Back to Explore
GuideVibeFix editorialUpdated Nov 16, 2026

One Search Is Not Enough: Saved Searches and New-Item Alerts in Practice for Vibe Projects

Of 1,000 searches, 187 are saved, 132 turn on new-item alerts, 96 open an alert, and 61 finish what they came to do. A saved search stores conditions, not items: keywords plus filters plus a price range, so results keep growing on their own. This guide covers the funnel, one-tap saving, trigger rules, alert copy, the management page, and three tiers. The product, people, and numbers in this walk-through are fictional.

A real photo: a smiling woman looking at the phone in her hands while searching on it.

Not finding it in one search does not mean the user stopped wanting it

First, a clear statement: the product, people and numbers in this walk-through are fictional examples, used to make the reasoning concrete. They are not operating data from any real product. The example product is "Secondhand Market", a fictional secondhand trading app, and the example user is Xiaobei, who wants a used 27-inch 4K monitor for under 1,500 yuan, will accept light wear, but does not want dead pixels. The first time he opens the app and searches for "27-inch 4K monitor under 1500", the results are either over budget, have dead pixels, or are too far away. He closes the app and leaves — but that does not mean he stopped wanting one. On the contrary, he is exactly the user worth keeping: his need is specific, his budget is specific, and he is only waiting for the right item to appear.

"Saved search" is built for people like Xiaobei: save this search with its conditions, give it a name, and let the app tell you when something new matches. Without it, Xiaobei has to come back and search manually every day, and after three days of nothing he gives up. With it, he searches once, saves it, turns on alerts, and the app does the rest. Many vibe projects build search as "one search only" — an input box, a results list, some filters — and miss the part that comes after the search. A user who leaves empty-handed has not lost the need; your product has simply failed to catch it.

A saved search is not the same as three features it is often confused with. A favourites list stores "things already seen", such as a specific monitor Xiaobei saved. Search history stores "words searched", which can only be searched again. Subscriptions and follows store "people or shops". A saved search stores "conditions": keywords plus filters plus sorting plus a price range. When the conditions change, the results change with them — that is the fundamental difference from a favourites list. What the user saves is not an item but a question that keeps growing on its own.

The funnel first: where 1,000 searches go

Let us look at the problem with fictional numbers. Suppose "Secondhand Market" sees 1,000 searches in a month, and they break down like this: after searching, 187 searches are saved; of those 187 saved searches, 132 turn on new-item alerts; of those 132, 96 are opened after an alert arrives; and of those 96 opens, 61 finish what the user came to do — contacting the seller or placing an order. The step conversion rates are 19%, 71%, 73% and 64%. The losses split like this: 813 searches leave without saving anything; 55 are saved but never turn on alerts; 36 have alerts on but are never opened; 35 are opened but lead to no action. 55 plus 36 plus 35 equals 126, which is exactly 187 minus 61. The arithmetic balances.

Funnel chart: of 1,000 searches, 187 are saved, 132 turn on new-item alerts, 96 open an alert, and 61 complete the goal. The biggest leak is the first step, where 813 searches leave without saving. Numbers are fictional examples.

The first step is the one to stare at: 813 searches leave without saving. That does not mean 813 people stopped wanting anything; it means your product gave them too little reason, or the wrong moment, to save. The further down the funnel you go, the healthier it looks: 71% of people who save are willing to turn on alerts, and 73% of those who turn alerts on are willing to open them. Once a user gets past "save", the rest of the path works. Many teams see low alert open rates and reach for better push copy first. But the numbers say the real fix is step one: get more people to save a search. Fix the wrong step and no amount of copywriting will save you.

The last step is the second one worth watching: 35 of 96 opens lead nowhere. Most likely the alert was "almost right" — the price was 50 yuan over budget, or one line in the condition description made the buyer hesitate. That is not the alert's fault; it is the matching being too coarse. The section on trigger rules below exists to close exactly that gap.

Making saving easy: timing, names, and one tap

Where the save button sits decides whether anyone taps it. The most common mistake is burying it at the bottom of the filter panel or inside a "more" menu. After a search, the user's eyes are on the results list, not in a menu. The right approach is to put a "Save this search" button right at the top of the results, next to the search terms — and to put one on the zero-result page too, with the copy changed to "Tell me when one arrives". Zero results is precisely when the desire to save is strongest: the user has just been told "nothing right now", and the next sentence they most want to hear is "I will tell you when there is".

After the tap, do not throw up a form asking for a name, a category, an expiry date and an alert switch. Saving must be light: one tap saves it, the button changes to "Saved", and a default name is filled in from the search terms plus the key filters — "27-inch 4K monitor, under 1500". The name can be changed later, but it works unchanged. Leave naming for the management page. Many products make the mistake of turning saving into a ceremony — title, description, category, expiry, alert switch, five fields — and users give up at field two. Remember: the first goal of saving is "it got saved", not "it got saved neatly".

One more detail: do not let a user save the same conditions twice. If the search closely matches an already-saved one, show "Saved" on the button and let the tap open that saved record. Once duplicate records pile up, the management page becomes a junk drawer that nobody opens again. Deduplication can be one gentle prompt at save time: "You already saved a similar search (27-inch 4K monitor). Want to see it?" One sentence saves a lot of cleanup later.

When alerts fire, what they carry, and how often

Turning alerts on does not mean pushing once for every new item. The trigger rules decide whether this feature feels "thoughtful" or "harassing". Start with the trigger condition: only newly listed items that match the saved conditions fire an alert. Price changes, relistings and edited descriptions do not trigger by default, unless the user explicitly turned on an advanced switch such as "price drop alerts". After triggering, three more gates apply. First, deduplication: the same item is pushed once; a seller taking an item down and relisting it does not count as new. Second, merging: if more than three matching items appear in a short window, combine them into one message — "3 new items match your search" — instead of three pushes in a row. Third, a frequency cap: at most two pushes a day, with anything beyond that held for the next day's digest. And there must be quiet hours: a new-item alert at 2 a.m. wakes the user up and achieves nothing else.

Beyond the trigger rules sits a more fundamental rule: alerts may only push what the user asked for, never what you want to sell. That is the dividing line between saved-search alerts and marketing pushes. The user saved "27-inch 4K monitor under 1500"; pushing a 32-inch monitor at 2,000 yuan, even on sale, crosses the line. Cross it once and trust in alerts drops a notch; cross it three times and the user turns alerts off, and the app's notification permission with them. Trust is this feature's only asset. Spend it carefully.

Freshness needs honest thinking too. Secondhand goods are first-come, first-served; an alert that arrives half a day late is an alert for an item that is already sold. If you cannot deliver in real time, do not promise real time. It is more honest to write "listed 2 hours ago" in the alert and let the user judge whether it is still worth opening. Being clear about freshness beats pretending to be instant.

Writing the alert: three seconds to decide

Before an alert is opened, the user has about three seconds: one glance at the title, then tap or swipe away. The title must carry the key matched facts. A good title reads: "27-inch 4K monitor, 1,350 yuan, 90% new, 3 km away". Price, condition and distance are all conditions Xiaobei set, so he can judge at a glance that this is what he wants. A bad title reads: "New items matching your interests are here, come and see!" It carries no information at all, and curiosity runs out after two taps.

The landing page after the tap matters just as much: do not dump the user on a generic search results page and make them scroll again. Land directly on that item's detail page, with one line at the top explaining why this was pushed: "Matches your saved search: 27-inch 4K monitor, under 1500". That line does two jobs: it confirms "this alert came for me", and it gives the user a chance to correct the match — change the conditions, or delete the saved search right there. Every alert is a chance to check whether the conditions are still right. Do not waste it.

Do not skip the image either. A push with the item's main photo opens at a completely different rate from one without. The photo must be real — the one the seller uploaded, not a stock image from your library. The image in the alert and the image on the detail page must be the same one; a mismatch instantly reads as "not as described".

The better the conditions, the more the alert is worth

The quality of a saved search depends on how well the user wrote the conditions in the first place. Conditions that are too broad flood the user with alerts: Xiaobei saved only "monitor" — no size, no budget, no condition — and gets a dozen alerts a day, including 32-inch models, items over 2,000 yuan, and ones with dead pixels. The more you push, the faster trust drains away. Conditions that are too strict starve the alerts: Xiaobei caps the budget at exactly 1,500 yuan, accepts only "almost new" condition, and only within 3 km — and gets no alert for a month, concluding the feature is broken. A good saving experience helps the user tune the range to "just right" at the moment of saving: suggestions for keywords, a price range, and sensible filter defaults.

Three concrete things help. First, when the user saves, show a line like "With these conditions, 12 new items matched in the past 7 days", so they know what frequency to expect. If only 1 item matched in the past week, suggest "These conditions may be too strict — try raising the budget to 1,800 yuan". If 80 matched, suggest "These conditions may be too broad — adding a condition filter would make alerts much more precise". That line costs very little — one run of the match against history — but it turns "save and forget" into "save with a clear expectation". Second, do not ask for a single price ceiling; ask for a range. Under 1,500 yuan is the ceiling, but under 1,200 yuan might be a "bargain price" worth its own switch: "push me immediately below 1,200 yuan". A range lets alerts have urgency levels: routine pushes for in-range matches, immediate pushes for bargains. Third, keywords should support "not this": Xiaobei does not want dead pixels, so let him add an exclusion term for them. Exclusion terms are among the most-used conditions in secondhand trading; without them, users must eyeball every listing themselves.

Conditions also need a regular check-up. If a saved search has matched nothing new for 30 days, the product should ask: are the conditions too strict, or is this item simply rare? For the former, suggest loosening; for the latter, suggest leaving it alone — some secondhand items appear once every six months, and the user can wait, but needs to know "the feature is not broken; the goods are just rare". That check-up message is itself a form of alert: it tells the user the subscription is still working, there is just nothing new right now. "No news is also news" beats a month of silence. Conversely, if a saved search pushes three or four alerts a day, ask proactively whether to tighten it. Users rarely notice alert frequency themselves; only the product sees the full picture.

The management page: the more saved, the better it must be managed

Once a user has more than five saved searches, the management page stops being decoration and becomes essential. It has to answer three questions: what did I save, what state is each one in, and how do I change it? Each entry in the list should show its name, its core conditions, when it last matched something new, and whether alerts are on. States must be readable at a glance: matched something new this week, quiet for a long time, expired — kept apart. For searches quiet for a long time, prompt proactively: "No new items for this search in 30 days. Loosen the conditions — for example, raise the budget to 1,800 yuan — or delete it?" That sentence helps the user tidy up and helps you retire a subscription that has been occupying push quota for nothing.

Editing conditions must be easier than creating: tap a saved search and you are back on the search page with the conditions already filled in; change them, save, and it updates rather than creating a duplicate. Pausing and deleting must be equally easy: pausing means "no alerts for now, but I am not ready to delete", deleting means "gone for good". Unsubscribing must take one tap; any design that asks for a questionnaire first and then an email confirmation is spending trust. Expiry needs a default too — 90 days, for example — after which alerts stop automatically and the user is prompted to renew, instead of pushing forever. A subscription the user has forgotten is the subscription most likely to become harassment.

Three tiers: from working to good

Tier one is the minimum viable version: saved search plus a daily email or in-app digest. The user saves conditions and gets one message at a fixed time each day — "2 new items match your search today" — and taps through to the list. This tier is the cheapest to build: the core is "store the conditions" and "run the match on a schedule". It fits a project that has just shipped search and wants to learn whether anyone wants to save before investing in real-time pushes.

Tier two is the standard version: tier one plus real-time push, deduplication, merging, a frequency cap and quiet hours, plus a management page. This tier solves three things — push accurately, push sparingly, manage easily — and it is where most products should stop. Alert open rates and completion rates start to mean something here, and the five numbers in the funnel become the way to measure them.

Tier three is the advanced version: price-range alerts ("push me when one drops below 1,400 yuan"), price-drop alerts on saved items, smart merging across similar saved searches, and proactive suggestions when conditions are too strict ("widening to 32-inch would triple your options"). Each item in tier three is only worth building with data behind it: build "loosen the conditions" suggestions after seeing users manually loosen conditions on the management page; build price-drop alerts after seeing support tickets about price drops. Do not build tier three from imagination.

Launch checklist and metrics: watch five numbers

Run through this checklist before launch: does the save button appear at the top of the results page and on the zero-result page; does one tap save it, with a default name that needs no typing; are duplicate saves intercepted or flagged; do alerts fire only for newly listed items that match the conditions; do deduplication, merging, the frequency cap and quiet hours all work; does the alert title carry key facts such as price, condition and distance; does the tap land directly on the item page and state why it matched; can the management page edit conditions, pause, and unsubscribe in one tap; do saved searches have a 90-day default expiry; and on privacy, are saved conditions used only for matching, never for ad targeting?

After launch, watch only five numbers — the five in the funnel: of 1,000 searches, how many are saved; of saved searches, how many turn alerts on; of those, how many open; of opens, how many complete the goal; plus the unsubscribe rate. If the save rate will not rise, check the button placement and the save flow before touching push copy. If the open rate falls, check whether the trigger is too coarse and whether the title carries key facts before raising push frequency. If the unsubscribe rate rises, stop immediately and check what crossed the line recently. Metrics exist to locate problems, not to decorate reports.

One last word on scope. This guide covers one stretch only: "save the conditions and wait for new results". It does not cover how to build a single search (that is the search guide's job), how to build a notification centre (that is the notification system guide's job), or how to build favourites (that is the favourites guide's job). The value of saved search is not the feature itself but the change it makes in the relationship between user and product: the user moves from "coming back every day to try my luck" to "just tell me when". For a vibe project, that change costs very little — a save button, a scheduled match, a restrained set of push rules — but it keeps precisely the users whose needs are clearest. Someone like Xiaobei may buy only one monitor in a month; but afterwards, he will remember who told him at exactly the moment he wanted to buy.

Views 10Comments 0

Comments (0)

ME
0/1000
Loading comments...
Browse projectsPublish your project

Related articles

Real photo: a laptop, tablet, and phone on a wooden table showing the same green-leaf website, the phone view with navigation in one column, a potted plant nearby
Guide
Squashed on phones? Responsive layout and media queries, explained plainly

One page should stand up on a phone, a tablet, and a computer. This guide follows MDN's Responsive web design, Using media queries, and the viewport meta reference: fluid HTML, the viewport tag, fluid grids and max-width, images that never burst their container, media query syntax, breakpoints set by content, mobile first, responsive type, testing capability instead of width, and a pre-launch checklist.

Mobile AppsDesign ExperienceFrontend Engineering
Real photo: a bank of light gray metal filing cabinets with one drawer pulled open, revealing rows of hanging file folders inside
Guide
A real database inside the browser: IndexedDB local storage, explained plainly

localStorage only stores strings and has no indexes. IndexedDB is a real database in the browser: structured data, far larger capacity, indexes, and transactions. Following MDN's Using IndexedDB, IDBDatabase, and IDBObjectStore: the five-step routine, opening, upgrades, keys, indexes, transactions, add versus put, get, cursors, ranges, and multi-tab upgrades.

Tool TipsData WorkflowsFrontend Engineering
Real photo: hands typing on a MacBook Pro showing JavaScript code that changes HTML content, with a white mouse, papers, and binders on a white desk
Guide
Functions without the confusion: JavaScript functions, parameters, and arrow functions, explained plainly

You already call functions every day; this guide names what you have been doing. Following MDN's Functions, Arrow function expressions, and Rest parameters: declarations versus expressions, parameters and arguments, default values, rest parameters, spread syntax, return values, arrow function shorthand, the this trap, callbacks, and scope.

Tool TipsFrontend EngineeringAI Coding