> For the complete documentation index, see [llms.txt](https://docs.zigpoll.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zigpoll.com/guides/what-brands-have-learned.md).

# What Brands Have Learned with Zigpoll

What thousands of Zigpoll surveys have found, across the whole customer base and brand by brand: which channels get missed, what gets a survey answered, why visitors leave, and what brands did with th

The guides in this section tell you what to ask. This page is about what came back when brands asked, and what they did next.

Everything here comes from one of two places:

* **Aggregate data across the Zigpoll customer base.** The [Attribution Index](https://www.zigpoll.com/attribution-index) (3.3 million order-linked "how did you hear about us?" answers from 3,237 stores), the [Response Rate Benchmarks](https://docs.zigpoll.com/response-rate-benchmarks.md), and a September 2026 study of every post-purchase and exit intent survey with meaningful traffic. These numbers are pooled and anonymous; no individual store can be identified from them.
* **Published case studies.** Where a brand is named, the result comes from a case study the brand took part in. The link is next to each one.

***

## Post-purchase attribution: the channels you fund are not the channels that work

### Across the customer base

When customers are asked at checkout how they heard about a brand, the answers over the twelve months to September 2026 looked like this:

| Stated channel       | Share of answers | Stores reporting it |
| -------------------- | ---------------- | ------------------- |
| Instagram            | 19.4%            | 2,292               |
| Word of mouth        | 17.0%            | 2,552               |
| Facebook             | 16.9%            | 1,888               |
| Search               | 14.0%            | 2,447               |
| Don't know / other   | 8.1%             | 2,411               |
| YouTube              | 4.6%             | 1,151               |
| TikTok               | 3.8%             | 1,304               |
| Creator / influencer | 1.2%             | 288                 |
| Podcast              | 0.8%             | 113                 |
| Press / editorial    | 0.8%             | 275                 |
| AI assistant         | 0.5%             | 371                 |

Word of mouth is the second-largest channel in ecommerce, and it is the one no analytics tool can see. The same responses were matched against the source the store's own analytics recorded for that order:

* **51.8% of orders arrived with no usable source** (direct, unknown, or absent).
* Those customers were not forgetting. Only 8.3% of them answered "don't remember," against 7.3% of customers whose source was tracked.
* The untracked half skews toward channels that produce no click. Word of mouth made up 19.3% of untracked answers against 14.6% of tracked ones. YouTube 5.4% against 3.8%. Facebook ran the other way: 13.1% untracked against 21.1% tracked.
* Per channel, analytics lost **60% of YouTube** and **59% of word-of-mouth** sales, against **40% of Facebook** sales. A recommendation is one and a half times as likely to vanish from your reporting as a Facebook visit.
* **AI assistants** went from 0.23% of stated discovery in December 2025 to 1.10% in August 2026, a 4.8x rise. The number of stores seeing at least one AI-attributed order in a month rose from 59 to 237. Still small, but the fastest-moving line on the board.

The gap holds in every vertical. Three examples from the per-vertical tables:

| Vertical                                                                                         | Stores | Responses | Orders with no source | Word of mouth lost | Facebook lost |
| ------------------------------------------------------------------------------------------------ | ------ | --------- | --------------------- | ------------------ | ------------- |
| [Supplements & health](https://www.zigpoll.com/attribution-index/2026-09/supplements-health)     | 129    | 225,203   | 43.3%                 | 47.1%              | 34.8%         |
| [Beauty & personal care](https://www.zigpoll.com/attribution-index/2026-09/beauty-personal-care) | 190    | 186,931   | 44.8%                 | 51.9%              | 31.2%         |
| [Baby & kids](https://www.zigpoll.com/attribution-index/2026-09/baby-kids)                       | 75     | 102,251   | 48.0%                 | 50.5%              | 38.1%         |

### What brands did with it

* [**OMHU**](https://www.zigpoll.com/blog/omhu-case-study) (Danish furniture) found that 20% of sales came from word of mouth and recommendations from past customers, a number their media reporting had never shown. They now raise and lower budgets by market based on the survey, and are investing in referrals.
* [**Dalfilo**](https://www.zigpoll.com/blog/dalfilo-case-study) (Italian linens, four country stores) saw around 9% of customers arriving from friends and family, and put money into a referral programme to grow the number.
* [**Jolly Mama**](https://www.zigpoll.com/blog/jolly-mama-case-study) (French maternity nutrition, Shopify Plus) learned that 5% of customers found them through press, a channel they had assumed was for awareness only. It justified keeping their PR agency. The same survey showed how many customers were health professionals, which became its own marketing channel.
* [**Kanga Coolers**](https://www.zigpoll.com/blog/kanga-coolers-case-study), with 60–70% of traffic paid, allocates 10–15% of budget each quarter to experimental channels and uses the survey to decide which ones to keep.
* [**Manucurist**](https://www.zigpoll.com/blog/manucurist-case-study) (nail polish, 10,000+ responses a month) follows social-media answers with "Do you remember where in particular?" to separate organic, influencer and paid. A Good Morning America appearance showed up as "GMA" write-ins the same day.
* [**TabooGrow**](https://www.zigpoll.com/blog/taboogrow-case-study) measured Reddit at 8.2% of sales for a client, a channel with no pixel to argue the case for it.
* [**NextPercent**](https://www.zigpoll.com/blog/nextpercent-case-study) (CRO agency, DACH) installs one fixed-choice question plus an optional text field on every client store and treats it as the referee no ad platform can inflate.

### What to copy

One **Single Choice** question first, with specific options (Instagram, TikTok and YouTube rather than "social media"), word of mouth as its own option, and "Other" with a write-in enabled. Compare the shares against your ad platforms' numbers every month; the channels where the survey share far exceeds the platform share are the under-funded ones. See [How to Measure Marketing Attribution](https://docs.zigpoll.com/guides/how-to-measure-marketing-attribution.md).

***

## Post-purchase: why they bought, what almost stopped them, who it's for

### Across the customer base

Of the 3,620 Shopify thank-you-page surveys with meaningful traffic in the 90 days to September 2026, 29% open with an attribution question. The rest lead with the shopping experience, purchase motivation, the customer's age or situation, or whether they have bought before. Most ask between one and five questions; a three-question survey is the single most common shape.

The study also measured which open-ended questions actually get typed into. When an optional text question is shown after a one-tap question, the share of people who answer it depends almost entirely on how specific the ask is:

| Optional text question                           | Answered when shown |
| ------------------------------------------------ | ------------------- |
| "Anything else you'd like to share?"             | 18%                 |
| "How can we improve?"                            | 27%                 |
| "What almost stopped you from buying?"           | 51%                 |
| "Why did you buy today?" / "What convinced you?" | 60%                 |

A specific question is answered about three times as often as a generic closer.

### What brands did with it

* [**Tumeri**](https://www.zigpoll.com/blog/tumeri-case-study) (turmeric skincare) asked "What made you buy from Tumeri today?" Two findings: customers kept citing reviews that said the product was non-staining, so reviews were moved onto the key landing pages; and customers kept naming melasma, a condition the founder had to look up. The Kojic Acid Soap launched for it sold out twice.
* [**Kanga Coolers**](https://www.zigpoll.com/blog/kanga-coolers-case-study) asked "What almost stopped you from buying today?" and "Who is this for?" The buyer turned out to be older than assumed and to change by season (gifts for dads in June, grandparents buying for students in Q4). Matching landing pages and ad creative to the season raised landing page conversion 15–20% and ROAS 10%, and the objections surfaced by the first question led to new product lines that now drive six figures in incremental revenue.
* [**AMARO**](https://www.zigpoll.com/blog/amaro-case-study) (Brazilian fashion, around 8,000 responses a month) asked "Is this item for you or is it a gift?" About 20% of orders were gifts, which reshaped bundles, email copy and paid creative.
* [**Manucurist**](https://www.zigpoll.com/blog/manucurist-case-study) uses an open-ended question about the website to catch problems that customers hit and bought through anyway. It also asks which colours are missing, which is how they launch new shades knowing demand exists, now across 70+ colours.
* [**PAJ GPS**](https://www.zigpoll.com/blog/paj-gps-case-study) asked "How do you plan on using our product?" and found an animal-tracking use case they had never marketed. One affiliate partnership with a vet drove 200 extra sales in two months, and the answers now feed six use-case landing pages.
* [**Perfect Presents**](https://www.zigpoll.com/blog/perfect-presents-case-study) (gifting, Philippines) asks customers what to launch next, takes the top three requested products from the AI summary, and builds them. One of them, a bracelet, doubled Valentine's Day revenue year over year.
* [**Dalfilo**](https://www.zigpoll.com/blog/dalfilo-case-study) added a question about a missing bed-sheet size and another about gift boxes, and launched both once the survey confirmed demand.

### What to copy

Pick the one question whose answer would change a decision this quarter, put it second after attribution, and make its follow-up text box specific. Leave the text box required if its answer is what you need. See [How to Find Out Why Customers Buy](https://docs.zigpoll.com/guides/how-to-find-out-why-customers-buy.md), [How to Measure Gift Buying](https://docs.zigpoll.com/guides/how-to-measure-gift-buying.md) and [How to Build Your Ideal Customer Profile](https://docs.zigpoll.com/guides/how-to-build-your-ideal-customer-profile.md).

***

## What gets a post-purchase survey answered

### Across the customer base

The September 2026 study covered 3,433 thank-you and order-status surveys from 2,972 stores and 10.9 million presentations. The typical survey is engaged with by 29% of buyers; the top quarter clears 42% and the top tenth 53%. Two things explained almost all of the spread, and neither was design:

|                                         | Median engagement |
| --------------------------------------- | ----------------- |
| First question is a **Single Choice**   | 34%               |
| First question is a **Multiple Choice** | 33%               |
| First question is a **rating**          | 17%               |
| First question is an **open text box**  | 14%               |

The first question is the whole story. Colours, answer count, randomisation, title length and auto-advance moved engagement by five points or less, and those differences vanished when the same store was compared with itself.

The second finding is what a text question in the middle of a survey costs. Measured question by question over the 30 days to October 26, 2026, on thank-you page surveys (1,677 pairs of questions with no logic between them):

| Text question before the end | Answered when shown | Go on to the next question |
| ---------------------------- | ------------------- | -------------------------- |
| **Required**                 | 85%                 | 85%                        |
| Optional                     | 58%                 | 87%                        |

A required text box does not end the survey. Making it optional costs about a third of its answers and keeps about 2 more buyers in 100 for the questions after it. The rating question is not lost by being asked second: asked after a one-tap question it collects more answers per 100 buyers than when asked first, because more people are still there.

### What brands see

* [**Jolly Mama**](https://www.zigpoll.com/blog/jolly-mama-case-study) reports over 50% of buyers answering, helped by a surprise discount code revealed at the end of the survey.
* [**ThumbPRO**](https://www.zigpoll.com/blog/thumbpro-case-study) runs the thank-you page survey and a follow-up email 72 hours later and estimates that around 95% of customers answer.
* An [**8-figure beauty brand**](https://www.zigpoll.com/blog/beauty-brand-case-study) collecting 100,000+ responses a month sees about 80% of respondents continue past the first question.

### What to copy

Open with a click, ask text questions that are specific, and leave a text question required when its answer is what you need: it is answered far more often, and almost nobody stops at it. The [Post-Purchase Survey guide](https://docs.zigpoll.com/guides/how-to-build-a-post-purchase-survey.md) covers the order; the [Reward question](https://docs.zigpoll.com/tutorials/reward-question.md) is how the discount at the end is built.

***

## Exit intent: what makes visitors hesitate

### Across the customer base

Exit intent surveys are a different instrument from post-purchase surveys. In the 90 days to September 2026, 717 exit intent surveys from 644 stores were shown 12.1 million times and answered 197,000 times. The typical survey is engaged with by 1.6% of the visitors who see it, which sounds low until you remember who those visitors are: people hesitating on a product or cart page, caught by a pop-up at the moment they move to leave, and whom nothing else on the site was going to reach.

The first-question rule is even stronger on-site than on the thank-you page. An exit intent survey that opens with a **Single Choice** question engages 2% of leavers; one that opens with a text box engages 0.4%, five times fewer. Of the 29 stores running both kinds, the choice-first survey won in 29 of 29. Every extra question is also more expensive on-site: 70% of people who answer the first question see the second, against 83% post-purchase and 99% on a link or email survey. Put the question that matters first.

What brands ask is remarkably consistent. The most common first questions are "Why are you leaving without making a purchase?", "We noticed you left an item in your cart. Why is that?", "Was there anything that stopped you from making a purchase?" and "Did you find everything you were looking for today?" The most common answer options, across all stores:

1. Not enough information about the product
2. Couldn't find what I was looking for
3. The price is too expensive
4. Still deciding / planning to purchase later
5. Just browsing
6. Other (with a write-in)

### What brands did with it

* [**AMARO**](https://www.zigpoll.com/blog/amaro-case-study) targeted the exit survey at visitors with items in their cart and asked "Why are you leaving without checking out?" Over 50% were waiting for a coupon or discount. That is a very different problem from a usability one, and it told them exactly what the cart-abandonment email should contain.
* [**Melvin & Hamilton**](https://www.zigpoll.com/blog/melvin-hamilton-case-study) (leather shoes, Germany \~70% of revenue) saw German conversion fall after a redesign and assumed navigation was the cause. The exit survey said stockouts: visitors were leaving because their size was unavailable. The redesign was cleared, and the fix was inventory.
* [**Yay's Snacks**](https://www.zigpoll.com/blog/yays-snacks-case-study) asked "What kept you from buying today?" Price was expected. Flavour confusion was not: people did not know which flavour to choose. Content explaining the differences improved conversion.
* [**Partify**](https://www.zigpoll.com/blog/partify-case-study) (auto parts) asks "Leaving so soon? Tell us why!" with options for price, couldn't find my part, out of stock, and need more time. The CEO reports a significant conversion lift from acting on the answers, which are routed to Slack channels per topic.
* [**GAÂLA Paris**](https://www.zigpoll.com/blog/gaala-paris-case-study) (sustainable luxury fashion) A/B-tested the exit survey itself. A newsletter sign-up converted poorly; a variant pointing leavers at the limited-stock "Last Chance" collection was clicked by 6% of the visitors who saw it.
* [**Blikket**](https://www.zigpoll.com/blog/blikket-case-study) (CRO agency) used on-site quizzes and exit surveys, with discount codes gated behind completion, to drive a 2% absolute lift in conversion rate for a detergent brand.

### What to copy

One single-choice question with the honest options (including price), a write-in "Other", and a cart condition so the survey fires for people who matter. See [How to Find Out Why Visitors Aren't Buying](https://docs.zigpoll.com/guides/how-to-find-out-why-visitors-arent-buying.md) and the [Exit Intent Survey](https://docs.zigpoll.com/tutorials/exit-intent-survey.md) tutorial.

***

## After delivery: product, packaging and the delivery experience

### Across the customer base

Surveys sent after the order arrives behave like link surveys, not pop-ups: the typical email survey link is opened by 8% of recipients (the top quarter clear 20%), and once opened, around three quarters of people answer. Order and required text questions barely matter here; a required text box is answered 91% of the time when shown. This is the place to ask the questions that need a sentence.

### What brands did with it

* [**AMARO**](https://www.zigpoll.com/blog/amaro-case-study) sends a post-fulfilment survey about the delivery and shares the qualitative feedback with its carriers, which lets them hold partners to account in a country where logistics is the top ecommerce complaint.
* [**Dalfilo**](https://www.zigpoll.com/blog/dalfilo-case-study) learned from post-fulfilment answers that customers did not understand that 100% cotton shrinks on first wash, despite it being on the site. They added a care sticker inside the box.
* The [**beauty brand**](https://www.zigpoll.com/blog/beauty-brand-case-study) above routes happy post-fulfilment respondents to leave a review, which has produced over 1,200 positive reviews, and uses the same survey for packaging and product-performance feedback.
* [**Perfect Presents**](https://www.zigpoll.com/blog/perfect-presents-case-study) installed Zigpoll to troubleshoot the website and ended up using the fulfilment survey to run its delivery experience.

### What to copy

Keep the thank-you page for attribution and intent, and move satisfaction, fit, quality and delivery questions to an [Order Delivered email](https://docs.zigpoll.com/tutorials/order-delivery-survey.md) a few days after arrival. See [How to Measure the Delivery and Product Experience](https://docs.zigpoll.com/guides/how-to-measure-the-delivery-and-product-experience.md).

***

## Turning answers into segments and flows

* [**Jolly Mama**](https://www.zigpoll.com/blog/jolly-mama-case-study) asks each buyer where they are (trying to conceive, pregnant, postpartum) and builds Klaviyo segments from the answer, with rules such as "said pregnant in the last three months, so probably still pregnant." Fewer emails, higher open rates, fewer spam reports.
* [**Yay's Snacks**](https://www.zigpoll.com/blog/yays-snacks-case-study) segments Klaviyo by acquisition channel, having found that Facebook-acquired customers need steeper discounts to return while organic customers are more loyal.
* [**ThumbPRO**](https://www.zigpoll.com/blog/thumbpro-case-study) and [**GrowthTrigger**](https://www.zigpoll.com/blog/growth-trigger-case-study) build Klaviyo segments straight from survey answers.
* [**Jobtest**](https://www.zigpoll.com/blog/jobtest-case-study) (SaaS) triggers conditional workflows from open-ended answers, which lifted conversion on its paywall by about 8%, and uses NPS to model referral rates against the $250,000+ a month it spends on ads.
* An [**8-figure beauty brand**](https://www.zigpoll.com/blog/how-a-beauty-brand-collects-data-before-purchases) runs pre-purchase pop-ups that trade a 10% coupon for two or three questions, including credibility tests such as "Does our claim to easily achieve salon quality from home sound too good to be true?", with a follow-up asking what would remove the fear.

Every integration action is set up in [Question Logic](https://docs.zigpoll.com/questions/question-logic.md). See [Klaviyo Segmentation](https://docs.zigpoll.com/tutorials/klaviyo-segmentation.md) and the [Integrations](https://docs.zigpoll.com/integrations.md) list for Customer.io, Braze and the rest.

***

## How the aggregate numbers were produced

* **Attribution Index.** Order-linked responses only (every response carries a Shopify order), so what the customer said can be compared with the source the store's analytics recorded for that order. Answers are mapped to channels by an open rule set; unrecognised answers are reported as unclassified rather than folded into "Other". Channels with fewer than 50 contributing stores are never published. Full method on the [Attribution Index](https://www.zigpoll.com/attribution-index) page.
* **Engagement and exit intent study.** Trailing 90 days to 20 September 2026, one data point per survey, surveys under 200 presentations excluded, surveys with inconsistent counters excluded, synthetic surveys excluded. Differences were checked within the same store as well as across stores so that merchant mix could not masquerade as a setting.
* **Benchmarks.** See [Response Rate Benchmarks](https://docs.zigpoll.com/response-rate-benchmarks.md).

Stores using Zigpoll are not a random sample of ecommerce, and these figures describe that population rather than the market as a whole.

***

## Related

{% content-ref url="/pages/jep3ZyGAGh7Un6jTVncW" %}
[How to Build a Post-Purchase Survey Customers Actually Answer](https://docs.zigpoll.com/guides/how-to-build-a-post-purchase-survey.md)
{% endcontent-ref %}

{% content-ref url="/pages/Iwn5ywxmSvccwtT1njra" %}
[How to Measure Marketing Attribution](https://docs.zigpoll.com/guides/how-to-measure-marketing-attribution.md)
{% endcontent-ref %}

{% content-ref url="/pages/70rirBfJvG59hsGptkAE" %}
[How to Find Out Why Visitors Aren't Buying](https://docs.zigpoll.com/guides/how-to-find-out-why-visitors-arent-buying.md)
{% endcontent-ref %}

{% content-ref url="/pages/2Xl17hwPIDij54Vg1KPD" %}
[Response Rate Benchmarks](https://docs.zigpoll.com/response-rate-benchmarks.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.zigpoll.com/guides/what-brands-have-learned.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
