How B&Q is reinventing growth with AI-powered Search

We’re seeing a few key shifts. First, customers are looking for far more choice and doing much more comparison. We recently launched our marketplace, which allows us to offer millions of products, and naturally that leads to increased comparison.

We’re also seeing more informational searches, with customers looking for product features or how to install something, alongside traditional product searches.

Search has also become much more unpredictable. We used to plan budgets rigidly with a clear sense of seasonality, but that’s changed. You might have a very warm Easter and suddenly need to pull budget forward because demand spikes in March. Later in the year, demand might soften. That flexibility has become essential because consumer behaviour is far harder to predict.

We were seeing consistent instances where demand was clearly there, but without flexible budgets we couldn’t capture it. That meant we weren’t showing customers what they could be buying from us or helping them complete their home projects.

When we visualised this, you could see big spikes in demand while investment stayed flat. There was a clear gap where we weren’t meeting customers at the moment they were searching. That’s what made us realise now was the right time to test demand-led growth. Without it, we risked staying stuck in a rigid model that no longer reflected how people actually behaved.

It’s useful to start with what demand-led growth isn’t. Sometimes people think it just means having extra pockets of budget for unexpected demand events. But in reality, it’s about being completely fluid.

Rather than trying to predict demand, you allow budgets to flex naturally in response to real-time signals. As long as you’re meeting efficiency and profitability targets, you can invest when demand appears, instead of forcing spend into a rigid plan.

We use AI and data in several ways. One of the most impactful is using total business sales data to spot emerging trends and then aligning our search activity to capture those opportunities.

Recently, a grey vase we sell was picked up by an influencer and received significant press coverage. Demand spiked very quickly. That uplift first appeared in our total business data, and AI helped us identify the trend early. This meant we could rapidly build strong Search coverage, so that when customers searched, we were visible and aligned with our wider marketing activity.

SA360 is the platform we run all of our paid Search activity through. The impact of the test was significant. Between April and July, we saw a 33% increase in online revenue from SA360-driven paid Search alone.

Importantly, our return on ad spend (ROAS) remained stable throughout the period, which made this a very successful test.

Reaching that point didn’t happen overnight. Over many years, we built a robust evidence base through geo-testing and incrementally testing to demonstrate the value of paid Search. This year, we were finally able to secure approval to run flexible budgets over peak periods. The three-month test involved close collaboration with teams across finance and data, and the results were extremely positive.

A key part was recognising that different teams started from different places, and meeting each of them there. Our finance team, for example, wanted full visibility of the data, so we shared everything from weekly spend updates to monthly deep dives.

We also broke performance down by fulfilment type, such as click and collect versus home delivery, which helped finance understand profitability in more depth.

Rather than asking for uncapped budgets, we focused on reassurance: spend would only increase when demand and returns were there. Weekly check-ins were invaluable in understanding concerns and tailoring reporting to build confidence.

We use a combination of measurement approaches. For weekly reporting, we rely on in-platform SA360 data and Google Analytics. For less frequent reporting, we use total business sales data to assess profitability.

More recently, we’ve been exploring media mix modelling (MMM) alongside causal impact studies. One key learning has been ensuring everyone understands the differences between these attribution models. Once that understanding is in place, it becomes much easier to align on how success should be measured.

Listening has been crucial. Understanding concerns and addressing them directly helps put teams at ease.

With the data teams in particular, it’s been powerful to show the real-world impact of the data they provide and better understand how it helps us grow the business and meet customers at the moment they’re searching.

One of B&Q’s values is bravery, and that became important at moments when things felt uncomfortable. We reminded ourselves that this was an exciting test and worth trying. That mindset helped galvanise teams and keep momentum going.

Securing buy-in for a small initial test is key, perhaps over a month or six weeks during a peak period. Once you have that first test, you can build the case for ongoing adoption.

Before asking for it, proving the value of the channel through geo tests or modelling helps reduce perceived risk. The reassurance comes from knowing the budget will only be used if returns are delivered. That first test really is the foundation.

AI will become even more powerful as we feed in richer signals, whether that’s audience data, profit data or other performance indicators. The more signals we provide, the more effectively AI can optimise.

From a demand-led perspective, we’re still early in the journey. I’m excited to continue demonstrating the impact and highlighting moments where flexible budgets allow us to capture opportunities we would otherwise miss. That storytelling is key to helping the business see Search as a growth lever, not just a cost line on a spreadsheet.