Sequence product data, merchandising, and feeds before peak traffic hits.

U.S. shoppers spent a record $257.8 billion online during the 2025 holiday season, and Cyber Monday alone drove $14.25 billion, the largest online shopping day the country has ever recorded.¹ Every retailer is pouring budget into driving a share of that traffic to their site. Far fewer are preparing the systems that decide what those shoppers see once they arrive.

During Black Friday and Cyber Monday (BFCM), visits climb faster than any team can hand-rank products, rewrite thin product descriptions, or catch a broken product feed. The work that turns a visit into an order does not scale with the visit count, and that is where peak revenue leaks. It is also the work AI is now good enough to handle, if you start on the right pieces early enough.

At a glance: three AI jobs to do before BFCM (callout box)

• Now, product data: Enrich your product catalog so onsite search returns results and AI assistants can read your products. This has the longest lead time and the highest payoff.
• Through the fall, merchandising: Let AI draft your peak campaigns and product rankings, then have your team edit and keep control.
• Before peak, product feeds: Run continuous AI feed audits so nothing is disapproved or stale across Google, social, and marketplaces.
• At peak, personalization: Personalize in real time for first-time and AI-referred shoppers.

Start with a product data and feed audit to see where revenue is leaking today.

Where peak revenue leaks

For most retailers, the constraint at peak has never been demand. It is how much skilled manual work a small team can do in a compressed window. Merchandisers hand-rank products for holiday campaigns. Analysts tune search rules query by query. Someone reconciles a product feed that a channel rejected the night before a sale. It all drives revenue, and it all hits a ceiling exactly when traffic is highest.

The category is moving from that manual model toward AI-assisted and, increasingly, AI-executed work, where the operator sets strategy and reviews output while AI handles the volume. The retailers who benefit most this BFCM prepare early, clearing three chokepoints on a schedule: product discovery, merchandising, and product feeds. Each caps revenue differently and has its own lead time, so sequence matters as much as tool choice.

Start now: enrich your product data so shoppers can find products

Product data enrichment is the highest-impact fix and the one that needs the most lead time, which is why it belongs at the top of the list in August rather than November. When a shopper searches your site for “waterproof winter boots” and your product descriptions only say “boots,” the search returns nothing useful and the shopper leaves. Multiply that by peak traffic and the lost revenue adds up fast. The same thin product data also keeps you invisible off your site, on Google Shopping, marketplaces, and the AI assistants shoppers use to research purchases.

Suhas Gudihal, CTO of Athos Commerce, estimates that fewer than 10% of merchants have product data that is structured and machine-readable today.

If I don’t have that metadata, structured and machine-readable, that means your product is not visible to the agent at all.

Suhas Gudihal

CTO, Athos Commerce

AI closes that shortfall faster than a manual attribute project ever could. It reads product images and existing copy to generate the attributes buyers search on, standardizes colors and synonyms so “coffee” also returns under “brown,” and formats the catalog correctly for each channel. Retailers that fixed onsite discovery this way have seen search-led revenue climb 300% (John Smedley) and revenue from search rise 74% (Paul Smith). Enriched now, that product catalog pays off on your own site and everywhere else shoppers look for you at peak.

Next: let AI draft your merchandising campaigns

Building merchandising campaigns for peak is where skilled time disappears. A single BFCM campaign, with its curated collections, promotional rankings, and category rules, can take a merchandiser about three days by hand. Run several across regions and product categories, and the calendar fills before the season starts.

AI changes the economics of that work. It generates a strong first draft of a campaign, collection, or set of product rankings, and the merchandiser edits rather than builds from scratch. Suhas estimates AI gets a campaign “70 to 80% right,” turning a three-day build into roughly one day while the judgment calls stay with the person who owns the results. The merchandiser keeps control, and their scarce peak-season hours move from assembly to the strategy that improves the work.

Athos customers already see this. Early Settler saved more than 10 hours a week, and Aje freed up two to three full days a week through merchandising automation. At peak, that reclaimed time decides whether you run the campaigns you planned or only the ones you had hours left to finish.

Before peak: validate your product feeds across every channel

A product feed error is a minor annoyance in July and a revenue problem on Black Friday. When a channel disapproves products or shows stale prices and stock, those products vanish from the placements you are paying to win at peak. More of peak now happens off your website. U.S. social commerce reached about $87 billion in 2025, up 21.5% year over year, and TikTok Shop alone sold more than $500 million over the four-day BFCM window.²³ Each of those channels reads your product feed against its own format and rules.

Checking dozens of product feeds against dozens of channel specifications by hand is not realistic in the weeks before peak. AI feed audits run continuously instead, catching missing attributes, disapprovals, and errors before they cost you traffic and tuning each feed for its channel. Accent Group saw revenue rise 60% within eight weeks after improving feed accuracy, and River Island grew revenue 49% after launching product feeds on Google Shopping. Because Athos syndicates to more than 1,500 channels from one product data model, the same enriched catalog stays consistent wherever a shopper finds it.

The feed is no longer a syndication artifact. It is your agent-facing operating system.

Suhas Gudihal

CTO, Athos Commerce

At peak: personalize in real time, for humans and AI shoppers

Peak brings a flood of first-time visitors who have no purchase history with you, the traditional blind spot for personalization. AI removes that limitation. It reads live behavioral signals, including time on page, scroll depth, and click patterns, and combines them with social proof to tailor what a brand-new shopper sees in the moment, rather than waiting for history it will never collect in a single session. Michael Stars saw revenue per visit rise 951% once its experience adapted this way, with the team setting the rules and AI executing at volume.

There is a second kind of shopper to prepare for this year, and it is not human. Traffic to U.S. retail sites from generative-AI sources grew 693.4% year over year last holiday season, and shoppers who arrived through those AI sources were about 38% more likely to complete a purchase on Black Friday than shoppers from other sources.¹⁴ Those buyers convert well, but only when the AI can read your product data well enough to recommend you in the first place. The same enrichment that helps a person find the right product makes your catalog legible to an AI assistant deciding what to recommend.

You need to design for verifiability, not persuasion. Claims or specifications that are measurable, attributed, or structured will win.

Suhas Gudihal

CTO, Athos Commerce

One foundation behind all three fixes

These fixes sequence well because they share one foundation, your product data. The same enriched product data powers onsite search and AI-drafted merchandising, and it keeps feeds clean and your catalog legible to the AI assistants sending high-converting traffic. Improve that data in a stack of disconnected point tools and you repeat the work in each one. On a single intelligent discovery platform where search, merchandising, personalization, and feed management draw on the same catalog, one improvement reaches every downstream job.

That is why the sequence in the callout above runs as one motion rather than four separate projects. If you do only one thing this month, run a product data and feed audit. It is the lowest-effort way to see where peak revenue is leaking today and which of the three chokepoints will cost you most when traffic spikes.

The advantage this BFCM goes to teams that used the months before it to clear the manual chokepoints in discovery, merchandising, and feeds. When record traffic and a growing share of AI shoppers arrive, their enriched product data is ready to convert both. Peak season has always been the test of whether your discovery infrastructure works. This year, it also tests whether that infrastructure is ready for who, and what, is doing the shopping.

Sources & Further Reading

1. Adobe. “Holiday Shopping Season Drove a Record $257.8 Billion Online with Consumers Embracing Generative AI Tools.” Adobe News, January 7, 2026. https://news.adobe.com/news/2026/01/adobe-holiday-shopping-season
2. eMarketer. “TikTok Shop Makes Up Nearly 20% of Social Commerce in 2025.” eMarketer, 2025. https://www.emarketer.com/press-releases/tiktok-shop-makes-up-nearly-20-of-social-commerce-in-2025
3. Erdly, Catherine. “The TikTok Shop Sellers Behind Social Commerce’s $87 Billion Surge.” Forbes, February 16, 2026. https://www.forbes.com/sites/catherineerdly/2026/02/16/the-tiktok-shop-sellers-behind-social-commerces-87-billion-surge/
4. Adobe Analytics data, reported in Digital Commerce 360. “Generative AI Shifts Online Holiday Shopping Traffic in 2025.” Digital Commerce 360, January 13, 2026. https://www.digitalcommerce360.com/2026/01/13/generative-ai-online-holiday-shopping-traffic-2025/