Basket study 02
Where you save from says who you are
One save event, five live surfaces, four downstream systems.
The reframe
For about a year the roadmap argument kept returning: which save surface should win – Safari extension, share sheet, in-app browser, paste? Each had an advocate. Wrong question. People save from whichever surface fits the moment, and the same people use several; picking a winner throws away most of the cohort.
The real problem was schema unification, not UI consolidation: making any surface feed the same product. So we built the save layer as a polyglot system – one save event, tagged with its source, five input surfaces feeding it and four downstream systems reading it. I designed the surface UIs and the event schema, and owned the analytics properties so every save could be attributed back to a surface. Instrument every surface to feed that one event and the data settles the roadmap argument on its own – no advocate has to lose face.
Data signal – reach vs depth
First-touch by save method (365 days, 52,684 users). Bars are first-touch users; the dotted line is depth – lifetime saves per user.
Multi-surface users compound
Users who combine surfaces retain materially better than single-surface users.
| Surfaces used | Users | Lifetime saves | 90d active |
|---|---|---|---|
| Share only | 29,640 | 53 | 33.9% |
| Extension only | 6,271 | 883.1 | 21.2% |
| Extension + share | 6,421 | 911.8 | 57% |
| Share + paste | 2,454 | 81.5 | 41.9% |
| Extension + paste | 779 | 829 | 41.3% |
| All three | 1,286 | 1,108.5 | 65.9% |
Three-surface users (n=1,286) retain at 65.9% – over 3× the rate of extension-only users. Multi-surface adoption is the strongest retention signal in the data.
Design implications
Five active input surfaces
- Safari extension (web). Passive auto-capture on every product page view. Deepest input by saves per user. Full case study in case 03.
- OS share sheet (iOS + Android). User shares from any retailer app or Safari → item lands in an inbox → multi-select destination basket via the organise step. Biggest input by first-touch users (31,943).
- Paste-a-link (mobile). Edge-case input – copy URL, paste into save bar.
- FOMO feed / trending (in-app). Pull-based saves from a curated feed.
- AI inspo (in-app generative search). User describes intent, AI returns items, user saves. Small but live.
The share-sheet “organise” step
After the share: the app opens with the item in an inbox, and the user picks one or more destination baskets. 92.7% of users finished picking a basket (May–Oct 2025).
Downstream consumers
Four systems read the same save event:
- Organise. Basket-type stamping; assigns saves to the right destination.
- Price tracking. Every item gets historical price polling.
- Affiliate link generation. ~137K links per month across all surfaces.
- Recommendations. Every save is a co-occurrence signal for the recommendation graph.
Because saves are surface-agnostic downstream, surfaces can fail safely. When the in-app browser experiment ended, nothing downstream broke.
Caveats
- Figures come from a year of product analytics (through Nov 2025). Totals like first-touch and lifetime saves are reliable; a few recent months read directionally because tracking changed mid-window.
- Sum-of-users across methods is not the unique total. Users use multiple methods (see overlap table).
What I’d do next
- Share-sheet → extension graduation prompt. Users who adopt both surfaces (n=6,421) retain at 57% 90d – three times the extension-only rate. A share-sheet first-toucher with N+ saves should see “install the extension for auto-capture” at the right moment.
- Method-cohort retention curves. Per-method retention segmented by month would let surface fatigue show up earlier than first-touch averages can.