On October 23, the University of Michigan’s final consumer sentiment reading for October came in at 48.0, essentially flat with September and 10.4% below a year ago. Buying conditions for durables fell 18% in a single month, which the survey attributes to higher interest rates and concerns about affordability. Year-ahead inflation expectations rose to 4.7%.
Sentiment has kept drifting down since August, when the preliminary reading was 51.0. And yet every major holiday forecast still calls for a record season.
Record spending, cautious shoppers
Both things can be true at once. NRF expects November and December retail sales to grow 4.2% to $1.013 trillion, while warning that inflation, rising gas prices, and pressure on household budgets will hit consumers differently by income. Adobe forecasts a record $275.1 billion in U.S. online holiday spending, up 6.7%, with Cyber Week alone worth $47.5 billion.
The shape of the season has changed too. Adobe expects $95.8 billion in online spending in October alone, pulled forward by the October Prime Day event, and discounts of up to 30% off list price during Cyber Week. More of the season’s demand is arriving early and arriving on discount, which puts more pressure on every full-price conversion.
Two more shifts matter for whoever owns the storefront. Adobe expects mobile to drive 57.4% of holiday online spend, and 63% on Thanksgiving. It also expects traffic to U.S. retail sites from AI chat and browser services to rise 130% year over year. Yottaa’s own Cyber 5 research found that while shoppers spent $44.2 billion online over Cyber 5 2025, shopping sessions fell every day of the weekend, down as much as 16.6% year over year.
Put those together and the picture for a mid-market brand is specific: fewer casual browsers, more comparison-shopping, more of it on a phone or through an AI assistant, and no macro tailwind to hide behind.
What this does to the Q4 playbook
Most peak-season plans are built on a traffic assumption. Spend more, land more sessions, convert at roughly last year’s rate, hit the number. That plan has two problems this year.
The first is that paid traffic is not getting cheaper, and a cautious consumer is going to convert at a lower-than-expected rate per session, so the same acquisition budget buys less revenue than the model assumes. The second is that the traffic side of the equation is increasingly out of your hands. AI-mediated discovery is reshaping how shoppers arrive, and the retailers on the winning side of that shift are not necessarily the ones with the biggest media budgets.
This is where performance stops being an engineering concern and becomes a P&L concern. A shopper who is already hesitant about a discretionary purchase does not need much of an excuse to abandon. A product page that takes five seconds to become interactive, a checkout that stutters because a chat widget is blocking the main thread, a mobile experience that shifts under a thumb mid-tap: in a confident market, some of those shoppers push through. In a cautious one, fewer do. Shopify’s own analysis puts the cost at 3.5% of conversion rate for every 100 milliseconds of load time (April 2026).
And keep in mind: AI agents are even less tolerant of slow sites than human shoppers. So, as agentic shopping becomes more prevalent, it’s critical that your site is keeping up with their lofty expectations.
The uncomfortable part: most teams cannot see where it is leaking
Here is our POV, and it is the part that tends to land badly in planning meetings. Deciding to compete on conversion is straightforward. Knowing where conversion is actually breaking is harder, because the data most teams use to answer that question is incomplete in a specific, non-random way.
Ad blockers, browser privacy tools, and script failures create a 15–30% visibility gap in standard real user monitoring implementations. That gap skews toward the sessions where something went wrong: the blocked scripts, the failed loads, the broken experiences. The dashboard shows you the sessions healthy enough to report on themselves, but omits the sessions where more problems are occurring.
Add two more distortions: If your performance signal is CrUX, you are seeing Chrome on a 28-day rolling window, and across Yottaa-monitored sites roughly 38% of sessions come from Safari. And a growing share of what hits a commerce storefront is not a shopper at all. Cloudflare reported in September that more than half the traffic reaching sites on its network is now automated. Fold that into your averages and both your Core Web Vitals and your conversion rate describe a population that is partly (mostly?) machines.
So the team commits to a conversion-led quarter, opens the performance dashboard, and makes decisions from a partial, Chrome-weighted, bot-diluted, month-old picture. Then ot wonders why the optimization did not move the number.
What to do in the weeks before Cyber 5
- Audit coverage before you audit performance. Find out what share of your sessions your tooling actually reports on, and which browsers, devices, and traffic types are missing. It needs no code change, so it fits inside a freeze, and every other decision depends on it.
- Separate humans from machines before calculating a single metric. If you cannot filter automated traffic out of your performance and conversion data, you do not know your baseline, and that gap widens as AI-driven visits climb through the season.
- Attribute third-party impact per vendor, not in aggregate. Peak season is when tag sprawl gets expensive. ‘Our scripts are slow’ is not something anyone can act on. ‘This specific vendor adds 900ms to the product page’ is.
- Instrument the conversion path specifically. Category, product, cart, checkout. Sitewide averages hide the pages where money is actually lost.
- Freeze what you cannot measure. A tag that is not live and measured by now should wait until January. A new script added in November without per-vendor visibility puts risk into your highest-revenue weeks with no way to diagnose it.
- Make sure someone is watching in the evening. In Yottaa’s Cyber 5 2025 data, AI-referred shopping traffic spiked every night between about 2 and 4 a.m. UTC, which is evening in the U.S. Point alerts at whoever is actually on shift then.
The season rewards operators
A cautious consumer should change where the plan gets its growth. When demand is soft and traffic is expensive, the brands that win are the ones that convert the shoppers they already have, and that comes down to execution.
The catch is that you cannot optimize what you cannot see. With Cyber 5 a few weeks out, the highest-return work is making sure the data you will run peak season on describes your actual shoppers.
Where is your team placing its bet this season: more traffic, or better conversion on the traffic you already have?
Subscribe to our blog
Performant Experimentation – A Guide to A/B Testing
Every A/B test you run is a purchase. You’re buying information—an answer to a hypothesis—and you’re paying for it in degraded performance and lost conversion during the test window.
Download eBook
