Analytics, AI, and Reporting

How to review and analyze your data using Zigpoll's built-in reporting features

Zigpoll comes with a built in suite of analytics that makes it easy to understand responses at high volume. Here's an overview:

Available on every plan, including Lite. Automatic account insights, automatic survey insights, write-in themes & summaries, and Z-GPT chat with your data are all included on every Zigpoll plan. Plans only differ in the monthly AI analysis credit allotment used for on-demand insights — see Subscription Plans.

The AI features below (Z-GPT Insights, Z-GPT Chat, and AI action items) write their replies in the language set under Preferred language in your account's General Settings. Change it there if you'd like your AI summaries and chat replies in another language.

Charts

Zigpoll has built in charts and data visualization.

You can see these in the dashboard and the charts are auto-generated for each question.

Lifetime tiles and the date range

The tiles along the top of your account Dashboard and a survey's Dashboard always show lifetime totals, however you set the date range. They are Answers, Write-Ins, Participants, and, for email or SMS surveys, Emails Sent or SMS Sent. The small change figure beside each one compares it with the previous period.

Pick a date range on the survey card or a question card below them, and those cards switch to the range you chose. That covers Views, Opens, Closes, Responses, Participants, Completions, and the rates. The charts and per-answer totals follow it too. The Charts page has its own date picker.

Response Rate, Answer Rate, and the other rates

Rate	Where	What it measures
Response Rate	Survey card	Responses, one per person who took the survey, divided by the number of times the survey was shown.
Engagement Rate	Survey card	Times people engaged with the survey, divided by the number of times it was shown.
Completion Rate	Survey card	Times the survey was finished, divided by the times people engaged with it.
Answer Rate	Question card	The share of views of this question that produced an answer.
Engagement Rate	Question card	The share of participants who engaged with this question in any way, including skipping past it.
Dropoff Rate	Question card	The share of participants who stopped filling out the survey at this question.

Views on a question card counts how many times that question was viewed. Views on the survey card counts how many times the survey was viewed.

Share of Answers and Penetration Rate

On Multiple Choice, Mosaic, and Image Choice questions, click the percentage beside any answer to switch how percentages are worked out:

• Share of Answers: the answer's count divided by all answers picked on the question. The percentages add up to 100%.

• Penetration Rate: the answer's count divided by the number of people who engaged with the question. It reads as "what share of people picked this". Because one person can pick several answers, the percentages can add up to more than 100%.

The toggle isn't available while a metadata filter is applied.

Cross tabs

A cross tab shows how people who gave one answer answered another question. For example, "what % of people who answered X answered Y".

1. On a survey's Dashboard, go to a fixed-choice question's card.

2. Click Cross Tab, or choose Cross Tabulate Answers from the question's ⋯ menu.

3. In the Cross Tabulate window, pick the two questions to compare and a date range.

4. Click Generate Cross Tab.

The result is a heatmap. Use Flip Axis to swap the questions, or the Download CSV icon to save the table. Cross tabs aren't available for ranking questions.

Revenue and AOV on the Charts page

For questions answered on a Shopify thank-you or order status page, the Charts page adds a metric picker next to the interval picker: Responses, Attributed Revenue or Average Order Value. Every chart on the page follows the metric you choose, including the previous-period comparison, so you can see revenue by answer week over week for an attribution question. Each answer is credited with the full order total behind it, so on a multi-select question the revenue across answers can exceed store revenue; AOV divides by the answers that carry an order value.

Answers by field

Below the time series, every fixed-choice question whose responses carry attribution or traffic metadata shows an Answers by module. Pick a field and the module splits the question's answers by it. The first two fields, Reported Channel and Reported Source, are what the customer said: Zigpoll reads them from each customer's answer to the survey's attribution question, as described under Reported channel and source. The rest are what was tracked. On-site surveys offer the fields Zigpoll reads from the URL the customer arrived on: UTM source, medium, campaign, content and term, referrer, landing page, and Meta or Google ad parameters (campaign, ad set, ad, placement). Post-purchase, abandoned checkout and order-triggered email or SMS surveys also offer what Shopify recorded on the order: the last-click source and source type, the order's landing page and referrer, and the discount code. Only fields present on the question's responses are offered, and a question with none of them shows no module. Each row is one value with its responses, share and, where the responses have an order behind them, attributed revenue and AOV; expand a row to see how those customers answered, and use the heatmap for the full answer-by-value grid. Common spellings of a channel (for example "Facebook", facebook.com and the Facebook app) are folded into one row. The first tile, Has UTM medium or Has last-click source, is the share of responses that carry the field (Used a discount code for the discount pivot), and the empty bucket is named to match: No UTM medium, No tracked source, No discount code. With Compare to previous period on, each row also shows its change in responses and attributed revenue against the previous window. On a "How did you hear about us?" question this is the stated-versus-tracked attribution check; on a "Who is this purchase for?" question it shows which campaigns bring gift buyers. The same report is available to the API as GET /poll/attribution and to AI assistants as the analyze_attribution MCP tool.

Projected totals. Only some of the customers who see a survey answer it. The Surveyed tile gives how many customers the survey reached in the selected range and the share who answered this question, and the Projected totals switch beside the field picker lifts every row from those who answered to everyone surveyed: responses are divided by the response rate, and revenue is each row's AOV times its projected responses, so shares and AOV do not move. It is an estimate that assumes the customers who skipped the survey are like the ones who answered, and the table says so underneath while it is on. The switch is offered only when the range can be projected honestly: views were recorded, the question has at least 30 responses, the response rate is 5% or more, no filter is applied (a filter narrows responses but not views), and the range does not have more responses than views. Add ?projected=1 to the Charts page URL to open it projected. The Reporting API returns the same projected figures, and the reason when a projection isn't available.

You have the ability to change the format of the chart by pressing the icons underneath it, on the right:

The generated charts can be easily exported for reporting or adding to question decks. Hover over the chart and click the menu icon that appears underneath the date range to do so:

Reported channel and source

Reported Channel and Reported Source in the Answers by picker come from the survey itself. Zigpoll reads each customer's answer to the survey's attribution question ("How did you hear about us?"), including "Other" write-ins and any follow-up question such as "Which podcast?", and adds two metadata keys to the response a few minutes after it is submitted:

• attribution_channel is one value from a fixed list, such as instagram, search, podcast or word_of_mouth. "Instagram" picked as a fixed answer and "insta" typed into "Other" both land on instagram.

• attribution_source is the name the customer gave, when they gave one, such as "The Daily", "@somehandle" or "ChatGPT".

There is nothing to set up. No AI model is involved, and your sentiment and AI settings don't affect it.

The other fields in the picker are tracked: they hold what the URL, the UTMs or Shopify recorded about the visit. The reported fields hold the customer's own answer. The two can disagree, and where they do is the useful part. A customer whose last click was a Google search may say a podcast sent them, and word of mouth has no tracked value at all. Pick Reported Channel on any other question, such as "Who is this purchase for?", to split its answers by how customers say they found you. Or filter the page to one reported channel and then pick a tracked field to see how that channel's customers actually arrived.

Reports trail new responses by about a minute, or up to ten minutes for a survey that was left unfinished. Responses collected before this was released don't have the two fields unless Zigpoll support backfills them for your account. The full list of values is in Special Variables, and How do I find out how customers heard about my store? covers how a question is recognized and how to collect the named source.

Filtering results by metadata

You can narrow any question's results down to the responses that carry a particular piece of metadata — for example, only the customers who purchased a specific product, ordered above a certain total, or responded from a particular country or device.

Look for the filter icon next to the date range. It appears in three places:

• Dashboard — on each question's results panel

• Answers — above the list of answers

• Charts — next to the date picker

Clicking it opens a Filter Responses window. Pick a field, then pick a value — the dropdown shows how many responses match each value, and the filter applies as soon as you choose one. The active filter shows on the button itself, with an x to clear it.

The fields available depend on what your responses actually carry:

• Attribution: Reported Channel and Reported Source, which Zigpoll reads from the customer's answer to the survey's attribution question (see Reported channel and source). Reported Channels also appears when a multi-select attribution question had more than one channel picked

• Shopify: Products Purchased, Order Total, and Locale are attached automatically to post-purchase survey responses

• Location & Language, Device, and Page: country, region, device, browser, the page the survey was answered on, and similar fields collected with each response

• Custom: any metadata you attach through the embed code or survey links

Products Purchased supports selecting multiple products. Tick as many as you like — the results include any response whose order contains at least one of the selected products.

While a filter is active, everything recalculates from the matching responses: the per-answer tallies and charts on the Dashboard, the list on the Answers page, and every chart on the Charts page — including the Compare to previous period overlay, which compares filtered against filtered so the deltas stay apples-to-apples. The filter follows you as you switch between questions in the same survey, so you can walk a segment through the whole survey question by question.

The field and value lists are built from your most recent responses, so a value that stopped appearing a long time ago may not be offered — and on very high-volume questions the Dashboard tallies reflect the most recent responses rather than all history. The Charts page always computes filtered results over the full selected date range.

Z-GPT Insights - embedded AI summarization on surveys

Write-ins are some of the most valuable data, while also being the type of responses that are the hardest to analyze.

Z-GPT is a built-in analytics feature that will summarize responses and help extract trends on your behalf, instead of asking an intern or manually reading yourself.

Once you hit a certain volume of responses (~25) we send it to an LLM to extract summaries. This runs once a week.

These summaries are present in the Dashboard or when you click into a specific survey. On a survey's Dashboard, scroll below a question's results to find them.

Here are examples of AI summarization on one of Zigpoll's attribution surveys.

Z-GPT Chat

You can ask questions directly on your data using chat. This can be found in the "Insights" tab in the sidebar.

You have the ability to change which survey the data is being pulled from by selecting the survey at the top.

You can also export a chat conversation to Markdown to share or archive an analysis, and the Insights page shows your remaining monthly insight credit balance (with an upgrade link when you're running low). Credit allotments per plan are listed in Subscription Plans.

Report generation and data export

You can generate a custom report with the columns you care about then export to CSV.

To start, go to "Table" in the sidebar.

Select the survey you want to build a table for:

From here, you can select the fields that you want to export and date range you want to export:

Sending into other services

Many users want to perform deeper dives into their data. Zigpoll has a wide set of integrations supporting this.

The most common of these are exporting to Google Sheets, logging events into product analytics platforms like Segment, and pulling response data from the Reporting API. Please see Integrations for more.