customers arrive because an existing customer recommended you, and it is one of the hardest product-market fit signals to fake. Setup takes a day or two, a reading needs four to eight weeks of data, and a week of calls explains the numbers. A survey shows what people say they would do, and this SOP counts what they actually did. People recommend products that make them look good to their friends, so real word of mouth is proof the product delivers.

What is customer referral analysis and when should you use it?

Customer referral analysis tracks organic word of mouth (unprompted recommendations) and invited signups, then studies who refers, why and with what result. It answers three questions: what share of growth comes from customers, which customers spread the product, and whether referred customers turn out better or worse than the rest.

It differs from the NPS survey, which asks how likely people are to recommend you. Saying "9 out of 10" costs nothing; sending a colleague a link costs a little reputation. Comparing both shows how much of the stated intent turns into action.

Research backs the value of referred customers. Philipp Schmitt, Bernd Skiera and Christophe Van den Bulte studied a German bank's referral programme in the Journal of Marketing and found referred customers were more loyal and brought in more profit than customers acquired through other channels.

Use this SOP:

• After the Sean Ellis test shows a fan base, to see whether fans bring others.
• Before building a referral programme, so you reward behaviour that already exists.
• When acquisition costs rise and you want to know how much growth is free.
• For products used in teams or shared outward, where spread is built into the use.

What do you need before you start?

• A "How did you hear about us?" field at signup or checkout.
• Invite links or referral codes if the product lets users invite others, tracked as events.
• Customer data with signup date, plan and retention status, to compare groups.
• At least 100 customers, so referral counts are big enough to read.

How do you run it, step by step?

1. Add the attribution question

At signup, ask "How did you hear about us?" with options (search, social media, ad, article, a friend or colleague, other) and a short text box. Make the text box required when someone picks "a friend or colleague", and ask for the referrer's name or company. Analytics tools miss word of mouth because recommendations happen in chats, meetings and dinners.

2. Tag existing customers

For customers who joined before the question existed, use what you have: referral codes, sales notes, first-call notes. Email the rest one question. Partial data still gives you a baseline.

3. Calculate the referral share every month

Divide new customers who came through a recommendation or invite by all new customers that month. Track it next to paid acquisition. A rising referral share while ad spend stays flat is one of the strongest fit signals you can show an investor.

4. Find who refers

Calculate the referring customer rate: the share of customers who brought at least one new customer. Then split by segment, plan and tenure. Referrals concentrate in a small group, and that group describes your best customer more precisely than any persona exercise.

5. Calculate the viral coefficient

For invite-based products, the viral coefficient is K = invites sent per user × share of invites that convert. A K above 1 means each user brings more than one new user and growth compounds on its own. Most products sit below 1, and even a K of 0.3 lowers your acquisition cost by bringing at least three extra customers for every ten you pay for.

6. Interview referrers and the people they referred

Call five referrers and five referred customers. Ask referrers what made them share, at what moment and in which words. Ask referred customers what convinced them. The exact sentences become your marketing copy.

7. Compare referred customers with the rest

Compare retention and revenue per customer for referred and non-referred groups over the same period. If referred customers stay longer, every referral is worth more than its acquisition saving, and a programme earns its cost.

Template you can copy

SIGNUP QUESTION
How did you hear about [Product]?
( ) Search engine   ( ) Social media   ( ) Ad
( ) Article/podcast ( ) A friend or colleague recommended it
( ) Other: ______
If "friend or colleague": Who recommended us? (name or company) ______

REFERRAL ANALYSIS SHEET
Tab 1: Monthly sources
Month | New customers | Via recommendation | Via invite link |
Paid | Other | Referral share %

Tab 2: Referrers
Customer | Segment | Plan | Months as customer | Customers referred |
Moment of sharing (from call)

Tab 3: Viral coefficient (invite products)
Active users | Invites sent | Invites per user (i) |
Invites converted | Conversion (c) | K = i x c

Tab 4: Referred vs non-referred
Group | Customers | Retention month 3 | Revenue per customer | Referred others %

REFERRER CALL QUESTIONS
1. When did you first tell someone about [Product]? What was happening?
2. What exactly did you say?
3. Who else would you tell, and what would stop you?

Worked example: a shared shopping list app for families

This example is illustrative. A shared shopping list app has 1,800 paying households. The founder adds the attribution question and invite tracking, then reads eight weeks of data: 420 new paying households, 151 of them via a recommendation or invite. Referral share: 36%.

Only 9% of households referred anyone, and 70% of those referrers are parents with two or more children. Invites: 0.6 per active user, 25% conversion, so K = 0.15. Calls reveal the sharing moment: a partner or grandparent gets added to the list and then sets up their own household.

Referred households retain at 64% at month 3 against 49% for the rest. The founder runs a growth experiment on a two-sided reward, a free month for both households, aimed at large families only.

What should you measure, and when do you stop?

The 20% threshold is this SOP's working rule for a clear word-of-mouth signal. Stop the analysis phase after eight weeks of clean data and ten calls. If referral share stays under 5% while the Sean Ellis score is high, people love the product privately; find out what makes sharing awkward before adding rewards.

Which mistakes should you avoid?

• Launching a referral programme before measuring organic referrals. Rewards on a product nobody shares buy a few signups and nothing lasting.
• Trusting analytics "direct" traffic as word of mouth. It mixes bookmarks, typed URLs and broken tracking. Ask people.
• Counting invites sent as referrals. Only converted invites count.
• Ignoring referral quality. If referred customers churn faster, rewards attract the wrong people.
• Copying another company's reward. Dropbox's two-sided storage bonus worked because storage was the product. Reward with what your customers value in your product.

How can AI speed this up?

Export the free-text answers to "How did you hear about us?" and ask an AI assistant to sort them into channels, flag named referrers and count each group. Ask it to cluster referrer call notes by the moment of sharing and pull exact phrases. Check the sorting on 30 random answers, because AI places "saw it on a friend's Instagram" in social media when the friend was the real reason.

Where does this SOP fit in your validation plan?

Referral analysis follows the Sean Ellis test and the NPS survey in the Product-Market Fit stage. Related steps:

1. Add referral share to the referral stage of your AARRR dashboard.
2. Test a reward or invite flow with the growth experiment SOP.
3. Factor free referral growth into customer acquisition cost validation.

See all Product-Market Fit SOPs. The F/MS social media module shows how to make sharing easy, and the free elevator pitch generator helps you turn your referrers' words into a pitch everyone can repeat.

FAQ

What is a good referral rate for a startup?

This SOP treats 20% or more of new customers from recommendations and invites as a clear word-of-mouth signal. Compare against your own trend more than against other companies.

How do I calculate the viral coefficient?

Multiply the average number of invites each user sends by the share of invites that become new users. A K of 1 or more means each user brings at least one more.

How do I track word-of-mouth referrals?

Ask "How did you hear about us?" at signup with a "friend or colleague" option and a field for who referred them. Analytics alone can't see recommendations made in conversation.

Should I offer rewards for referrals?

Only after you see organic referrals. Rewards speed up sharing that already happens; they rarely start it. Test a reward as a growth experiment and compare the quality of rewarded referrals.