How Merciv’s AI Connects Sales Data, Reviews and Social Signals for Retailers

Shaia Erlbaum, co-founder and CEO of Merciv, breaks down how connecting internal archives with real-time web signals helps merchants spot fit issues early, filter social noise and protect open-to-buy commitments.

In modern apparel retail, brand leaders aren’t suffering from a lack of data — they are overwhelmed by disconnected signals. Between fragmented style codes, misread sizing complaints and fleeting social trends, crucial insights often remain trapped in organizational silos until after markdowns hit or reorder windows close.

Here, Shaia Erlbaum, co-founder and CEO of Merciv Inc., discusses how the platform functions as a “correlations engine” that unifys internal archives, POS history, review language and social conversation into cited, merchant-ready strategy without replacing human creative judgment.

Street Talk: How does Merciv’s AI unify sell-through, reviews, resale and social conversation data to help merchants identify the root cause of slow-performing styles and adjust OTB commitments before markdowns lock? 

Shaia Erlbaum: The starting problem is that a single style lives under at least three different identities–a wholesale item number, an internal ERP style code, and whatever consumers actually call it in a review or a TikTok comment. Traditionally, those never get reconciled fast enough. Merciv resolves all of them into one entity so that retailer sell-through, DTC performance, review language, resale activity and social conversation about that style can be read against each other, and, more importantly, against what a brand already knows. Most brands are sitting on years of their own research, POS history and survey work that never gets opened again. That archive is the control group for everything happening on the open web, and it’s the half of the equation most tools ignore entirely. 

If DTC sell-through is holding up but wholesale is softening, that usually points to a distribution or execution problem, not a demand problem. If it’s the reverse, DTC drops while wholesale holds, that’s often an early signal of a perception shift, and wholesale typically catches up within a quarter or two if nothing changes. Merchants have historically had to wait 60 to 70 days for that kind of signal to surface through manual channels, and we’re compressing that to minutes or days. 

How does Merciv’s platform analyze search, creator, and consumer data to separate short-lived social feed noise from compounding demand, ensuring marketing teams optimize campaign spend and drop timing? 

Social feed activity is the loudest signal and often the least reliable one, because a single creator or a single platform can make anything look like a trend for a week. We treat a spike as noise until it’s confirmed across multiple independent layers: has search behavior evolved from broad, branded queries into more specific, comparison-level intent; is the conversation showing up across more than one platform rather than riding one creator’s reach; and is it actually converting, is it showing up in reviews and repeat purchase behavior, not just impressions. 

Another valuable check in our validation analysis is looking at a shift teams documented two years ago and shelved or an insight teams registered and then stored away in sharepoint are often the best pieces of evidence they have that a current shift or insight is real. Correlating across layers is how you get to confidence sooner. 

How does Merciv leverage AI to surface early-season fit and sizing issues from reviews and customer conversation, allowing brands to cut fit-driven returns before reorder windows close? 

Fit and sizing complaints are one of the largest and most misread categories in apparel returns as roughly half of returns trace back to sizing in some form, but brands typically lump it into one bucket when it’s actually three distinct problems: inconsistent grading across the size run, a mismatch between the product detail page and the actual size chart, and genuine fit preference. Each one needs a different fix, and none of them show up clearly in a returns dashboard until well after the damage is done. 

The real signals live across multiple surface areas. A review tells you one thing, and that might be validated by the way consumers change their keyword choices on google search (for example). But simply searching for keywords won’t give you a complete answer; recognizing shifting language across reviews, socials, and forums together allows our agents to identify the new terminology customers are using or might use, and the combination of all of these data points provide the full picture of what’s actually changing, what’s being said, and what that should indicate strategically. 

How does Merciv differentiate itself from legacy social listening and trend forecasting tools by providing cited, merchant-ready evidence rather than non-actionable sentiment charts? 

Brand teams may have more consumer information than ever, but that does not always translate into clarity. Merciv uses AI to bring disconnected signals together and surface insights that can inform decisions across brand, product, growth, marketing, and research.  

If the objective is to find where something lives, several excellent knowledge management solutions might help you find the file. If you need one or two clean documents synthesized into a clean answer, any frontier AI model can do that now. If you need baseline social listening, there are point solutions, and same for CX reporting, same for innovation research, etc. 

Importantly, Merciv isn’t just a database. It’s a correlations engine. A database recalls what you already know. A correlations engine drives strategy your team can act on by connecting the full corpus of research you’ve built/collected to what consumers are actually doing and saying on the open web this week. 

In practice that means your internal knowledge, POS data, survey data and live consumer signal all working together. The comparison that matters is that our correlations inform strategy  rather than simply providing answers. And where legacy systems provide a limited, manually-directed lens into a portion of consumer conversation, Merciv covers the landscape. Lastly, where incumbent systems require months of work to stand up dashboards, then even more manual time maintaining them every time a category or a competitor set shifts, Merciv enables day-one reporting interfaces, mapped to the exact preferences of each customer brand. 

How does Merciv balance AI-driven consumer insights with a merchant’s creative judgment while maintaining enterprise-grade data privacy and zero-training guarantees? 

We’ve never positioned Merciv as a replacement for a merchant’s or a designer’s judgment, and I don’t think that’s the right ambition for AI in this category. The risk we see isn’t AI overruling good taste. It’s AI laundering a weak signal into false confidence, or an answer delivered fluently enough that nobody thinks to ask what it rests on. That’s the whole reason provenance is architectural for us. If a merchant can’t click into why, they shouldn’t be asked to act on it. Our job is to shorten the distance between “here’s what’s changing and here’s the evidence” and a human decision, not to remove the human from that decision.  On privacy: our customers’ data never trains our models, or anyone else’s. Every customer sits in an isolated tenant, data is encrypted at rest and in transit. 

Related article: How Agentic AI is Rewiring the Global Retail Engine

Street Talk is the weekly newsletter for retail and fashion executives who want analysis, not noise. New issues every Thursday. Subscribe today.