6 searches a day.

For a business that welcomed 170 million visitors across its destinations in 2024, that's the search volume Hammerson's website was generating. Not per hour. Per day. 

Hammerson is an FTSE 250 real estate company that owns and operates some of the UK's most prominent retail destinations, including Bullring in Birmingham, Brent Cross in North London, Cabot Circus in Bristol, and Westquay in Southampton. Across their portfolio, occupancy sits above 95%, retailers consistently report their Hammerson locations ranking among their top-performing stores, and the brand attracted over 200 premium brand partnerships and events in 2024 alone. 

In short, this is not a business with a demand problem.

So 6 daily searches wasn't a signal that nobody was interested in what was on the sites. It was a signal that the experience was making it too hard to look. Visitors were arriving, running a search, getting nothing useful back, and either navigating away or giving up entirely. The intent was there the whole time. The search bar just wasn't equipped to catch it.

That number is now 1,000 searches per day. 

This post covers what changed, why it matters, and what it might mean for your own site if any of this sounds familiar.

Why Keyword Search Fails Retail Destinations

The instinct when site metrics underperform is to look at traffic. Not enough visitors, not enough conversions. But Hammerson's sites weren't short of people. The issue was what happened once those people arrived.

Their search was built on keyword matching. Type in the exact name of a shop, get a result. That works fine if every visitor already knows what they're looking for and remembers the specific brand name. In practice, that's not how most people shop. Today, that is one of the top reasons people leave a website: getting the “No results found” after their search entry.

The data backs this up at scale. According to Salesforce, 76% of customers expect companies to understand their needs and expectations, not just respond to what they literally type in the search bar. And Forrester found that around 43% of site visitors go straight to the search bar when they land on a website, making it one of the most used features on any retail site. When that experience fails them, most don't try again with different wording. They leave for a competitor’s website.

That's the distinction worth holding onto: low search volume on a busy site usually means suppressed intent, not absent demand. The people are there, they are just not getting what they want. That was the case with Hammerson.

The Gap Between What Shoppers Type and What They Mean

Traditional site search treats every query as a lookup, not a request. It's looking for an exact string match in a database, not trying to understand what the person actually wants.

For a shopping destination with hundreds of retailers across food, fashion, leisure, and entertainment, that creates a lot of dead ends. A visitor looking for "something for the kids" or "casual lunch" shouldn't need to already know which specific retailer to visit. The search should figure that out for them. Primarily, because the visitor today has changed their behaviour, and, know, this is what they expect. They no longer want to search for everything manually, but want what they are looking for to find them - with as little friction as possible.

This is a widespread problem. In Coveo's recent Commerce Relevance Report, they found that 72% of shoppers will abandon a site if they can’t find what they need quickly. Of those who leave, 36% head straight to a competitor's site (while 53% bounce back to Google).

And internal data from e-commerce platforms consistently shows zero-results rates of 10-30% on sites using basic keyword matching, meaning roughly one in five searches is returning nothing useful. 

Hammerson Solution

The rebuild for Hammerson focused on closing this gap. 

The new search layer introduced AI tagging across the retailer's catalogue, which meant the system could now recognise category and intent terms rather than just brand names. Someone searching for "birthday dinner" would surface relevant restaurant options. Someone looking for "sports gear" would find the right retailers even if they didn't know any of them by name.

Smart suggestions were built in for searches that didn't return an exact match. Instead of a blank results page, shoppers would see related options, nearby alternatives, or relevant offers, effectively avoiding the “no results” message. 

Image 1: Bullring Search Bar, Source: Bullring.co.uk

Automated pipelines pulled retailer content directly from each site's CMS, and retailer profiles were enriched with live information, including opening hours from Google Places.

Accurate, current data is the kind of detail that determines whether people trust what they're looking at.

When Search Volume Becomes a Commercial Signal

After the new search experience went live, daily searches went from 6 to 1,000. That's a 9x increase, and it held.

The headline number is strong, but the more interesting shift was in how people used search. Before, the people who did search were mostly navigating to retailers they already knew. After, shoppers started using search to discover retailers they hadn't visited before. The search went from being a navigation tool to a discovery tool, which is a much more commercially valuable thing for a retail destination to have.

Image 2: Brent Cross Search Bar, Source: brentcross.co.uk

That shift has real revenue implications. 

Visitors who use site search convert at a significantly higher rate than those who don't, with some studies putting it as high as 2-3x the conversion rate of non-search visitors. 

The site also saw a 6% increase in new users over the same period, and the platform maintained 99.99% uptime across all destinations.

Clare Cooper, Senior Growth Marketing Lead at Hammerson, put it plainly: “the increase in daily search volume far exceeded expectations.” 

That quote is useful not because it's enthusiastic, but because it suggests even the people closest to the project didn't fully anticipate how much demand had been sitting there, waiting for a better experience to surface.

Is Your Site Search Suppressing Demand?

Here's a practical question: do you know what your daily search volume looks like?

Not the headline sessions or bounce rate figure. The actual number of searches being run on your site each day, and what percentage of those return no results.

Most brands don't track this closely. 

A study by the Baymard Institute found that 70% of e-commerce sites don't support common search query types effectively, including category terms, symptom-based queries, and natural language phrases. That's a significant share of the market essentially ignoring one of the most direct signals visitors send about what they want to buy.

A few things worth checking on your own site:

• What are people actually searching for? The query log tells you a lot about how visitors think about your catalogue versus how you've organised it. If you're seeing lots of category and intent terms but your search only handles brand names, that's a structural mismatch.
• What's your zero-results rate? Anything above 10-15% is worth investigating. For many sites, it's considerably higher, and each of those is a visitor who expressed intent and got nothing back.
• What happens after a failed search? If visitors who hit a zero-results page bounce at a significantly higher rate than those who get results, that's a direct revenue case for improving it. 

The Hammerson data is useful partly because it shows how much latent demand can exist on a site that's already performing well by most conventional measures. The problem wasn't that people weren't interested. The experience just wasn't giving their interest anywhere to go.

What to Do Next

Pull your search query data and look at the zero-results rate first. That's the quickest way to size the problem. Then check whether your search handles intent and category terms, or only exact brand matches.

If the gap looks significant, the Hammerson build is a useful reference point for what a modern AI search layer actually involves: intent recognition, smart suggestions, live data enrichment, and reliable infrastructure underneath. None of it is exotic, but it needs to be built properly rather than bolted on with a basic plugin.

You can read the full breakdown of what was built for Hammerson on the case study page. And if you want a quick read on where your own search stands, we run 30-minute search audits — book one here.

Frequently Asked Questions

1. How do I know if my site search is actually underperforming?

Start with two numbers: your daily search volume and your zero-results rate. If your zero-results rate is above 10-15%, that's a clear sign your search isn't keeping up with how visitors are actually looking for things. Most analytics platforms track this. If yours doesn't, that's worth fixing first. A high bounce rate from search results pages is another reliable indicator that something isn't working.

2. What's the difference between keyword search and AI-powered semantic search?

Keyword search looks for an exact match between what someone typed and what's in your database. If a shopper types "casual dining" and your database only has restaurant names, keyword search returns nothing. Semantic search understands intent, so it can connect "casual dining" to the right results even without an exact string match. For retail and destination sites with large, varied catalogues, that difference tends to show up directly in search volume and conversion.

3. How long does it take to replace a legacy search with AI search?

It depends on the size of your catalogue and how your data is structured, but a phased build typically runs over 60-90 days. The first 30 days focus on auditing your current search data and setting baselines. The next phase covers architecture decisions and build. The final phase is testing, rollout, and iteration. The Hammerson build followed a similar timeline.

4. Is this only relevant for large retail destinations, or does it apply to smaller e-commerce sites too?

The search intent problem applies at any scale. Smaller catalogues can actually see faster wins because there's less data complexity to work through. The core issue, which is search failing to understand what visitors actually mean, is just as present on a 500-product e-commerce site as it is on a destination with hundreds of retailers. The Hammerson numbers are dramatic, partly because of the scale, but the underlying dynamic is the same.