Most retail media pages are run by two systems that have never met.

A lot of vendors talk about unifying ranking. Fewer of them can do it without adding a visible delay to the page, because most weren't built to call two systems at once and reconcile the results before the shopper finishes their search. Ask any vendor to walk you through that sequence step by step and the answer tells you whether they've actually solved it.

They've never met because nobody built them to. One decides what's organic. One decides what's sponsored. Instead, the two sides settled for a truce, drawn up in advance and never revisited: fixed tiles. Position 3 is always an ad. Position 7 is always an ad. Nobody asked if the shopper wanted an ad there. Nobody asked if it's relevant.

That's not a ranking strategy. That's a treaty.

This piece is about what replaces the treaty: how Pentaleap actually sits inside your stack, what happens in the moment a shopper searches, and what to ask any vendor, including us, before you commit to one.

TL;DR: Pentaleap sits between your search engine and your ad server as one decision layer. It requests organic and sponsored products in parallel, ranks the whole grid against one relevance and margin objective, and hands back a single result. Your search engine still searches. Your ad server still serves. Pentaleap just decides what goes where. If you're evaluating vendors right now, the questions at the end of this piece should help.

‍

The problem: two systems, one grid

A search engine is good at one job: understanding what a shopper is looking for. An ad server is good at a different job: understanding who paid to be seen. Neither was built to referee between the two, so most retailers don't ask them to. They just carve up the page and call it a compromise.

The result is predictable. Sponsored products get shown whether or not they're relevant, because the tile was theirs regardless of fit. Organic products get buried under products nobody was looking for. Shoppers notice. Legacy fixed-tile serving delivers click-through rates that are 50% or less of what unified ranking achieves, according to Pentaleap's own A/B testing.

Fixed tiles aren't a ranking decision. They're an absence of one. And an absence of a decision is still a decision, it's just one that costs you revenue and relevance at the same time.

Here's the part that makes evaluating vendors harder than it should be: almost everyone in this category now says "unified ranking" and "AI-powered." Three years ago this was a two-vendor market. Today it's ten to fifteen specialized companies, and most of them are pitching similar language regardless of what's actually running underneath. The question worth asking isn't who built the AI. It's what that AI can actually see when it makes a ranking decision, and whether it's making one decision or reconciling two after the fact.

‍

What Pentaleap actually is

Pentaleap is the optimization layer that sits between your incumbent ad server and your product grid. Not a replacement for either. A referee that both report to.

Your search engine keeps scoring organic relevance. Your ad server keeps managing bids and budgets. Pentaleap owns the one decision neither was ever built to make: what actually gets shown, and where, once both sides have spoken.

No rip and replace. No new frontend to learn. The same A/B testing that showed the CTR gap also showed 80% to 140% more ad revenue once relevance improved and more paid clicks followed, without touching the systems already in place.

Michael Krans, VP of Retail Media at Macy's, put it this way: the partnership helps Macy's build a more open and flexible retail media ecosystem, one that keeps ads relevant and protects the shopper experience while opening access to new demand sources that drive growth. That's the test worth applying to any vendor's pitch: does it make the page more relevant, or does it just add more ways to sell inventory?

‍

The request, step by step

Unifying ranking doesn't mean unifying systems. It means putting one layer in between that talks to both, and it's worth understanding what that actually requires technically, because "unified ranking" is easy to claim and harder to do without adding latency.

Here's what happens in the time it takes a shopper to hit enter:

1. The shopper searches. The retailer's backend requests products for that keyword.
2. Pentaleap calls both sides, in parallel. Organic products from the search engine. Sponsored products with bids from the ad server and any connected demand sources. Same moment, not sequential.
3. Both responses land back at Pentaleap. Relevance scores on one side. Bids on the other.
4. Pentaleap unifies the ranking. Relevance scores and bids don't get compared as separate line items; they're weighed together, in one grid, under one logic. A high-bid product still has to earn its position on relevance. A high-relevance organic product doesn't lose its spot to a lower-relevance ad just because that ad paid for a reserved tile. 
5. The retailer's backend serves the unified grid. The shopper sees one ranked list. No visible seam between where a product came from.
6. Tracking splits by design. Impressions and clicks are tracked in the browser, where the page already knows what's on screen. Conversions are tracked in the backend, where that data is harder to tamper with.

Parallel, not sequential. That's the detail that keeps this fast enough to run in real time without shoppers noticing a thing.

‍

Where the demand comes from

An optimization layer is only as useful as what it can plug into. Pentaleap connects through whatever method fits the source, not a single proprietary pipe:

Legacy ad servers, by contrast, typically only reach the first row on this list, commerce management platforms, and stop there. That's the real gap: it isn't that legacy setups have no demand connections, it's that most of the rows above are closed to them.

RTB is the one that changes the ceiling. It pulls bids from sources like Google and Amazon in real time, at the moment of the request, not on a batch job from the night before. That's the difference between being limited to whoever your incumbent ad server already had a deal with, and being open to the wider market, without a rebuild to get there.

This matters more than it sounds. A retailer locked into one vendor's demand relationships is a retailer whose ad revenue ceiling was set by someone else's business development team. Ask a prospective vendor for their version of this table. If they can't name specific demand sources and connection types, they're describing an ambition, not a capability.

‍

Two ways to deploy it, and how to pick

Not every retailer needs the same slice, and the right starting point depends on what you're trying to prove first.

Not every retailer needs the full stack on day one, and the right starting point depends on what you're trying to prove first.

Relevance layer only is where every retailer should start, and for good reason: it's the fastest path to a first result, because it proves the concept without asking anyone, internally or on the brand side, to change how they already work. Add unified ranking on top of your existing ad server and search engine. No new demand yet. Choose this if you want evidence before you touch anything else.

Full stack is the right call once relevance is proven and the constraint has shifted to demand. Pentaleap governs ranking, opens new demand through RTB and API, and can host the campaign UI directly. This is the only configuration that touches all three levers at once: relevance, inventory, and demand. Choose this when you already have proof, not as a first move.

The signal to move from one stage to the next isn't a calendar date. It's whether the current stage has actually proven itself in your own numbers. Retailers who skip ahead because a demo looked impressive tend to be the ones who end up troubleshooting in production.

‍

What this looks like in practice

The Home Depot took a different path, and it's worth walking through because it shows the decoupling in action rather than just describing it. Microsoft PromoteIQ originally served Sponsored Products into fixed tiles, disconnected from organic search entirely. Today, The Home Depot runs cross-channel campaigns through frontend orchestration tool Vantage, while Pentaleap unifies the ranking underneath, independently of which orchestration tool sits on top. Melanie Babcock, VP of Retail Media at The Home Depot, says Pentaleap helps the retailer serve more relevant ads onsite, and that collaborating with Vantage let the team streamline work across every function into one platform. Pacvue is next, coming on as an API demand partner that plugs directly into the same layer.

Two retailers, two different starting stacks, two different sequences, but the same underlying move in both cases: unify the ranking first, then decide separately what to do about orchestration and demand. Neither one required tearing anything out to get there.

‍

Questions to ask vendors before you commit

Most vendors in this category now use the same words: unified ranking, relevance, AI-powered. The pitch decks read almost identically. What actually separates a real unified ranking system from one that's relabeled a legacy setup isn't the vocabulary. It's whether the mechanism underneath survives specific questions. We already asked one of them earlier in this piece: walk through the request sequence, step by step. Here's the rest of the list.

• Does ranking happen in one decision, or two decisions reconciled afterward? Ask the vendor to walk through their request flow the way this piece just did. If they can't get specific about sequencing, they likely haven't built it.
• What's their batch size for pagination, and why that number? A vague answer usually means duplicate products and broken cross-page boosting in production, not in the pitch deck.
• What does their demand connection table actually look like? Ask for the equivalent of the table above, by name. "We're open" is not an answer. "Here are our six demand sources and how each connects" is.
• What do their current customers say, specifically, about relevance, not just revenue? Revenue lifts are easy to claim. Ask for a named reference who'll speak to whether shopper experience actually improved.
• Can you start with the smallest deployment and prove it before expanding? If a vendor's only offer is a full rebuild, that's a sign that relevance improvements aren't separable from a bigger commitment, which should raise questions about how much of the pitch is architecture versus lock-in.

‍

Key takeaways

• Pentaleap doesn't replace your search engine or ad server. It sits between them and makes the one decision neither was built to make: what gets shown, and where.
• With some vendors now using similar "unified ranking" language, the differentiator to check for is what the system can actually see when it ranks, not the marketing around it.
• The request flow runs in parallel, not sequentially, which is what keeps unified ranking fast enough for real-time serving. Ask any vendor to explain their version of this sequence.
• Demand connections use whatever method fits the source: API for commerce management platforms, orchestration tools, and DSPs; RTB for ad networks, incumbent retail media vendors, and traditional search companies.
• Deployment isn't all-or-nothing. Start with the smallest step that proves the concept, then expand once you have your own numbers, not a demo, to justify the next stage.
• Macy's and The Home Depot took different paths to the same underlying move: unify ranking first, decide on orchestration and demand separately.

‍

Frequently Asked Questions

Does this replace my ad server or search engine?

Not required to. Pentaleap integrates alongside both from day one, and can take over ad serving over time if you choose. Either way, the ranking decision runs through Pentaleap. 

‍Does this slow down page load? 

No. Search engine and ad server calls run in parallel, not one after the other, which is what keeps response times where real-time serving needs them.

What if I'm not ready to open new demand sources? 

Then don't. The relevance-layer-only deployment unifies ranking using only what you already have. New demand is a later step, not a requirement.

How does programmatic demand like Google or Amazon actually connect? 

Through RTB. Bids come in real time, at the moment of the request, not through a batch process or a manual handoff.

Do I need to change my campaign UI to use Pentaleap?

No, not for the relevance-layer-only deployment. Brands can keep managing campaigns through your existing UI, as both Macy's and The Home Depot have done with their respective platforms. 

How much of my product grid does Pentaleap need to see at once? 

Somewhere between 40 and 80 products per request, starting from page one, position one. That range is what balances proper cross-page boosting against response time.

What should I actually ask a vendor before signing anything? 

Start with the five questions above. If a vendor can answer all five with specifics rather than reassurance, you're evaluating a real system. If they can't, you're evaluating a pitch.

‍

Get in touch with Pentaleap to learn more.

‍