On this page
Use three signals together: a unique code per creator, UTM links where you can, and a “How did you hear about us?” question after checkout. Codes and links undercount, because people forget codes and don’t click links in podcasts. The survey catches what they miss. Where buyers name a channel much more often than your last-click data shows, that channel is undercounted.
Why these channels look worse than they are
Someone hears your brand on a podcast while driving. Two days later they type your name into Google and buy. Shopify records “Google search” or “direct”. The podcast gets nothing, and at renewal time it looks like a waste of money.
The same happens with creators (watched on TikTok, bought later on a laptop), with friends (“you have to try this”), and with group chats. None of it leaves a click.
What each method catches
| Method | Catches | Misses |
|---|---|---|
| UTM links | Clicks from links you control (bio links, emails, ads) | Podcasts, word of mouth, anyone who searches instead of clicking |
| Creator discount codes | Buyers who remember and use the code | Buyers who forget it; inflated by code-sharing sites |
| Shopify’s last click | The last tracked visit before the order | Everything that happened before that visit |
| Post-purchase question | What the buyer remembers, including offline | Precision: memory is fuzzy, and some don’t answer |
No single row is the truth. Where they agree, you can trust the number. Where they disagree, you learn something.
A worked example: did the creator pay off?
You paid a creator $600 for a video. Here’s what each method says after 30 days:
The code alone says 14 orders, roughly $800 in sales: a loss after product costs. The survey says 33 buyers named the creator, and since only part of buyers answer, the real number is likely higher. Same campaign, opposite decision.
To get “typed the creator’s name”, ask the follow-up “Which creator?” only to buyers who picked Influencer or creator.
Compare what buyers say with last click
HearAbout (our app) records, for each answer, the last click Shopify saw before the order, and puts the two side by side.
How to read it:
- Buyers name it far more than last click shows (podcasts, friends, creators): undercounted. Give it more credit than your dashboard does.
- Last click shows it far more than buyers name it (often Google search or direct): the closer, not the opener. People searched your name after hearing of you somewhere else.
- Both roughly agree (often email, paid search): trust your existing numbers.
Last click can take about an hour after an order to appear, and some orders never get one. Only orders where both are known are compared.
Look past the first order
Some channels bring buyers who come back. HearAbout’s Value by channel after 90 days shows what buyers from each channel spent in total, counting their later orders, once they’ve had 90 days to order again. Friend-or-family buyers often look modest on the first order and strong at 90 days.
Guest checkouts can’t be linked to later orders, so they’re not in this number.
HearAbout by Centubo. Ask how buyers found you, compare it with Shopify's last click, and see value by channel after 90 days. Free during launch.
Install on ShopifyGetting cleaner answers
- Put the creator or podcast in the survey only while you’re working with them, so old answers don’t linger.
- Randomize the answer order.
- Don’t give a reward for answering while you’re judging a campaign: some people will tap anything for a code.
- Compare equal periods: 30 days before the campaign, 30 days during.
Questions people ask
Aren't discount codes enough to track influencers?
Codes only count buyers who remember and use the code. Many people watch a creator, search your brand days later, and check out without it. Codes also leak to coupon sites, which credits the creator for sales they didn't make.
Can a survey answer be wrong?
Yes, people misremember. That's why it works best next to your click data. When buyers name a channel far more often than your analytics credit it, that channel is undercounted; small differences aren't worth acting on.
How many answers do I need before deciding?
Enough per channel that one or two answers don't swing the share much. As a rough rule, wait for at least 30 answers in the channel you're judging, and compare periods of the same length.