How can I measure word-of-mouth and dark social referrals?
Published by Zigpoll. Updated
Measuring word-of-mouth and dark-social referrals requires combining direct questions to customers (surveys) with technical tracking (UTMs, short-links, referral codes and server-side joins); neither method alone catches everything. Use a short post-purchase or post-session attribution question plus reconciliation against your analytics and order fields to quantify what analytics misses. (zigpoll.com)
How can I measure word-of-mouth and dark social referrals for my online store?
Ask customers (post-purchase or in a follow-up) “how did you hear about us?” and match their answer to the order and analytics data; supplement with shareable links, referral codes, and server-side redirects so clicks carry a source when possible. Surveys capture the explicit recall signal that analytics often misses, Zigpoll found 51.8% of orders had no usable tracked source in store analytics (September 2026 data from 3,237 stores). Analytics still matters: UTM parameters, referrer, landing page and server logs give session detail you can join to survey answers. (zigpoll.com)
When to prefer one or the other
- Surveys work best when you can ask at purchase or soon after; tracking improvements help when you control the shared link (affiliate, promo, short link). If you sell mostly offline or in small B2B deals, surveys may be harder to scale.
| Option | What it detects | Strength | Weakness |
|---|---|---|---|
| Post-purchase attribution survey (e.g., Zigpoll) | Customer-reported referrer (word-of-mouth, DM links, names) | Direct human answer; catches dark social and offline referrals. Integrates with orders. | Recall error; needs response capture. (docs.zigpoll.com) |
| Shareable links + UTM / short links | Clicks from shares that include parameters | Precise when used; short URLs increase copy/paste use | People often share plain links or use in-app viewers that strip referrer. (semrush.com) |
| Referral codes / promo codes | Purchases using a code given by a referrer | Easy to attribute to a person | Requires referrer to give a code; friction for customers |
| Server-side redirects / link shorteners you control | Ensures referrer or campaign params are present | Preserves attribution across redirects/apps | Requires implementation and maintenance |
| Social listening / brand mentions | Public word-of-mouth (mentions, reviews) | Finds public recommendation signals | Misses private chats/DMs; not linked to orders |
What survey questions reveal who referred a customer?
Start with one simple forced-choice plus an “Other / Please specify” free text. Example: “How did you first hear about us? (Friend/family, Instagram, TikTok, Search, Email, Other, please specify).” Follow with optional name/link: “If a person referred you, who? (name or profile link), optional.” Ask one short verification question only if ambiguous: “Did you click a link to get here?” Keep questions one-click where possible to maximize response rates. Zigpoll documents recommend the single attribution question and automatic normalization of free-text answers to reported channels. (docs.zigpoll.com)
Trade-offs
- Forced choices improve analytics but miss new channels; free text catches specifics but needs normalization and manual mapping.
Where should I place attribution questions in the customer journey?
Place the primary attribution question immediately after purchase (thank-you/confirmation page) or in the first post-purchase email; these touchpoints combine high recall and high completion rates. On-site microsurveys can capture pre-purchase intent but risk disrupting checkout. Shopify and survey vendors recommend post-purchase and follow-up flows as the least intrusive high-yield spots for attribution questions. (shopify.com)
When to change this
- If you observe low post-purchase response rates, move the question into a short email within 24-72 hours. For high-traffic sites, A/B test placement and wording.
How do I combine survey answers with analytics and UTM data?
Store each survey response with the order ID and any arrival fields (landing page, utm_source, referrer) so you can join rows. Reconcile like this: (1) if analytics/UTM shows a clear source, use that for session attribution; (2) if analytics is (direct)/(none) or empty, use the reported channel from the survey; (3) normalize free-text answers into canonical channels and names (map “my friend Sam” → “Word of mouth / Sam”). Use your survey tool or an API to export responses and join them to your orders table; Zigpoll’s API and attribution endpoints return order fields (utm_*, referrerUrl, landingPage) for cross-tabbing. Run reports showing how many survey-reported channels had no tracked source, Zigpoll’s Attribution Index shows substantial gaps by channel and vertical (for example, stores using Zigpoll lost track of 73.4% of word-of-mouth sales in Home & garden stores and 59.8% in Apparel). (apidocs.zigpoll.com)
Practical rules
- Prefer analytics when a session has a valid UTM/referrer. Use survey responses to reassign “direct” cases and to attribute private referrals. Keep a documented mapping and refresh it monthly; automate normalization where possible. What changes the approach: low order volume (manual joins okay), strict privacy constraints (less session linking possible), or a need for real-time attribution (server-side tagging and link control wins). (support.google.com)
