Sell the sample, then the order

This summer I designed a paid funnel for an architectural products retailer. Their finishes cost enough that buyers want to see one in their own light before ordering a room's worth. So the ads sold a sample first, and the full order was the second sale. The agency team built and ran the channels. The design and the reporting were mine.

Samples were about 70% of the store's orders (count) and under 4% of its revenue. The sale the ads could see was the small one. The order that paid for it all came weeks later.

Build each ad on a worry

The Meta ads were built on four things that stop someone ordering: will it look real, will it work in my room, what if I pick the wrong finish, and can I install it myself. Each ad opens with the worry and answers it with customers' own photos where we had them, or the sample itself. The offer stays out of the opening line.

I asked for one campaign with each worry as a named ad, optimizing to a sample purchase in any line and weighted by value, because the lead line's sample buyers were worth two and a half to five times as much in later full orders as the other lines' (June snapshot). In the first week the fit worry had the best click-through rate. Separate ads made that visible.

Give each worry a page

The click should land on a page that answers the same worry with more proof and makes the sample the easy next step. Ours started with one sample page for every ad, and a page per worry is what I'd build next.

Check the page before you spend on it. Our test pages said 7 in 10 customers go on to a full order, while our data said roughly 10 to 23% of sample buyers (count), depending on how long you wait. Session recordings later found script errors in 81.8% of mobile sessions on the paid page.

A promo code toward the full order

Every sample order earns a promo code by email, which takes the price of up to five samples off a full-size order at checkout. It expires after six months and can't be combined with other offers. The client asked us to call it a promo code, never a credit, rebate or refund, because the buyer has to type it in.

In QA, our pages' FAQs had this right while their product badges still said "refunded". The welcome emails were designed around a second code, 30% off the price of a first sample order. Keep the two codes clearly apart, and decide in advance what the promo code takes off when the samples were bought at the discount.

The emails between sample and order

For people who sign up without buying, I designed four welcome emails that branch on one signup question, what they're shopping for. The promo code toward the full order leads. The sample discount is a side note until the last email, when it's about to run out.

After the sample come post-purchase emails by product line, and two segments of sample buyers with no full order, day 25 to 90 and a long tail to day 270. These were built to feed the closing audiences in search ads, and anyone who has placed a full order should come out of every acquisition and closing audience. Give sample carts their own cart emails too. Ours sent them full-order copy.

Follow buyers forward

I grouped buyers by the month of their first sample and followed each group to its first full order. The median wait was about 25 days, between 24 and 30 across the February to May cohorts.

Counting backwards is an easy mistake. Measured from a month's full orders back to each buyer's first-ever sample, the median nearly doubles to about 48 days. Looking back from the most recent sample gave 23 days, close to the forward number.

Short windows mislead the other way. For the February to April cohorts, looking only 30 days out gave a median of 13 days, 90 days gave 19, and 180 days gave the full 25. A young cohort always looks fast, because its slow buyers haven't arrived.

Judge every cohort at the same age

At about ten days old, 1.3% of the paid cohort's sample buyers (count) had placed a full order, against 0.9% for June's cohort and 2.5% for July's at the same age. It sat inside the normal range, with most of its orders still to come.

Our plans used later full-order revenue per sample buyer by line, with those who never order counted in. It doesn't prove the sample caused anything, and a paid cohort may convert differently. We didn't get to find out, because the new owners closed both sites in September.

Reconcile before you report

The store had no "sample order" type, so I defined one: an order where every item is a sample. Counted up from individual orders, it matched the email platform's monthly totals within 0.1% on settled months. Label that definition every time. The board's count used a category tag that includes mixed carts and ran about 1.5% higher (count). A newer "Samples" tag only reached the whole catalog mid-year, so a year-over-year view on it showed 241 sample orders against 2.

Put the top line on store revenue. Meta credits the ad that started a purchase and web analytics credits the last click, so they won't agree. Judge Meta on what its cohorts go on to buy. August revenue did grow, mostly from returning trade orders, and I didn't credit it to the funnel.

Do this Monday: one table, one worry

1. Pull a year of first-time entry-product buyers by month, follow each month forward, and note the median days to a full order and the share (count) ordering by day 10, 30 and 60.
2. List the worries that stop someone ordering your full product. Reviews and support emails will have them.
3. Pick one. Write its ad and landing page, check every claim against your data, and tap everything on your phone.
4. Launch one ad first and check it after review. Our first 45 came back with their media stripped after an integration rewrote the tracking.
5. Keep the campaigns that bring in full orders today at full budget until the first sample cohort passes its median.

---

In one line: let the ads sell the sample by answering a worry, then judge each cohort forward and at the same age.