.jpg)
of the United Kingdom’s capitol city.
A stale product page, a sluggish site, and an over-stuffed warehouse may seem like small problems, and many teams treat them as such.
They don’t look like major breaks on the surface. So they get ignored in lieu of other tasks and priorities.
But during the holiday season, these “small issues” can lead to major ramifications. The revenue stakes are higher, and everyone is paying closer attention to every flaw and delay. They also expose a major flaw in retail operations, especially in managing infrastructure.
That’s because most retailers follow a two-phased calendar that I’ll call “the quiet” and “the spike.” The quiet phase is the long runway, where teams build their budgets, forecasts, and campaigns. During the Q4 spike, they use these pre-established components to survive and ride out the end of the year.
The split makes sense for many categories because demand typically thins out in the spring and swells starting in November. But infrastructure never has a quiet season, and that’s why many retailers fall apart.
The Permafrost Problem
Permafrost serves as the perfect metaphor for retail infrastructure. When you look at the surface, there’s no clear evidence that the land is flawed. It looks solid, but the temperature underneath is slowly creeping up a fraction of a degree at a time. Over decades, the land deteriorates until one random summer, when the road buckles.
In reality, this “sudden catastrophe” was years in the making.
Retail infrastructure thaws the same way. Imagine a retailer’s product sells out on a random Tuesday in April. The PDP insists it’s in stock for hours because the page was cached for speed and nobody built a way to sense the change. A new shipment of that same product comes in the following week, but the catalog doesn’t recognize that change. The underlying issue is that the system tracking the warehouse and the system that talks to customers were never properly integrated.
One audit of a mid-sized retailer found that category pages took close to six seconds to load, with stock data trailing the warehouse by three hours on average—and by a full day at the worst. On a random Tuesday, the loss is hard to measure. But spread across the busiest week of the year, the impact is felt. For this one brand, the channel did a little over €1 million a year, so that gap equated to €250,000 in lost sales.
More Isn’t Just More
The same erosion sets in at the level of raw capacity because a retailer doesn’t lose speed in proportion to how many people show up—the loss compounds.
A thousand shoppers browsing and comparing products produce database requests that are essentially elbowing each other for the same memory and CPU power, the same locks, and the same rows in the same tables. But a system can look laughably small by the third week of November because nobody is looking at the same handful of pages. They’re sorting and filtering through thousands of unique combinations, and every one of them needs a fresh answer from an overwhelmed database.
Procurement sees a similar failure, but in slow motion. If a team can’t fully trust real-time inventory numbers, the “safe bet” is to over-order. They pad the warehouse and hedge against a stock-out, especially ahead of peak season. But this tactic can cost the business in the long run. One retailer was sitting on more than €2 million of dead stock because that inventory was never insurance against real demand. It was the retailer’s way of disguising what the team didn’t know. Like the stale pages, it’s the symptom, but not the problem itself.
A Season With No Calendar
None of these issues are due to Black Friday itself. Black Friday, and the broader holiday season, is just turning the volume up on decisions made, and not made, back in March. Everyone remembers the week the site went down, but almost nobody remembers the ordinary Tuesday in spring when the same crack first opened.
The issue is only intensifying because agents are driving a growing share of retail traffic, comparing prices, checking stock levels, and even buying on a human’s behalf. In June, 57.5% of web traffic measured by Cloudflare came from automated requests. It was a threshold that CEO Matthew Prince predicted—it just came 18 months early.
And an agent visit can put incredible pressure on an eCommerce site if retailers aren’t ready. In one study, OpenAI crawled a retail site 198 times for every visit it sent back to a user, a staggering difference from the six crawls from Google. Retailers haven’t planned for the incredible volumes at which their storefronts are being read and judged. And these agents don’t crawl based on a retail calendar. They show up whenever someone wants an answer, whether it’s on a random Tuesday in April, or on Thanksgiving Day.
Agent traffic feels like an attack when an infrastructure is built for the traditional retail calendar. The first instinct is to tighten the firewall, add another layer of bot management, throttle the crawlers, and watch the server bill shrink. It feels like good housekeeping, but it’s actually locking the door on your next best customer.
Those crawlers are how ChatGPT, Perplexity, and Gemini know a store’s prices exist in the first place, and how shopping agents learn what’s for sale and whether your products are worth suggesting. Your customers (both current and potential) use these platforms and, in turn, these crawlers, as their own personal consultants and concierges. Block the crawler, and a retailer disappears from the answer, right as a buying decision is forming.
There’s no quiet season left to wait this out in. This traffic is, at best, the smallest it will ever be—and McKinsey puts the agent-driven share of global retail at $3–$5 trillion by 2030. For comparison, the entire global online retail market today is worth somewhere around $6.9 trillion.
This kind of traffic hits broken systems harder. A human may get annoyed if a sold-out item is marketed as “in stock.” At worst, they’ll complain and eventually come back because finding an alternative product and retailer takes effort. But when an agent hits the same roadblock, it simply crosses the retailer off the list, removing the retailer completely from this new discovery layer. Previously, the feedback loop and effects took the whole season. Now, all it takes is a single request.
The Next Quiet Season Is Already Loading
Agents comparing prices on a human’s behalf are, in some categories, already starting to negotiate with each other. A procurement agent for one business is checking terms against a sales agent for another, and no human is in the loop to monitor the exchange.
Storefronts are being interrogated by something that never sleeps. Retailers’ data must be able to survive being cross-examined by another piece of software at three in the morning, in a completely different currency and language. Otherwise, they’ll be falling victim to the same fallacy as the retailer lacking inventory visibility.
Which brings us back to where we started. The retailers who come out ahead over the next few years will be the ones who stopped writing off quiet revenue loss as background noise and started treating it as the actual problem. If you’re just building for a season you can picture, you can guarantee an unexpected blind spot—and that’s true whether that visitor is a human or a bot.
Gala Pustova is the Co-founder and CEO of Catomize, a Swiss solution for online retailers. She is a technology entrepreneur and experienced C-level executive with a strong track record across eCommerce technology, GovTech, and digital transformation. Gala previously held senior leadership roles in the Government of Ukraine and has been recognized among GovInsider’s Women in GovTech 2025. She holds an Executive MBA from the University of Oxford and serves in non-executive and advisory roles, helping businesses to digitalize and prepare for the agentic era.
A stale product page, a sluggish site, and an over-stuffed warehouse may seem like small problems, and many teams treat them as such.
They don’t look like major breaks on the surface. So they get ignored in lieu of other tasks and priorities.
But during the holiday season, these “small issues” can lead to major ramifications. The revenue stakes are higher, and everyone is paying closer attention to every flaw and delay. They also expose a major flaw in retail operations, especially in managing infrastructure.
That’s because most retailers follow a two-phased calendar that I’ll call “the quiet” and “the spike.” The quiet phase is the long runway, where teams build their budgets, forecasts, and campaigns. During the Q4 spike, they use these pre-established components to survive and ride out the end of the year.
The split makes sense for many categories because demand typically thins out in the spring and swells starting in November. But infrastructure never has a quiet season, and that’s why many retailers fall apart.
The Permafrost Problem
Permafrost serves as the perfect metaphor for retail infrastructure. When you look at the surface, there’s no clear evidence that the land is flawed. It looks solid, but the temperature underneath is slowly creeping up a fraction of a degree at a time. Over decades, the land deteriorates until one random summer, when the road buckles.
In reality, this “sudden catastrophe” was years in the making.
Retail infrastructure thaws the same way. Imagine a retailer’s product sells out on a random Tuesday in April. The PDP insists it’s in stock for hours because the page was cached for speed and nobody built a way to sense the change. A new shipment of that same product comes in the following week, but the catalog doesn’t recognize that change. The underlying issue is that the system tracking the warehouse and the system that talks to customers were never properly integrated.
One audit of a mid-sized retailer found that category pages took close to six seconds to load, with stock data trailing the warehouse by three hours on average—and by a full day at the worst. On a random Tuesday, the loss is hard to measure. But spread across the busiest week of the year, the impact is felt. For this one brand, the channel did a little over €1 million a year, so that gap equated to €250,000 in lost sales.
More Isn’t Just More
The same erosion sets in at the level of raw capacity because a retailer doesn’t lose speed in proportion to how many people show up—the loss compounds.
A thousand shoppers browsing and comparing products produce database requests that are essentially elbowing each other for the same memory and CPU power, the same locks, and the same rows in the same tables. But a system can look laughably small by the third week of November because nobody is looking at the same handful of pages. They’re sorting and filtering through thousands of unique combinations, and every one of them needs a fresh answer from an overwhelmed database.
Procurement sees a similar failure, but in slow motion. If a team can’t fully trust real-time inventory numbers, the “safe bet” is to over-order. They pad the warehouse and hedge against a stock-out, especially ahead of peak season. But this tactic can cost the business in the long run. One retailer was sitting on more than €2 million of dead stock because that inventory was never insurance against real demand. It was the retailer’s way of disguising what the team didn’t know. Like the stale pages, it’s the symptom, but not the problem itself.
A Season With No Calendar
None of these issues are due to Black Friday itself. Black Friday, and the broader holiday season, is just turning the volume up on decisions made, and not made, back in March. Everyone remembers the week the site went down, but almost nobody remembers the ordinary Tuesday in spring when the same crack first opened.
The issue is only intensifying because agents are driving a growing share of retail traffic, comparing prices, checking stock levels, and even buying on a human’s behalf. In June, 57.5% of web traffic measured by Cloudflare came from automated requests. It was a threshold that CEO Matthew Prince predicted—it just came 18 months early.
And an agent visit can put incredible pressure on an eCommerce site if retailers aren’t ready. In one study, OpenAI crawled a retail site 198 times for every visit it sent back to a user, a staggering difference from the six crawls from Google. Retailers haven’t planned for the incredible volumes at which their storefronts are being read and judged. And these agents don’t crawl based on a retail calendar. They show up whenever someone wants an answer, whether it’s on a random Tuesday in April, or on Thanksgiving Day.
Agent traffic feels like an attack when an infrastructure is built for the traditional retail calendar. The first instinct is to tighten the firewall, add another layer of bot management, throttle the crawlers, and watch the server bill shrink. It feels like good housekeeping, but it’s actually locking the door on your next best customer.
Those crawlers are how ChatGPT, Perplexity, and Gemini know a store’s prices exist in the first place, and how shopping agents learn what’s for sale and whether your products are worth suggesting. Your customers (both current and potential) use these platforms and, in turn, these crawlers, as their own personal consultants and concierges. Block the crawler, and a retailer disappears from the answer, right as a buying decision is forming.
There’s no quiet season left to wait this out in. This traffic is, at best, the smallest it will ever be—and McKinsey puts the agent-driven share of global retail at $3–$5 trillion by 2030. For comparison, the entire global online retail market today is worth somewhere around $6.9 trillion.
This kind of traffic hits broken systems harder. A human may get annoyed if a sold-out item is marketed as “in stock.” At worst, they’ll complain and eventually come back because finding an alternative product and retailer takes effort. But when an agent hits the same roadblock, it simply crosses the retailer off the list, removing the retailer completely from this new discovery layer. Previously, the feedback loop and effects took the whole season. Now, all it takes is a single request.
The Next Quiet Season Is Already Loading
Agents comparing prices on a human’s behalf are, in some categories, already starting to negotiate with each other. A procurement agent for one business is checking terms against a sales agent for another, and no human is in the loop to monitor the exchange.
Storefronts are being interrogated by something that never sleeps. Retailers’ data must be able to survive being cross-examined by another piece of software at three in the morning, in a completely different currency and language. Otherwise, they’ll be falling victim to the same fallacy as the retailer lacking inventory visibility.
Which brings us back to where we started. The retailers who come out ahead over the next few years will be the ones who stopped writing off quiet revenue loss as background noise and started treating it as the actual problem. If you’re just building for a season you can picture, you can guarantee an unexpected blind spot—and that’s true whether that visitor is a human or a bot.
Gala Pustova is the Co-founder and CEO of Catomize, a Swiss solution for online retailers. She is a technology entrepreneur and experienced C-level executive with a strong track record across eCommerce technology, GovTech, and digital transformation. Gala previously held senior leadership roles in the Government of Ukraine and has been recognized among GovInsider’s Women in GovTech 2025. She holds an Executive MBA from the University of Oxford and serves in non-executive and advisory roles, helping businesses to digitalize and prepare for the agentic era.
A stale product page, a sluggish site, and an over-stuffed warehouse may seem like small problems, and many teams treat them as such.
They don’t look like major breaks on the surface. So they get ignored in lieu of other tasks and priorities.
But during the holiday season, these “small issues” can lead to major ramifications. The revenue stakes are higher, and everyone is paying closer attention to every flaw and delay. They also expose a major flaw in retail operations, especially in managing infrastructure.
That’s because most retailers follow a two-phased calendar that I’ll call “the quiet” and “the spike.” The quiet phase is the long runway, where teams build their budgets, forecasts, and campaigns. During the Q4 spike, they use these pre-established components to survive and ride out the end of the year.
The split makes sense for many categories because demand typically thins out in the spring and swells starting in November. But infrastructure never has a quiet season, and that’s why many retailers fall apart.
The Permafrost Problem
Permafrost serves as the perfect metaphor for retail infrastructure. When you look at the surface, there’s no clear evidence that the land is flawed. It looks solid, but the temperature underneath is slowly creeping up a fraction of a degree at a time. Over decades, the land deteriorates until one random summer, when the road buckles.
In reality, this “sudden catastrophe” was years in the making.
Retail infrastructure thaws the same way. Imagine a retailer’s product sells out on a random Tuesday in April. The PDP insists it’s in stock for hours because the page was cached for speed and nobody built a way to sense the change. A new shipment of that same product comes in the following week, but the catalog doesn’t recognize that change. The underlying issue is that the system tracking the warehouse and the system that talks to customers were never properly integrated.
One audit of a mid-sized retailer found that category pages took close to six seconds to load, with stock data trailing the warehouse by three hours on average—and by a full day at the worst. On a random Tuesday, the loss is hard to measure. But spread across the busiest week of the year, the impact is felt. For this one brand, the channel did a little over €1 million a year, so that gap equated to €250,000 in lost sales.
More Isn’t Just More
The same erosion sets in at the level of raw capacity because a retailer doesn’t lose speed in proportion to how many people show up—the loss compounds.
A thousand shoppers browsing and comparing products produce database requests that are essentially elbowing each other for the same memory and CPU power, the same locks, and the same rows in the same tables. But a system can look laughably small by the third week of November because nobody is looking at the same handful of pages. They’re sorting and filtering through thousands of unique combinations, and every one of them needs a fresh answer from an overwhelmed database.
Procurement sees a similar failure, but in slow motion. If a team can’t fully trust real-time inventory numbers, the “safe bet” is to over-order. They pad the warehouse and hedge against a stock-out, especially ahead of peak season. But this tactic can cost the business in the long run. One retailer was sitting on more than €2 million of dead stock because that inventory was never insurance against real demand. It was the retailer’s way of disguising what the team didn’t know. Like the stale pages, it’s the symptom, but not the problem itself.
A Season With No Calendar
None of these issues are due to Black Friday itself. Black Friday, and the broader holiday season, is just turning the volume up on decisions made, and not made, back in March. Everyone remembers the week the site went down, but almost nobody remembers the ordinary Tuesday in spring when the same crack first opened.
The issue is only intensifying because agents are driving a growing share of retail traffic, comparing prices, checking stock levels, and even buying on a human’s behalf. In June, 57.5% of web traffic measured by Cloudflare came from automated requests. It was a threshold that CEO Matthew Prince predicted—it just came 18 months early.
And an agent visit can put incredible pressure on an eCommerce site if retailers aren’t ready. In one study, OpenAI crawled a retail site 198 times for every visit it sent back to a user, a staggering difference from the six crawls from Google. Retailers haven’t planned for the incredible volumes at which their storefronts are being read and judged. And these agents don’t crawl based on a retail calendar. They show up whenever someone wants an answer, whether it’s on a random Tuesday in April, or on Thanksgiving Day.
Agent traffic feels like an attack when an infrastructure is built for the traditional retail calendar. The first instinct is to tighten the firewall, add another layer of bot management, throttle the crawlers, and watch the server bill shrink. It feels like good housekeeping, but it’s actually locking the door on your next best customer.
Those crawlers are how ChatGPT, Perplexity, and Gemini know a store’s prices exist in the first place, and how shopping agents learn what’s for sale and whether your products are worth suggesting. Your customers (both current and potential) use these platforms and, in turn, these crawlers, as their own personal consultants and concierges. Block the crawler, and a retailer disappears from the answer, right as a buying decision is forming.
There’s no quiet season left to wait this out in. This traffic is, at best, the smallest it will ever be—and McKinsey puts the agent-driven share of global retail at $3–$5 trillion by 2030. For comparison, the entire global online retail market today is worth somewhere around $6.9 trillion.
This kind of traffic hits broken systems harder. A human may get annoyed if a sold-out item is marketed as “in stock.” At worst, they’ll complain and eventually come back because finding an alternative product and retailer takes effort. But when an agent hits the same roadblock, it simply crosses the retailer off the list, removing the retailer completely from this new discovery layer. Previously, the feedback loop and effects took the whole season. Now, all it takes is a single request.
The Next Quiet Season Is Already Loading
Agents comparing prices on a human’s behalf are, in some categories, already starting to negotiate with each other. A procurement agent for one business is checking terms against a sales agent for another, and no human is in the loop to monitor the exchange.
Storefronts are being interrogated by something that never sleeps. Retailers’ data must be able to survive being cross-examined by another piece of software at three in the morning, in a completely different currency and language. Otherwise, they’ll be falling victim to the same fallacy as the retailer lacking inventory visibility.
Which brings us back to where we started. The retailers who come out ahead over the next few years will be the ones who stopped writing off quiet revenue loss as background noise and started treating it as the actual problem. If you’re just building for a season you can picture, you can guarantee an unexpected blind spot—and that’s true whether that visitor is a human or a bot.
Gala Pustova is the Co-founder and CEO of Catomize, a Swiss solution for online retailers. She is a technology entrepreneur and experienced C-level executive with a strong track record across eCommerce technology, GovTech, and digital transformation. Gala previously held senior leadership roles in the Government of Ukraine and has been recognized among GovInsider’s Women in GovTech 2025. She holds an Executive MBA from the University of Oxford and serves in non-executive and advisory roles, helping businesses to digitalize and prepare for the agentic era.
A stale product page, a sluggish site, and an over-stuffed warehouse may seem like small problems, and many teams treat them as such.
They don’t look like major breaks on the surface. So they get ignored in lieu of other tasks and priorities.
But during the holiday season, these “small issues” can lead to major ramifications. The revenue stakes are higher, and everyone is paying closer attention to every flaw and delay. They also expose a major flaw in retail operations, especially in managing infrastructure.
That’s because most retailers follow a two-phased calendar that I’ll call “the quiet” and “the spike.” The quiet phase is the long runway, where teams build their budgets, forecasts, and campaigns. During the Q4 spike, they use these pre-established components to survive and ride out the end of the year.
The split makes sense for many categories because demand typically thins out in the spring and swells starting in November. But infrastructure never has a quiet season, and that’s why many retailers fall apart.
The Permafrost Problem
Permafrost serves as the perfect metaphor for retail infrastructure. When you look at the surface, there’s no clear evidence that the land is flawed. It looks solid, but the temperature underneath is slowly creeping up a fraction of a degree at a time. Over decades, the land deteriorates until one random summer, when the road buckles.
In reality, this “sudden catastrophe” was years in the making.
Retail infrastructure thaws the same way. Imagine a retailer’s product sells out on a random Tuesday in April. The PDP insists it’s in stock for hours because the page was cached for speed and nobody built a way to sense the change. A new shipment of that same product comes in the following week, but the catalog doesn’t recognize that change. The underlying issue is that the system tracking the warehouse and the system that talks to customers were never properly integrated.
One audit of a mid-sized retailer found that category pages took close to six seconds to load, with stock data trailing the warehouse by three hours on average—and by a full day at the worst. On a random Tuesday, the loss is hard to measure. But spread across the busiest week of the year, the impact is felt. For this one brand, the channel did a little over €1 million a year, so that gap equated to €250,000 in lost sales.
More Isn’t Just More
The same erosion sets in at the level of raw capacity because a retailer doesn’t lose speed in proportion to how many people show up—the loss compounds.
A thousand shoppers browsing and comparing products produce database requests that are essentially elbowing each other for the same memory and CPU power, the same locks, and the same rows in the same tables. But a system can look laughably small by the third week of November because nobody is looking at the same handful of pages. They’re sorting and filtering through thousands of unique combinations, and every one of them needs a fresh answer from an overwhelmed database.
Procurement sees a similar failure, but in slow motion. If a team can’t fully trust real-time inventory numbers, the “safe bet” is to over-order. They pad the warehouse and hedge against a stock-out, especially ahead of peak season. But this tactic can cost the business in the long run. One retailer was sitting on more than €2 million of dead stock because that inventory was never insurance against real demand. It was the retailer’s way of disguising what the team didn’t know. Like the stale pages, it’s the symptom, but not the problem itself.
A Season With No Calendar
None of these issues are due to Black Friday itself. Black Friday, and the broader holiday season, is just turning the volume up on decisions made, and not made, back in March. Everyone remembers the week the site went down, but almost nobody remembers the ordinary Tuesday in spring when the same crack first opened.
The issue is only intensifying because agents are driving a growing share of retail traffic, comparing prices, checking stock levels, and even buying on a human’s behalf. In June, 57.5% of web traffic measured by Cloudflare came from automated requests. It was a threshold that CEO Matthew Prince predicted—it just came 18 months early.
And an agent visit can put incredible pressure on an eCommerce site if retailers aren’t ready. In one study, OpenAI crawled a retail site 198 times for every visit it sent back to a user, a staggering difference from the six crawls from Google. Retailers haven’t planned for the incredible volumes at which their storefronts are being read and judged. And these agents don’t crawl based on a retail calendar. They show up whenever someone wants an answer, whether it’s on a random Tuesday in April, or on Thanksgiving Day.
Agent traffic feels like an attack when an infrastructure is built for the traditional retail calendar. The first instinct is to tighten the firewall, add another layer of bot management, throttle the crawlers, and watch the server bill shrink. It feels like good housekeeping, but it’s actually locking the door on your next best customer.
Those crawlers are how ChatGPT, Perplexity, and Gemini know a store’s prices exist in the first place, and how shopping agents learn what’s for sale and whether your products are worth suggesting. Your customers (both current and potential) use these platforms and, in turn, these crawlers, as their own personal consultants and concierges. Block the crawler, and a retailer disappears from the answer, right as a buying decision is forming.
There’s no quiet season left to wait this out in. This traffic is, at best, the smallest it will ever be—and McKinsey puts the agent-driven share of global retail at $3–$5 trillion by 2030. For comparison, the entire global online retail market today is worth somewhere around $6.9 trillion.
This kind of traffic hits broken systems harder. A human may get annoyed if a sold-out item is marketed as “in stock.” At worst, they’ll complain and eventually come back because finding an alternative product and retailer takes effort. But when an agent hits the same roadblock, it simply crosses the retailer off the list, removing the retailer completely from this new discovery layer. Previously, the feedback loop and effects took the whole season. Now, all it takes is a single request.
The Next Quiet Season Is Already Loading
Agents comparing prices on a human’s behalf are, in some categories, already starting to negotiate with each other. A procurement agent for one business is checking terms against a sales agent for another, and no human is in the loop to monitor the exchange.
Storefronts are being interrogated by something that never sleeps. Retailers’ data must be able to survive being cross-examined by another piece of software at three in the morning, in a completely different currency and language. Otherwise, they’ll be falling victim to the same fallacy as the retailer lacking inventory visibility.
Which brings us back to where we started. The retailers who come out ahead over the next few years will be the ones who stopped writing off quiet revenue loss as background noise and started treating it as the actual problem. If you’re just building for a season you can picture, you can guarantee an unexpected blind spot—and that’s true whether that visitor is a human or a bot.
Gala Pustova is the Co-founder and CEO of Catomize, a Swiss solution for online retailers. She is a technology entrepreneur and experienced C-level executive with a strong track record across eCommerce technology, GovTech, and digital transformation. Gala previously held senior leadership roles in the Government of Ukraine and has been recognized among GovInsider’s Women in GovTech 2025. She holds an Executive MBA from the University of Oxford and serves in non-executive and advisory roles, helping businesses to digitalize and prepare for the agentic era.
A stale product page, a sluggish site, and an over-stuffed warehouse may seem like small problems, and many teams treat them as such.
They don’t look like major breaks on the surface. So they get ignored in lieu of other tasks and priorities.
But during the holiday season, these “small issues” can lead to major ramifications. The revenue stakes are higher, and everyone is paying closer attention to every flaw and delay. They also expose a major flaw in retail operations, especially in managing infrastructure.
That’s because most retailers follow a two-phased calendar that I’ll call “the quiet” and “the spike.” The quiet phase is the long runway, where teams build their budgets, forecasts, and campaigns. During the Q4 spike, they use these pre-established components to survive and ride out the end of the year.
The split makes sense for many categories because demand typically thins out in the spring and swells starting in November. But infrastructure never has a quiet season, and that’s why many retailers fall apart.
The Permafrost Problem
Permafrost serves as the perfect metaphor for retail infrastructure. When you look at the surface, there’s no clear evidence that the land is flawed. It looks solid, but the temperature underneath is slowly creeping up a fraction of a degree at a time. Over decades, the land deteriorates until one random summer, when the road buckles.
In reality, this “sudden catastrophe” was years in the making.
Retail infrastructure thaws the same way. Imagine a retailer’s product sells out on a random Tuesday in April. The PDP insists it’s in stock for hours because the page was cached for speed and nobody built a way to sense the change. A new shipment of that same product comes in the following week, but the catalog doesn’t recognize that change. The underlying issue is that the system tracking the warehouse and the system that talks to customers were never properly integrated.
One audit of a mid-sized retailer found that category pages took close to six seconds to load, with stock data trailing the warehouse by three hours on average—and by a full day at the worst. On a random Tuesday, the loss is hard to measure. But spread across the busiest week of the year, the impact is felt. For this one brand, the channel did a little over €1 million a year, so that gap equated to €250,000 in lost sales.
More Isn’t Just More
The same erosion sets in at the level of raw capacity because a retailer doesn’t lose speed in proportion to how many people show up—the loss compounds.
A thousand shoppers browsing and comparing products produce database requests that are essentially elbowing each other for the same memory and CPU power, the same locks, and the same rows in the same tables. But a system can look laughably small by the third week of November because nobody is looking at the same handful of pages. They’re sorting and filtering through thousands of unique combinations, and every one of them needs a fresh answer from an overwhelmed database.
Procurement sees a similar failure, but in slow motion. If a team can’t fully trust real-time inventory numbers, the “safe bet” is to over-order. They pad the warehouse and hedge against a stock-out, especially ahead of peak season. But this tactic can cost the business in the long run. One retailer was sitting on more than €2 million of dead stock because that inventory was never insurance against real demand. It was the retailer’s way of disguising what the team didn’t know. Like the stale pages, it’s the symptom, but not the problem itself.
A Season With No Calendar
None of these issues are due to Black Friday itself. Black Friday, and the broader holiday season, is just turning the volume up on decisions made, and not made, back in March. Everyone remembers the week the site went down, but almost nobody remembers the ordinary Tuesday in spring when the same crack first opened.
The issue is only intensifying because agents are driving a growing share of retail traffic, comparing prices, checking stock levels, and even buying on a human’s behalf. In June, 57.5% of web traffic measured by Cloudflare came from automated requests. It was a threshold that CEO Matthew Prince predicted—it just came 18 months early.
And an agent visit can put incredible pressure on an eCommerce site if retailers aren’t ready. In one study, OpenAI crawled a retail site 198 times for every visit it sent back to a user, a staggering difference from the six crawls from Google. Retailers haven’t planned for the incredible volumes at which their storefronts are being read and judged. And these agents don’t crawl based on a retail calendar. They show up whenever someone wants an answer, whether it’s on a random Tuesday in April, or on Thanksgiving Day.
Agent traffic feels like an attack when an infrastructure is built for the traditional retail calendar. The first instinct is to tighten the firewall, add another layer of bot management, throttle the crawlers, and watch the server bill shrink. It feels like good housekeeping, but it’s actually locking the door on your next best customer.
Those crawlers are how ChatGPT, Perplexity, and Gemini know a store’s prices exist in the first place, and how shopping agents learn what’s for sale and whether your products are worth suggesting. Your customers (both current and potential) use these platforms and, in turn, these crawlers, as their own personal consultants and concierges. Block the crawler, and a retailer disappears from the answer, right as a buying decision is forming.
There’s no quiet season left to wait this out in. This traffic is, at best, the smallest it will ever be—and McKinsey puts the agent-driven share of global retail at $3–$5 trillion by 2030. For comparison, the entire global online retail market today is worth somewhere around $6.9 trillion.
This kind of traffic hits broken systems harder. A human may get annoyed if a sold-out item is marketed as “in stock.” At worst, they’ll complain and eventually come back because finding an alternative product and retailer takes effort. But when an agent hits the same roadblock, it simply crosses the retailer off the list, removing the retailer completely from this new discovery layer. Previously, the feedback loop and effects took the whole season. Now, all it takes is a single request.
The Next Quiet Season Is Already Loading
Agents comparing prices on a human’s behalf are, in some categories, already starting to negotiate with each other. A procurement agent for one business is checking terms against a sales agent for another, and no human is in the loop to monitor the exchange.
Storefronts are being interrogated by something that never sleeps. Retailers’ data must be able to survive being cross-examined by another piece of software at three in the morning, in a completely different currency and language. Otherwise, they’ll be falling victim to the same fallacy as the retailer lacking inventory visibility.
Which brings us back to where we started. The retailers who come out ahead over the next few years will be the ones who stopped writing off quiet revenue loss as background noise and started treating it as the actual problem. If you’re just building for a season you can picture, you can guarantee an unexpected blind spot—and that’s true whether that visitor is a human or a bot.
Gala Pustova is the Co-founder and CEO of Catomize, a Swiss solution for online retailers. She is a technology entrepreneur and experienced C-level executive with a strong track record across eCommerce technology, GovTech, and digital transformation. Gala previously held senior leadership roles in the Government of Ukraine and has been recognized among GovInsider’s Women in GovTech 2025. She holds an Executive MBA from the University of Oxford and serves in non-executive and advisory roles, helping businesses to digitalize and prepare for the agentic era.
Continue Reading...
Insights and futurism for executives in eCom and Retail
Those things we shouldn’t say out loud? We say them on the private feed. Bi-weekly “after dark” podcasts and a members-only newsletter, just for subscribers.
Our research reports combine visionary thinking with data-backed findings from our own advisory panel, made up of leaders at brands you know and trust.
Upskill, cross-skill, and future-proof your teams with Future Commerce Learning, the leading digital eCommerce learning platform, created by professional educators.
