Using Post-Purchase Surveys to Measure Ad Channel Attribution

Every ad platform claims credit for your sales. Meta says it drove 400 purchases. Google says 300. ChatGPT says 80. Add them up and you get 780, but your store only shipped 500 orders. This is the attribution problem: platforms count overlapping touchpoints, and the total is impossible.

Post purchase survey attribution offers a different approach. Instead of trusting platform pixels, you ask customers directly: "How did you hear about us?" Their answer is imperfect, but it is a real signal from a real buyer about what actually influenced their decision.

This guide covers how to design effective post-purchase surveys, what answer options to include for channels like ChatGPT ads and Meta, how to interpret results and how to reconcile survey data with platform reporting.

Why post-purchase surveys matter

Platform attribution has structural problems:

1. Credit for existing intent. If someone was already planning to buy and happens to see your ad first, the platform takes credit for a sale it did not cause.
2. Overlap with other channels. A customer who sees ads on three platforms gives credit to all three. The same sale is counted three times.
3. Black boxes. You cannot see how the platform decided to attribute a sale. You have to trust their model.

Post-purchase surveys bypass this. You ask the customer what they remember influencing their purchase. Their answer is not perfect, since memory is imperfect and multi-touch journeys are hard to recall, but it is independent of platform self-reporting.

For newer channels like ChatGPT ads, surveys are especially valuable. ChatGPT's reporting is less mature than Meta's, and conversion data may be sparse. Survey responses give you a second signal to validate whether the channel is actually reaching buyers.

Where to place the survey

The most common placement is on the order confirmation page, right after checkout. The customer has just completed their purchase, so their memory of the journey is fresh.

Alternative placements:

• Post-purchase email. Send a survey link 30 minutes to 24 hours after the order. Response rates are lower but still useful.
• Order status page. Some brands add a survey to the page customers visit to track shipping.
• In the packaging. A QR code leading to a survey can work, but timing is much later and response rates are lower.

For most brands, the order confirmation page is the best balance of response rate and data quality.

Designing the survey question

Keep it simple. One question is ideal; two is the maximum. Any more and completion rates drop.

The classic question:

How did you hear about us?

Variations that can work:

• "What brought you to us today?"
• "Where did you first learn about [brand name]?"
• "How did you discover us?"

The wording matters less than consistency. Pick a question and stick with it so responses are comparable over time.

Single-select vs multi-select

Single-select forces customers to pick one answer. This is easier to analyze but oversimplifies multi-touch journeys.

Multi-select allows customers to check all channels that played a role. This captures more information but makes analysis harder. If 40% of customers select both "Instagram ad" and "Friend or family," how do you allocate credit?

Most brands use single-select with an instruction like "Select the one that mattered most." It is not perfect, but it produces cleaner data.

Answer options to include

Your answer options should cover all meaningful channels a customer might have encountered. Missing an option means losing that signal.

Here is a starting list for a typical ecommerce brand:

ChatGPT as an answer option

For brands running ChatGPT ads, including ChatGPT as an explicit option is important. Otherwise, customers who found you through ChatGPT might select "Online search" or "Other," and you lose the signal.

You can make the option more specific:

• "ChatGPT or AI assistant" (captures Claude, Perplexity and similar)
• "ChatGPT ad" (if you want to isolate ads specifically)
• "AI chatbot recommendation"

The trade-off is specificity versus respondent understanding. Most customers know "ChatGPT" but may not distinguish between an ad and an organic mention. A broader option like "ChatGPT or AI assistant" may capture more responses at the cost of some precision.

Keep the list manageable

Ten to fifteen options is a good range. More than that overwhelms respondents and increases "Other" selections. Fewer than that may miss important channels.

Review your list quarterly. If a new channel becomes meaningful, add it. If an option consistently gets zero responses, remove it.

Interpreting survey results

Survey data should be viewed as directional, not precise. A few important caveats:

Memory is imperfect

Customers do not always remember every touchpoint. They may recall the last ad they saw but not the first. They may remember a friend's recommendation but forget the Instagram post that prompted the conversation.

Self-reported data has biases

Some channels are more memorable than others. A striking TikTok video may be remembered more than a static feed ad. A brand mentioned by a favorite podcaster may be attributed to "podcast" even if the customer also saw multiple ads.

Multi-touch journeys collapse to one answer

In single-select surveys, a customer who saw a Meta ad, then a ChatGPT ad, then bought from an email picks one. That does not mean the other touchpoints did not matter; it means they picked what felt most influential to them.

Use trends, not absolutes

Do not obsess over whether ChatGPT is at 4% or 5% of responses. Watch the trend: is it growing month over month? Is it stable? Is it declining after you paused ads? Trends tell you more than point estimates.

Reconciling survey data with platform data

Survey data and platform data will never match. They measure different things. The goal is to triangulate, not to pick one as truth.

Build a comparison dashboard

Track three views of each channel:

1. Platform-reported conversions. What Meta, ChatGPT, Google, etc., claim in their dashboards.
2. Survey-attributed conversions. Total orders multiplied by the percentage selecting that channel in surveys.
3. Blended metrics. Total revenue divided by total ad spend (MER), which does not attribute to channels at all.

A simple monthly table might look like this (hypothetical numbers for illustration):

These are hypothetical figures to show the comparison structure. In practice, your numbers will differ. Platform ROAS is typically higher because platforms overclaim. Survey ROAS is typically lower and may be closer to incremental truth, though still imperfect.

Use survey data to calibrate platform data

As a hypothetical example: if Meta claims 400 purchases but surveys attribute 250, you might apply a calibration factor (in this case 62.5%) to future Meta reporting. This is rough math and will vary for every brand, but it helps you avoid over-investing based on inflated platform claims.

Compare to MER

MER (marketing efficiency ratio) is total revenue divided by total marketing spend across all channels. It does not attribute anything; it just measures overall efficiency.

If you scale ChatGPT spend and MER improves, that is a positive signal regardless of what surveys or platform dashboards say. If you cut Meta spend and MER stays flat, maybe Meta was claiming sales it did not cause.

For a deeper look at blended metrics, see our guide to MER (marketing efficiency ratio) and the MER calculator.

Common post-purchase survey mistakes

Mistake 1: Not including ChatGPT as an option

If you run ChatGPT ads but do not list ChatGPT in your survey, customers who found you there will select "Other" or "Online search." You lose the signal. Always include every active paid channel as an explicit option.

Mistake 2: Too many options

A list of 25 channels overwhelms respondents. They skim, pick something familiar or select "Other." Consolidate options where possible. "Facebook ad" and "Instagram ad" can be combined into "Facebook or Instagram ad" if you do not need to separate them.

Mistake 3: Changing options frequently

If you add, remove or rename options every month, you cannot compare trends over time. Make changes sparingly and document when you do.

Mistake 4: Ignoring low response rates

If only 10% of customers complete the survey, your data may not be representative. The 10% who answer may differ from the 90% who do not. Work to improve response rates: keep the survey short, place it prominently, make the submit button obvious.

Mistake 5: Treating survey data as ground truth

Survey data is a signal, not an answer key. It has biases and limitations. Use it alongside platform data and blended metrics, not as a replacement.

Incrementality and surveys together

Post-purchase surveys measure perception: what customers remember influencing them. Incrementality tests measure causation: what would have happened without the ads.

The two complement each other. Surveys can tell you that 8% of buyers recall ChatGPT. An incrementality test (like a geo holdout) can tell you whether pausing ChatGPT ads actually reduces sales by 8%, or more, or less.

If surveys say ChatGPT matters but an incrementality test shows no lift when you pause it, maybe customers are remembering ChatGPT but would have bought anyway. If surveys show low attribution but incrementality tests show real lift, maybe ChatGPT influences purchases that customers attribute to other channels.

Neither data source is complete. Together, they give you a better picture. For more on incrementality testing, see our guide to incrementality testing for paid ads.

Tools for post-purchase surveys

Several tools make it easy to add surveys to your order confirmation page:

• Shopify apps. Fairing (formerly Enquire Labs), KnoCommerce and others integrate directly with Shopify checkout.
• Google Forms. Free, embedded via iframe. Works but looks less polished.
• Typeform. More design flexibility, can be embedded or linked.
• Custom build. If you have development resources, a simple form posting to a database works fine.

For most Shopify brands, a dedicated app like Fairing or KnoCommerce is the fastest path. They handle data collection, reporting and integration with ad platforms for automated syncing.

How SecondWin helps

SecondWin builds and manages ChatGPT ad campaigns for ecommerce brands. Because ChatGPT is a newer channel with less mature attribution than Meta or Google, we encourage tracking performance through blended metrics like MER and post-purchase surveys alongside platform reporting.

SecondWin does not claim credit for sales; ad spend is billed by OpenAI directly to you, and you see results in your own store data and surveys. That independence means you can validate ChatGPT's contribution without relying on platform self-reporting.

Want to see what ChatGPT ads would look like for your brand? Run a free analysis of your website to see the buyer questions and ad concepts we would start with. Plans start at $99 per month on the pricing page. SecondWin is independent and not affiliated with OpenAI.

FAQ

How accurate are post-purchase surveys for attribution?

Post-purchase surveys are directionally useful but not perfectly accurate. Customers may not remember every touchpoint, and memorable channels may be overrepresented. Use survey data as one signal alongside platform data and blended metrics, not as the definitive answer.

What percentage of customers typically complete post-purchase surveys?

Completion rates vary widely based on placement and design. Surveys on the order confirmation page generally see higher completion than follow-up emails, since the customer is still engaged and the question is right in front of them. Keep surveys short (one to two questions) to maximize response rates.

Should I include ChatGPT in my survey options?

Yes, if you run ChatGPT ads or expect organic mentions in ChatGPT to drive traffic. Without an explicit option, customers who found you through ChatGPT may select "Other" or "Online search," and you lose the signal. An option like "ChatGPT or AI assistant" captures this emerging channel.

How do I reconcile survey data with platform-reported conversions?

Build a comparison view showing platform-reported purchases, survey-attributed purchases (total orders times the percentage selecting that channel) and blended MER. The numbers will not match, but comparing them helps you understand how much each platform may be overclaiming. Use survey data to calibrate your interpretation of platform reports.

Can post-purchase surveys replace incrementality testing?

No. Surveys measure what customers remember, which is perception. Incrementality tests measure what would have happened without ads, which is causation. A customer might recall seeing a ChatGPT ad but would have bought anyway. The two methods complement each other; using both gives you a fuller picture.