November ecommerce conversion statistics depend on who arrives and when. In UCI’s online shopping archive, a purchase appeared in 24.6% of November returning visits and 9.6% of December returning visits. New visitors followed a different baseline in both months. The source contains 12,330 recorded shopping sessions (Sakar and Kastro, UCI, Aug 2018).
Here we compare visitor mix, observed month, weekday timing, product depth, coded traffic and duration. The results help recovery teams see how the baseline can shift before they try a new campaign. They describe sessions at one retailer rather than carts at checkout. The data and source scopes make it possible to reproduce each comparison without assuming that every unsuccessful session was a recoverable cart.
Key Takeaways
- Returning visits in November recorded a 24.6% purchase rate. (Sakar and Kastro, UCI, Aug 2018).
- The corresponding December return-visitor rate was 9.6%. (Sakar and Kastro, UCI, Aug 2018).
- November’s recorded all-session purchase rate was 25.4%. (Sakar and Kastro, UCI, Aug 2018).
- October’s recorded all-session purchase rate was 20.9%. (Sakar and Kastro, UCI, Aug 2018).
- November weekend visits recorded a purchase rate of 26.5%. (Sakar and Kastro, UCI, Aug 2018).
- December weekend visits recorded a purchase rate of 10.7%. (Sakar and Kastro, UCI, Aug 2018).
- Visits with one to three recorded product pages purchased at 3.8%. (Sakar and Kastro, UCI, Aug 2018).
- Visits with 51 to 100 recorded product pages purchased at 22.0%. (Sakar and Kastro, UCI, Aug 2018).
- November traffic code 2 visits purchased at 30.1%. (Sakar and Kastro, UCI, Aug 2018).
- December traffic code 2 visits purchased at 15.5%. (Sakar and Kastro, UCI, Aug 2018).
- Visits with zero recorded product-page time purchased at 1.7%. (Sakar and Kastro, UCI, Aug 2018).
- Visits with 301 to 600 seconds of recorded product time purchased at 15.3%. (Sakar and Kastro, UCI, Aug 2018).
How do November and December purchase rates differ by visitor type?
A pooled checkout rate can change through shopper mix as well as greater purchase intent. Comparing visitor types inside November and December exposes both differences before a team interprets its recovery results.

What do the individual visitor groups show?
The source records a binary purchase outcome for visits. Returning users in different months may arrive through different traffic and conditions, so preserve the observed visitor labels when comparing each period.
- Recorded sessions with a completed purchase. 15.5% (1,908 of 12,330) in All 12,330 sessions at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.9% (422 of 1,694) in All 12,330 sessions at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.9% (1,470 of 10,551) in All 12,330 sessions at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 25.4% (760 of 2,998) in November observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 30.5% (128 of 419) in November observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.6% (629 of 2,557) in November observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.5% (216 of 1,727) in December observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 22.4% (75 of 335) in December observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.6% (128 of 1,330) in December observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.3% (932 of 7,605) in Eight other recorded months at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 23.3% (219 of 940) in Eight other recorded months at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.7% (713 of 6,664) in Eight other recorded months at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
How do separate shopper surveys contextualize checkout friction?
DHL’s surveys asked distinct groups which barriers caused abandonment or were visible to businesses. Those reports can motivate store-level instrumentation without linking any reported reason to one of the UCI sessions.
- Shoppers reporting complicated checkout as an abandonment reason. 54.0% in DHL 2026 shopper or business survey, Dec 2025 to Feb 2026; groups polled separately, online shopper respondents (DHL, Jun 2026).
- Businesses reporting seeing complicated checkout as an abandonment reason. 31.0% in DHL 2026 shopper or business survey, Dec 2025 to Feb 2026; groups polled separately, business respondents (DHL, Jun 2026).
- Shoppers reporting discount codes that fail as an abandonment reason. 57.0% in DHL 2026 shopper or business survey, Dec 2025 to Feb 2026; groups polled separately, online shopper respondents (DHL, Jun 2026).
- Businesses reporting seeing discount codes that fail as an abandonment reason. 39.0% in DHL 2026 shopper or business survey, Dec 2025 to Feb 2026; groups polled separately, business respondents (DHL, Jun 2026).
Run recovery and control groups in the same calendar block and retain persistent shopper assignment across visits. A larger purchase share during November can change both groups. Keep the natural baseline in the same cohort separate from any effect attributed to the new message.
How unusual was November across the observed months?
Monthly comparisons are useful when the observation window is clear. The source records ten months and does not identify Black Friday or any of the other specific sales-event dates.

What does the full observed calendar look like?
The change in pooled baseline reflects the full mix of visitors and conditions in each observed month. Keep missing months visible instead of interpolating an unobserved rate.
- Recorded sessions with a completed purchase. 1.6% (3 of 184) in Feb observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.1% (192 of 1,907) in Mar observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.9% (365 of 3,364) in May observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.1% (29 of 288) in June observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 15.3% (66 of 432) in Jul observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 17.6% (76 of 433) in Aug observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 19.2% (86 of 448) in Sep observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.9% (115 of 549) in Oct observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 25.4% (760 of 2,998) in Nov observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.5% (216 of 1,727) in Dec observations at the archived site, all visitor types (Sakar and Kastro, UCI, Aug 2018).
How do rates differ within new and returning visits?
This segmentation can change the comparison when new traffic made up more visits in one month than the other. Small monthly groups need count context because one extra purchase can move the reported fraction substantially.
- Recorded sessions with a completed purchase. 0.0% (0 of 1) in Feb observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 1.6% (3 of 183) in Feb observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 15.5% (36 of 232) in Mar observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.3% (156 of 1,675) in Mar observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 27.6% (88 of 319) in May observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.1% (277 of 3,045) in May observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 23.3% (7 of 30) in June observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.6% (22 of 257) in June observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.4% (11 of 54) in Jul observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 14.6% (55 of 378) in Jul observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 29.2% (21 of 72) in Aug observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 15.2% (55 of 361) in Aug observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 25.9% (28 of 108) in Sep observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 17.1% (58 of 340) in Sep observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 22.6% (28 of 124) in Oct observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.5% (87 of 425) in Oct observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 30.5% (128 of 419) in Nov observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.6% (629 of 2,557) in Nov observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 22.4% (75 of 335) in Dec observations at the archived site, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.6% (128 of 1,330) in Dec observations at the archived site, returning visitors (Sakar and Kastro, UCI, Aug 2018).
A store should match periods by distance to a promotion or holiday and mark the launch and end dates. Compare within contemporaneous customer groups before treating those changes as a change in marketing effectiveness.
Is the weekend baseline stable across shopping periods?
The weekend flag gives a second comparison inside each month. Rates can look different when the test and comparison have different mixtures of weekday and weekend activity.

What is visible in the two holiday months?
Visits vary with traffic and customer schedules. The weekend flag is a useful date grouping without identifying the causal impact of weekend browsing.
- Recorded sessions with a completed purchase. 26.5% (210 of 791) in Nov, weekend sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 29.1% (37 of 127) in Nov, weekend sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 26.1% (173 of 662) in Nov, weekend sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.9% (550 of 2,207) in Nov, weekday sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 31.2% (91 of 292) in Nov, weekday sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.1% (456 of 1,895) in Nov, weekday sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.7% (39 of 366) in Dec, weekend sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 18.3% (13 of 71) in Dec, weekend sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.6% (25 of 290) in Dec, weekend sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.0% (177 of 1,361) in Dec, weekday sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 23.5% (62 of 264) in Dec, weekday sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.9% (103 of 1,040) in Dec, weekday sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
Does a pooled weekday and weekend view tell the same story?
Pooling all months can give more weight to the most active periods. Use rates inside month and group if the decision concerns sending on one specific day or in a particular holiday week.
- Recorded sessions with a completed purchase. 14.6% (250 of 1,711) in Other observed months, weekend sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 19.6% (55 of 281) in Other observed months, weekend sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.6% (195 of 1,430) in Other observed months, weekend sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 11.6% (682 of 5,894) in Other observed months, weekday sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.9% (164 of 659) in Other observed months, weekday sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 9.9% (518 of 5,234) in Other observed months, weekday sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 17.4% (499 of 2,868) in all, weekend sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 21.9% (105 of 479) in all, weekend sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 16.5% (393 of 2,382) in all, weekend sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 14.9% (1,409 of 9,462) in all, weekday sessions, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 26.1% (317 of 1,215) in all, weekday sessions, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.2% (1,077 of 8,169) in all, weekday sessions, returning visitors (Sakar and Kastro, UCI, Aug 2018).
For timing tests, give both groups the same period of calendar exposure. One strong weekend is not evidence that changing the message timing helped. Consider when recipients actually qualified for contact and when buyers placed a completed order.
Related research on Natural purchases and email recency.
How do complete-session product page counts relate to purchases?
Time and page counts can be strong descriptive measurements without being available at decision time. Variation by browsing depth gives a useful way to spot when a trigger depends on events that may arrive too late.

What occurs across full-session page-count bands?
A session can contain follow-up activity after the shopper reaches an initial cart decision. The exact temporal relationship between browsing count and checkout eligibility must be established in a merchant record.
- Recorded sessions with a completed purchase. 15.8% (6 of 38) in Sessions with 0–0 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 3.8% (58 of 1,545) in Sessions with 1–3 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.6% (223 of 2,590) in Sessions with 4–10 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 16.2% (568 of 3,498) in Sessions with 11–25 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.3% (509 of 2,507) in Sessions with 26–50 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 22.0% (315 of 1,429) in Sessions with 51–100 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 31.7% (229 of 723) in Sessions with 101–10000 recorded product-page visits, all visitor types (Sakar and Kastro, UCI, Aug 2018).
Does the relationship vary by visitor type?
Separate visitor populations can have distinct reasons to browse more deeply, and small cells offer fewer observations. This is not evidence that making someone visit more products is a conversion tactic.
- Recorded sessions with a completed purchase. 28.6% (2 of 7) in Sessions with 0–0 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.9% (4 of 31) in Sessions with 0–0 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.4% (25 of 186) in Sessions with 1–3 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 2.5% (33 of 1,322) in Sessions with 1–3 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 17.1% (86 of 504) in Sessions with 4–10 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 6.4% (133 of 2,065) in Sessions with 4–10 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 28.8% (183 of 635) in Sessions with 11–25 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.4% (381 of 2,847) in Sessions with 11–25 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 34.9% (97 of 278) in Sessions with 26–50 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 18.2% (405 of 2,220) in Sessions with 26–50 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 32.9% (25 of 76) in Sessions with 51–100 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 21.4% (289 of 1,352) in Sessions with 51–100 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 50.0% (4 of 8) in Sessions with 101–10000 recorded product-page visits, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 31.5% (225 of 714) in Sessions with 101–10000 recorded product-page visits, returning visitors (Sakar and Kastro, UCI, Aug 2018).
Freeze state at the moment a real recovery rule first qualifies. Record views, adds, started checkout and placed order independently. When the outcome is a purchase, never use a predictor that only becomes available after purchase or after the decision to contact.
How much can unobserved traffic mix affect the baseline?
Repeated numeric traffic codes give a way to examine the mix without guessing which channel the publisher meant. Each code can appear in several months with different observed purchase rates.

What do the November traffic splits show?
The code buckets may contain different shoppers and shopping intent. Exact visit counts keep large groups distinguishable from sparse categories.
- Recorded sessions with a completed purchase. 30.1% (374 of 1,242) in Source-defined traffic code 2, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.0% (93 of 466) in Source-defined traffic code 1, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 18.6% (76 of 408) in Source-defined traffic code 3, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.5% (4 of 32) in Source-defined traffic code 4, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.9% (21 of 193) in Source-defined traffic code 13, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 29.7% (62 of 209) in Source-defined traffic code 10, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 16.2% (11 of 68) in Source-defined traffic code 6, Nov, all visitor types (Sakar and Kastro, UCI, Aug 2018).
How do the same codes compare in December or other months?
Within-code variation can remain after separating these categories. The split cannot isolate a change in any marketing channel or discount offer, since those conditions are unrecorded.
- Recorded sessions with a completed purchase. 15.5% (121 of 783) in Source-defined traffic code 2, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 18.6% (352 of 1,888) in Source-defined traffic code 2, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.6% (33 of 383) in Source-defined traffic code 1, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.5% (136 of 1,602) in Source-defined traffic code 1, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 4.3% (7 of 161) in Source-defined traffic code 3, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 6.5% (97 of 1,483) in Source-defined traffic code 3, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 15.5% (161 of 1,037) in Source-defined traffic code 4, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 1.9% (1 of 53) in Source-defined traffic code 13, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 4.3% (21 of 492) in Source-defined traffic code 13, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.8% (14 of 109) in Source-defined traffic code 10, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.6% (14 of 132) in Source-defined traffic code 10, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.6% (6 of 44) in Source-defined traffic code 6, Dec, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 10.8% (36 of 332) in Source-defined traffic code 6, other months, all visitor types (Sakar and Kastro, UCI, Aug 2018).
In a real holdout, preserve the source or campaign fields that existed before assignment. Compare eligible shoppers from the same sources and weeks. Use the actual merchant attribution definition rather than mapping opaque codes to social, search, email or direct.
Can observed shopping duration become a recovery trigger?
Completed-session time can summarize engagement even when some of it happens after the message would be sent. The useful question is what was observable when the condition first became true.

What does elapsed time describe in the finished visit?
The archive permits a descriptive time-band comparison but does not expose the features present before a particular abandoned-cart decision. A long complete session may tell us that activity continued without telling us what caused it.
- Recorded sessions with a completed purchase. 1.7% (13 of 756) in 0–0 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 6.2% (25 of 406) in 0.01–30 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 4.7% (57 of 1,202) in 31–120 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 7.4% (133 of 1,807) in 121–300 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 15.3% (307 of 2,005) in 301–600 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 20.2% (753 of 3,722) in 601–1800 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 25.5% (620 of 2,432) in 1801–100000 seconds of observed product-page time, all visitor types (Sakar and Kastro, UCI, Aug 2018).
What differs among visitor types at the same duration?
These results let a reader inspect the extent to which the pattern accompanies a visitor category. A current store needs consent, event ordering and stable eligibility to measure a recovery action.
- Recorded sessions with a completed purchase. 11.9% (5 of 42) in 0–0 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 1.1% (8 of 697) in 0–0 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 22.4% (17 of 76) in 0.01–30 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 2.5% (8 of 325) in 0.01–30 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 8.2% (16 of 196) in 31–120 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 3.9% (39 of 988) in 31–120 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 13.0% (48 of 368) in 121–300 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 5.9% (84 of 1,429) in 121–300 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 26.4% (105 of 397) in 301–600 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 12.4% (198 of 1,591) in 301–600 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 37.6% (190 of 505) in 601–1800 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 17.4% (557 of 3,204) in 601–1800 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 37.3% (41 of 110) in 1801–100000 seconds of observed product-page time, new visitors (Sakar and Kastro, UCI, Aug 2018).
- Recorded sessions with a completed purchase. 24.9% (576 of 2,317) in 1801–100000 seconds of observed product-page time, returning visitors (Sakar and Kastro, UCI, Aug 2018).
Collect only pretrigger features when evaluating an intervention. Define a fixed follow-up window and count every assigned eligible checkout, including suppressed or uncontactable customers if that is how the policy works. Compare completed orders and net contribution alongside message metrics.
Related research on Email conversion denominators.
Methodology and Sources
We tabulate all 12,330 rows of UCI’s Online Shoppers Purchasing Intention archive. Revenue indicates a recorded purchase. Comparisons group sessions by the publisher’s observed Month and VisitorType labels, weekend flag, traffic code and product engagement. Pooled rates retain the Other group. Traffic codes do not have a verified channel mapping in the public data. Duration is rounded to whole seconds. Counts and duration cover the observed session and can include events after a merchant would choose contact. The archive supplies no purchase value, message exposure or eligibility information. Figures are a CC BY 4.0 adaptation with exact denominators.
Glossary
- Session purchase rate: Share of observed visits with Revenue marked true.
- Visitor type: The category supplied with an observed visit.
- Traffic code: A source category without a verified marketing-channel mapping.
- Baseline: Comparable outcome rate used to evaluate a change.
- Holdout: An otherwise eligible group that does not receive the action.
- Post-event measure: A feature measured after the decision it might try to predict.
- Eligibility window: Time between qualifying for measurement and final outcome.
- Selection bias: A group difference that exists before the measured intervention.