Farshad Amiri
← All case studies

Case study 04 · Provin · Alternatives

From an out-of-stock dead end to a way to keep buying

  • Problem

    When a product ran out, shoppers had one option: Notify me — a dead end at the moment they most wanted to buy.

  • Approach

    With no access to users and no A/B testing, I reframed the problem and added Alternatives next to Notify Me, validated with prototypes, guerrilla tests and a measurement plan.

  • Result

    39% of shoppers who opened Alternatives added a substitute to cart — with no drop in Notify Me.

Role
Senior Product Designer · Nizek
Method
Double Diamond + Lean UX
Surface
Product card + PDP, web and mobile
Measure
Add-to-cart from Alternatives
Reading
8 min
Visit Provin ↗
Provin Alternatives on web and mobile: the desktop product page with the “Available similar products!” dialog, an out-of-stock product page with Notify Me next to View Alternatives, and the Alternatives sheet listing similar products.

39%

added a substitute to cart

Of shoppers who opened Alternatives

No drop

in Notify Me clicks

Guardrail held over six months

≈ 2 : 1

Notify Me to Alternatives clicks

Both paths are used

0 A/B tests

needed to ship with confidence

Prototypes, guerrilla tests and a measurement plan

01 · The problem, Context · Discover

A routine purchase with no plan B

Provin is a curated supplements store in Kuwait. People take these products every day and come back every month, so running out breaks a routine, not just a basket.

Before: what happened when a product ran out

  1. Open productCard or product page
  2. Out of stockNo buy buttonPain point
  3. Only Notify meA promise for laterPain point
  4. Back to homeSearch the categoryPain point
  5. Compare againOr leave for a competitorPain point

The shopper with the highest intent left the funnel, at the moment they most wanted to buy.

What a dead end cost

  • Lost today · the transaction

    The shopper with the highest intent left the funnel.

  • Lost later · lifetime value

    A broken routine weakens the monthly return.

  • Lost to others · competitors

    The need gets met somewhere else.

Research without access to users

Two hard limits shaped the method. I treated them as design constraints, not as a reason to skip research.

  • Constraint 1 · No access to end users

    Interviews and field research were not possible.

  • Constraint 2 · No A/B testing

    Not enough traffic or infrastructure in production.

  • Response · Move the research standard

    Desk research, rapid prototypes, guerrilla tests and a pre-launch measurement plan.

What the sources and our own data said

What the sources and our own data said
SourceFindingSo what
Competitor review · xfix.ai, Straight NorthNotify as the main CTA, related alternatives at first glanceA proven pattern to start from
Baymard (via EcomHint)A good alternative meets the same need, not just any in-stock itemDefines the quality bar
StoreBeepBack-in-stock emails convert at about 22%Notify Me is worth protecting
Internal analyticsOut-of-stock status by category, and drop-off on out-of-stock pagesSized the problem

02 · The approach, Define · Develop

One reframe, two jobs

The team wanted to improve Notify Me. The data pointed to a different problem.

Before · solution-led
“Make Notify Me better.”
After · problem-led
“At the moment of no, we have no yes path.”

Two parallel jobs to be done

  • Primary job → Alternatives

    “When my routine product is out of stock, I want the fastest way to cover the same need without breaking my routine.”

  • Second job → Notify Me

    “I want to be the first to know when the original product is back.”

  • Key insight

    The two jobs run in parallel. Neither replaces the other, so the design needs two actions side by side.

  • How might we

    …give shoppers who need it today a way to buy at the moment of out-of-stock, without losing the ones loyal to that exact product?

Four options, one architecture

I weighed each option against both jobs and against risks we could not test.

  • Rejected

    Improve Notify Me only

    The primary job, buying today, stays unanswered.

  • Rejected

    Weaken Notify in favour of alternatives

    Risks losing loyal shoppers and demand data — and we could not test it.

  • Rejected

    Alternatives on the product page only

    Shoppers also meet out-of-stock in lists and search, so the session is lost early.

  • Chosen

    Alternatives next to Notify Me, on card and product page

    Covers both jobs with no extra navigation.

03 · What we did, Develop · Deliver

Two actions, two surfaces

Notify Me keeps its usual place. Alternatives sits next to it as a clear secondary action, never a small lost link.

The substitution ladder — the order alternatives appear in

  • 1

    Same product, another variant

    Other size, flavour or format in stock.

  • 2

    Same active ingredient, another brand

    Similar form and dose.

  • 3

    Same goal, e.g. sleep support

    Always shown with a “why this” flag.

Three rules the logic never breaks

  • No defensible alternative, no button

    A weak suggestion is worse than none.

  • Dose or form differences are always flagged

    The shopper never gets a surprise.

  • Allergens never pass the filter

    Safety comes before conversion.

  • Provin mobile product page for an out-of-stock supplement: an Out of stock tag on the product photo, and Notify Me with View Alternatives side by side at the bottom.

    Product page · mobile

    Clear status first, Notify Me in its usual place, View Alternatives right next to it.

  • The Alternatives sheet on mobile: “Available similar products!” with a grid of similar supplements, each with a price and an add button.

    Alternatives sheet · mobile

    Opens in place: “We’ll notify you. Here are similar products.”

  • Provin desktop product page with the “Available similar products!” dialog open over it, showing a grid of similar supplements.

    Web · product page and dialog

    The same two actions on desktop; Alternatives opens “Available similar products!” over the page.

Prototype and guerrilla test

Figma for the design, Lovable for a clickable build. 5 to 7 colleagues outside the product team, one-on-one, 15 to 20 minutes, think-aloud.

Prototype and guerrilla test
TaskScenarioSuccess looks like
1Dead end on the magnesium product pageSees both actions unaided and can say how they differ
2“I want it today”Alternatives → pick one → add to cart, without going back to home or search
3Alternatives on a card (sleep support)Reaches alternatives without opening the product page, list context kept
4“Only this exact product”Picks Notify Me and does not feel it was demoted
5Edge case: no defensible alternativeDoes not look for a button that is not there
6Synthesis: buying creatineBoth paths appear in their own story

What the test changed

What the test changed
FindingWhat I changedWhy it matters
An icon-only button was unclearAdded a text labelThe action is understood at a glance
Suggestions felt like a hard sellAdded “why this was suggested”Trust in the substitute
Nobody read Notify Me as removedKept its visual weightLoyal shoppers are not demoted

Measurement plan — every event answers one question

Measurement plan — every event answers one question
EventTriggerQuestion it answers
oos_card_view / oos_pdp_viewSeeing an out-of-stock card or product pageHow often do shoppers meet out-of-stock?
notify_click / notify_submitUsing Notify MeDoes the Notify hierarchy still work?
alternatives_click (card / pdp)Clicking Alternatives, and whereWhich surface gets more use?
alt_view / alt_clickSeeing or clicking an alternativeHow appealing are the suggestions?
alt_atc / alt_purchaseAdding or buying an alternativeCore funnel metric
session_recoveredAn out-of-stock shopper reaches the cart in the same sessionNorth Star
  • North Star · Funnel Recovery Rate

    Share of out-of-stock shoppers who get back on the purchase path in the same session.

  • Success

    Out-of-stock to add-to-cart, Alternatives click-through and recovered revenue.

  • Guardrails

    Notify Me must not drop — plus return-to-purchase after notify, and returns or cancellations of substitutes.

04 · The result, Six months after launch

The hypothesis held

Alternatives added a “yes” path without weakening the “wait” path.

  • Add-to-cart from Alternatives

    Each dot is 1 of 100 shoppers who opened Alternatives.

  • Guardrail · Notify Me click rate

    Monthly, six months after launch.

    1. M1: no drop
    2. M2: no drop
    3. M3: no drop
    4. M4: no drop
    5. M5: no drop
    6. M6: no drop
    No drop in any month: Alternatives did not steal from Notify Me.

39%

added a substitute to cart

Of shoppers who opened Alternatives

≈ 2 : 1

Notify Me to Alternatives clicks

Relative share only

No drop

in Notify Me

Six months after launch

What the numbers say

  • The ladder earned trust

    39% means the matching logic worked: the same ingredient and dose, a reason for each suggestion, and no button when no good match existed.

  • Two parallel paths was the right call

    Most shoppers stay loyal to their exact product and prefer to wait. Weakening Notify Me, as first planned, would have lost the largest group.

05 · What I learned, Reflection

Constraints move the standard, they don’t remove it

The question is how to build the most confidence with the tools you have.

  • Limits are not an exemption

    No users and no A/B test changed how I researched, not whether I did.

  • Code-based prototypes speed research

    Lovable let us test the concept in days, not weeks, and made handoff cleaner.

  • Keep usability and value apart

    Guerrilla tests show it is seen and understood. Only launch data shows people buy.

  • Can’t test a risk? Design it out

    We could not prove weakening Notify Me was safe, so we didn’t touch it.

Next Lean UX loop

  1. Measure the Funnel Recovery Rate across all out-of-stock shoppers.
  2. Compare Alternatives use on the card vs the product page.
  3. Track substitutes from add-to-cart to purchase, and their return rate.
  4. A/B test the button position if traffic allows.