New arrivals vs bestsellers is a false choice. How to solve the cold-start problem on Shopify collection pages with freshness slots, fair metrics and pins, without letting bestsellers lock the top row forever.
New arrivals vs bestsellers is the oldest argument in collection merchandising, and on most Shopify stores it is settled by accident. Either the collection is sorted by 'Best selling', so proven products hold the top row forever and new arrivals are buried below them, or it is sorted by 'Newest', so every untested upload takes the most valuable position on the page and proven sellers slide down with each new product. Both are wrong, and both are wrong in ways that get worse the longer they run.
The right answer is not a compromise between the two sorts. It is a page with a structure: a place where proven sellers earn their spot, a place where new products get a guaranteed, time-limited trial, and a fair metric that lets the new products that perform graduate into the main ranking. This post explains the cold-start problem behind the argument, how freshness slots solve it, which metrics treat new products fairly, and how to stop bestsellers locking the top row forever. We show how each piece works in SellRank, our Shopify collection sort app, and use illustrative numbers throughout, labelled as such.
The cold-start problem, in one paragraph
Any ranking based on performance needs data, and a new product has none. So it ranks near the bottom. At the bottom, it gets very few views, because most shoppers never scroll that far. With few views it collects few add-to-carts and fewer orders. With few orders it stays at the bottom. The ranking is not making a judgement about the product; it is simply reflecting the position the ranking itself put the product in. That loop is the cold-start problem, and it means a store with a pure performance sort can go months without a new product ever reaching the first screen, however good some of them are.
Why bestsellers lock the top row forever
The same loop runs in reverse at the top of the page. The product at position one gets the most views, so it collects the most orders, so it stays at position one. Some of that is merit: it may genuinely be your best product. Some of it is position: a product at position one would outsell a product at position forty even if the two were identical, because far more people see it.
Merchandisers call this position bias, and it is why 'Best selling' and other volume-based sorts are so sticky. Illustrative example: a skincare store's best-selling serum has held position one on 'Skincare' for a year. A new serum launched three months ago converts better per view, but it sits at position 28 and gets a small fraction of the views. On orders, the old serum wins every week. On conversion per view, the new one would. The store will never find out under a volume sort, because the new product never gets enough exposure to prove itself.
- Volume metrics, such as orders, revenue and units, reward exposure as much as quality.
- Exposure is decided by position, which is decided by the previous ranking.
- So the incumbents at the top keep winning, and challengers never get the traffic to challenge.
- Over time the first row becomes a museum of what sold well a year ago.
Why sorting by 'Newest' is not the answer either
The obvious fix, sorting by date, overcorrects. Newness is not evidence. A product uploaded this morning with no sales, no reviews and perhaps not even its final photography takes the best position on the page, purely for being recent. Proven sellers slide down one position every time a product is added, and on a store that uploads frequently, the first screen becomes a rotating showcase of untested stock.
Newest-first is right for exactly one kind of page: a genuine New Arrivals collection, where recency is the point. On every other collection it trades one kind of unfairness for another. Our breakdown of every Shopify collection sort option covers where the built-in date sort does and does not fit.
What each mistake costs over a year
Getting the balance wrong is expensive in both directions, but the costs show up in different places, which is why stores rarely notice them.
A store that lets bestsellers lock the top row pays in missed winners. Every new product that would have outperformed an incumbent, if it had been seen, is revenue that never happened. It also pays at buying time: new lines look like failures because they never got exposure, so the buyer orders less of them next season, and the catalogue slowly converges on whatever sold two years ago. Customers notice before the dashboard does. Returning visitors see the same first row every time they arrive and stop browsing.
A store that lets newness lead everything pays in conversion. The most valuable positions on the page are permanently filled with untested products, some of which will turn out to be weak. Proven sellers that would convert cold ad traffic are two or three scrolls down. The store looks fresh and converts worse, and because the change was gradual, nobody connects the two.
Neither cost appears as a line on a report. Both are real, and both are avoidable with the same structure.
New arrivals vs bestsellers: structure the page, not just the sort
The way out of the new arrivals vs bestsellers argument is to stop asking one sort to do two jobs. A collection page has room for both, if you divide it:
- A proven slot at the top, where products earn their place on performance and in-stock availability.
- A freshness slot close behind it, where new products get guaranteed exposure for a limited time.
- A fair ranking for everything else, one that does not simply reward whoever had the most exposure last month.
- A human override for launches and hero products that the data does not yet know about.
In SellRank this structure is a single smart sort built from percentage slots. Each slot has its own strategy and conditions; slots fill top-down; anything unclaimed follows an 'everything else' order. If a slot does not have enough qualifying products, its space passes to the next slot, so nothing is left empty. Our post on percentage-slot merchandising for Shopify collections goes deeper on slot design in general; here we focus on the freshness problem specifically.
How a freshness slot works
A freshness slot is a share of the page reserved for your newest products, ranked by Newness and usually restricted to in-stock items. Because it is defined by position rather than by a list of named products, it rotates on its own: as new products arrive, they enter the slot at the top, and the oldest products in it are pushed out into the main ranking. Nobody has to decide when a product stops being new.
Illustrative example: a 150-product 'Tops' collection with a store that uploads about ten new tops a fortnight. The first 20%, 30 products, is a proven slot ranked by conversion rate. The next 10%, 15 products, is a freshness slot ranked by Newness. That puts the store's fifteen newest in-stock tops on positions 31 to 45, well within reach of any shopper who scrolls past the first few screens, for roughly three weeks each before newer arrivals push them on. After that, each one competes in the main ranking on the numbers it earned during its trial.
How big should a freshness slot be?
Size the slot from two numbers: how many new products you add in a typical period, and how long you want each one to get a trial. If you add ten products a fortnight and want each to have about three weeks of exposure, the slot needs room for around fifteen products. Then express that as a share of the collection.
- Stores that upload rarely, a few products a month, can use a small slot of 5 to 10%, or rely on pins for each launch.
- Stores that upload weekly usually land between 10 and 20%.
- Stores built around constant newness, such as fast fashion or daily drops, may give freshness 30% or more, and should still keep a proven slot above it.
- A dedicated New Arrivals collection can be mostly or entirely freshness, with a Trending slot so the new products that are already selling stay visible.
Remember that SellRank's percentages scale with the collection: 10% is 5 products on a 50-product page and 20 on a 200-product page. If a collection is small, check the preview to make sure the slot holds a sensible number of products.
Where should the freshness slot sit?
Usually second, just behind a proven slot, not first. The first row has to convert cold traffic, and untested products are a gamble there. Placing freshness immediately after the proven slot means every new product is visible within the first couple of scrolls, which on most stores is enough for it to collect a meaningful number of views, while the very top of the page stays on products with evidence.
There are exceptions. A brand whose customers come specifically for novelty, or a collection linked from a 'just landed' email, may want freshness first. Try both and let the performance view decide.
Choosing a fair metric for the rest of the page
A freshness slot gives new products a trial. What happens when the trial ends depends on the metric ranking the rest of the page. If that metric is a volume measure such as orders or revenue, the graduated products drop to the bottom again, because they have had weeks of exposure while the incumbents have had months. The trial was pointless.
The fix is to rank on rates rather than totals. Conversion rate, orders divided by views, normalises for exposure: a product that got 300 views and 12 orders converts at 4%, whether it was at position 5 or position 45. Add-to-cart rate does the same one step earlier in the funnel. A product that converted well during its trial will rank well afterwards, even though its absolute orders are small.
The small-sample trap in conversion rate
Rates have a weakness: they are noisy on tiny numbers. A product with 3 views and 1 order has a 33% conversion rate, which would put it at the top of any collection, and it is almost certainly luck. A purely rate-based ranking can therefore throw brand new products to the top of the page on one lucky sale, which is the Newest problem again by a different route.
- Blend the rate with a volume signal, so a product needs both a strong rate and some real activity to lead.
- Normalise before blending, so the blend is not dominated by whichever metric has the biggest raw numbers.
- Keep the freshness slot as the place new products are shown, so the main ranking can afford to be a little conservative about them.
- Watch the preview after a sync: if a product with a tiny view count is sitting at the top, the blend needs more weight on volume.
SellRank's Balanced score is designed for this. It blends several metrics, for example conversion rate, views and revenue, with weights into one 0 to 100 score, and it rank-normalises each metric first so no single metric, and no single outlier product, dominates. Ready-made blends such as Trending, Revenue driver and Customer favourites give you a starting point, and the preview shows the reason for each product's position so you can see when a blend is rewarding noise. Our post on ranking Shopify products by conversion rate covers the rate-versus-volume trade-off in more detail.
Where the engagement data comes from
Fair metrics for new products depend on counting views and add-to-carts from the first day a product is live. Shopify's order data only tells you about purchases; the views that make a conversion rate possible have to be measured on the storefront.
SellRank installs its own engagement pixel through Shopify's Web Pixels API the first time the app opens. It counts product views and add-to-carts per product per day as aggregates, with no customer data and nothing to paste into your theme. A product launched today starts accumulating views today. On Scale and Max, connecting GA4 can backfill up to 90 days of views and add-to-carts for your existing catalogue, which helps rankings work from the day you install. It cannot help a product that did not exist yet, which is exactly why a freshness slot matters even with a full data history.
Owning this data matters more for new products than for old ones. A product with two years of history can survive a few weeks of patchy analytics without its ranking moving much. A product in its first fortnight is being judged on a few hundred views, and if a consent banner or an ad blocker hides a slice of them, the judgement is made on the wrong numbers at exactly the moment it matters most. We cover the wider argument in our post on first-party pixels versus Google Analytics.
Launches: when a new product should skip the queue
Some new products should not wait in a freshness slot. The product your whole campaign is built around, the collaboration you have teased for weeks, the restock of a cult item: these deserve the first position on the day they go live, and the data cannot know that.
- Pin the launch product to the top of its collections in the preview. A pin always beats the automatic order and survives every re-sync.
- Prepare the launch order as a draft and schedule it to go live at the launch time, so collection pages match the campaign the moment the email lands.
- Decide in advance when the pin comes off. A week is typical. After that, the product should compete on the numbers it has earned.
- If the launch is a limited event, set a revert date on the draft and the previous order returns automatically.
Pins, demotions, drag and drop and exact placement are available on every SellRank plan, including Free. Drafts with scheduled go-live and automatic revert are on Grow and above.
A worked example: one new product's first six weeks
Here is an illustrative walk-through of a single product moving through the structure on a 120-product 'Candles' collection, with a proven slot of 20% ranked by a Balanced score and a freshness slot of 15% behind it. The numbers are invented to show the mechanics.
- Day 1: a new fig and cedar candle is uploaded. It has no data, so in a pure performance sort it would sit around position 110. Here it enters the freshness slot at position 25, the top of the slot, because it is the newest in-stock product.
- Week 1: at position 25 it sits within the second scroll on most phones. The pixel counts a few hundred views and a steady trickle of add-to-carts. Its conversion rate is well above the collection average, but on a small sample.
- Weeks 2 and 3: newer candles arrive and it drifts down through the freshness slot, still within easy reach. Its view count grows, the conversion rate holds, and its Balanced score climbs as the volume component catches up.
- Week 4: it leaves the freshness slot as the oldest product in it. It does not fall to the bottom, because the main ranking uses a normalised blend weighted towards conversion rate. It lands at position 30, just outside the proven slot.
- Weeks 5 and 6: its conversion stays strong and its revenue grows. Its Balanced score overtakes a long-standing vanilla candle whose conversion has been slipping, and it enters the proven slot at position 18. The vanilla candle moves down to position 26.
Nobody dragged anything. The candle earned its place in six weeks; under 'Best selling' it would probably still be near the bottom, and under 'Newest' it would have taken position one on day one and been pushed off the first screen by the next upload a week later, whatever its numbers.
What counts as 'new'? Restocks, re-launches and seasonal returns
Newness strategies usually key off when a product was created or published. That works for genuinely new products, but real catalogues are messier. A seasonal product that returns every winter was created years ago. A best-seller that was out of stock for two months comes back as effectively new to your customers. A re-photographed product with a new name may deserve a fresh trial.
- For seasonal returns, use a tag such as 'back-for-winter' and a slot condition, rather than relying on creation date.
- For restocks of popular products, a short pin on the return date usually does more than any automatic rule, because you know the demand is there.
- For relaunched products, consider whether a trial is really needed. If the product has years of history, a performance ranking already knows about it.
- Keep tags tidy. A 'new' tag that is never removed turns the freshness slot into a permanent slot for whatever was once tagged.
Newness across a large catalogue
Stores with dozens of collections cannot hand-design a freshness structure for every one. The practical approach is to design bespoke smart sorts for the ten or fifteen collections that carry most traffic, and apply a sensible default everywhere else. In SellRank, the global rule on Scale and Max does exactly that: one strategy, for example a Balanced score with an in-stock condition, applies to every collection that does not have its own smart sort, and a specific rule always wins over it. Product-level pins are off on the global rule; when a collection needs hand curation, it gets its own rule.
Keeping bestsellers honest
None of this is an argument against bestsellers. Proven products should lead most collections, because they are the best evidence you have about what converts. The goal is to make them keep earning the position rather than inherit it. Three changes do most of the work.
- Rank on recent performance, not lifetime totals, so a product that stops selling stops leading.
- Rank the proven slot on rates or a normalised blend, not raw volume, so position bias does not keep the incumbent on top by default.
- Put an in-stock condition on the proven slot, so a bestseller that sells out gives up its place immediately instead of absorbing clicks it cannot convert.
Illustrative example: a pet supplies store's 'Dog Toys' collection has had the same rope toy at position one for eighteen months. After switching the proven slot to a normalised blend of conversion rate and recent revenue, the rope toy stays in the first row, because it still converts well, but drops from position one to position four. Two newer toys that had been converting strongly from a freshness slot move above it. Nobody decided that the rope toy should move; the evidence did.
Recipes for the collections most stores run
Recipe: a core category collection
- Slot 1, first 20%: in-stock products ranked by the Balanced score, weighted towards conversion rate and recent revenue.
- Slot 2, next 10 to 15%: in-stock products ranked by Newness, the freshness slot.
- Everything else: Balanced score, with out-of-stock products parked at the end.
- Human layer: pin launches for a week; demote anything with a known quality or supply problem.
Recipe: a New Arrivals collection
- Slot 1, first 40%: in-stock products ranked by Newness.
- Slot 2, next 30%: in-stock products ranked by a Trending blend, so recent arrivals that are converting stay prominent as newer ones push in.
- Everything else: Newness, with out-of-stock products at the end.
Recipe: a Bestsellers collection
- Slot 1, first 50%: in-stock products ranked by recent Revenue or Orders.
- Slot 2, next 10%: in-stock products ranked by Conversion %, which lets a newer product with a strong rate break into the bestsellers page before its volume catches up.
- Everything else: Revenue, out-of-stock at the end.
Recipe: the homepage or 'Shop all' collection
- Slot 1, first 15%: in-stock products ranked by a Trending blend. This is your shop window.
- Slot 2, next 15%: in-stock products ranked by Newness.
- Slot 3, next 10%: in-stock products ranked by Customer favourites, so highly rated products get visibility.
- Everything else: Revenue.
How to tell whether new products are getting a fair shot
Fairness can be measured. Once the structure above is live, check a few things each month:
- How many products launched in the last 90 days have reached the first screen of their main collection on merit, not through a pin?
- What is the average view count a new product collects during its freshness trial? If it is too low to produce a meaningful conversion rate, the slot is too small or too far down the page.
- How long has the current position-one product held that position? If it has not changed in six months, the proven slot is probably ranking on volume.
- Has overall collection conversion held up since the freshness slot went in? A slot that costs conversion is sitting too high or is too large.
The last question is the important one, because freshness is not free. Giving untested products exposure costs something in the short term, and the aim is for that cost to be repaid by finding new winners sooner. SellRank's per-version performance view makes the trade-off visible: the version with the freshness slot and the version without it, each with the daily views, add-to-carts, orders and revenue attributed to the days it was live. Per-version performance is on Grow and above. Our post on how to prove a collection order change worked covers how to read that comparison without being fooled by seasonality or ad spend.
Common mistakes when balancing new and proven products
- Putting the freshness slot first on every collection, so the first row is always untested.
- Sizing the slot so small that each new product gets a few days of exposure, too little to produce any evidence.
- Ranking the rest of the page by volume, so graduated products fall straight back to the bottom.
- Ranking purely by conversion rate, so a product with one lucky sale jumps to position one.
- Pinning launches and never removing the pins, so the first row fills with last season's launches.
- Changing the structure every few days, so no version runs long enough to measure.
Why this needs automation to work
Everything above can, in theory, be done by hand in Shopify's manual sort. In practice a freshness slot is the part that defeats manual maintenance first, because it has to rotate every time a product is uploaded. Someone has to move each new product into the right band, move the oldest one out, recompute conversion rates for the graduates and re-drag the proven slot, every week, forever. Stores that try it fall back to 'Newest' or 'Best selling' within a month.
The categories of tool available are Shopify's built-in sorts, which each pick a side in the new arrivals vs bestsellers argument; a manual order, which is flexible but static; client-side reordering scripts, which change the display rather than the real order; and server-side sorting apps, which vary a great deal in whether they can compose slots, blend normalised metrics and measure the result.
SellRank was built to make this structure run itself. Percentage slots with per-slot strategies and conditions, a Newness strategy for freshness, a rank-normalised Balanced score for fair ranking, a first-party pixel that counts views from day one, pins and scheduled drafts for launches, and per-version performance to prove it all worked. It writes the result into Shopify's own manual sort, keeps re-sorting while the smart sort is enabled, and confirms the order position by position after each sync. For stores that want both new products and proven sellers to get what they deserve, that is the most complete answer we know of.
Give your next launch a fair trial
Pick the collection where new products struggle most to be seen. Add a proven slot ranked on a Balanced score, a freshness slot of 10 to 15% behind it, and an in-stock condition on both, then open the preview and look at where your newest products land. If they are within the first two scrolls and your proven sellers still hold the first row, sync it and let it run for a few weeks.
SellRank is free on one collection with the full editor, live preview, pins and one live sync. Grow is $29 a month for 15 collections with drafts, scheduling, versions and performance analytics; Scale is $69 a month for 75 collections with the global rule, GA4 backfill and profit metrics; Max is $249 a month for unlimited collections. See the full feature list on the SellRank product page and try the structure on your own catalogue.
