‘Photographing’ big data: a word cloud visualisation of consumer demand for insurance, as expressed by internet searches.
COLLECTIVE CONSCIOUSNESS
Every day, people make around 5.6 billion Google searches. What’s more, they tell Google things that they wouldn’t share with their friends and family – perhaps not even their partner or their doctor. If you have your phone or laptop to hand, go to Google, type ‘I’ve just h’ into the search and you’ll see what I mean. As I’m writing this, here are some of the searches Google anticipates, based on what people have searched for before:
• ‘I’ve just had enough’
• ‘I’ve just had a car accident’
• ‘I’ve just had a panic attack’
• ‘I’ve just had the craziest week’
• ‘I’ve just had a baby, what am I entitled to?’
• ‘I’ve just had a period but feel pregnant’
• ‘I’ve just had a poo and there was blood’
What this set of searches highlights is that people turn to Google not just for practical information, but for advice at life’s most significant moments. Because we are unfiltered when we use Google, the data created by our internet searches is hugely powerful. The data scientist Seth Stephens-Davidowitz has called it ‘the most important data set ever collected on the human psyche’, and I don’t think he’s exaggerating. You can think of search data as a vast reservoir of human needs and desires, which grows deeper every minute. It’s a profound expression of our collective consciousness.
I started learning about search data when I was head of strategy and development at Experian, and quickly became obsessed. One of the company’s marketing businesses, Hitwise, had a software tool called Search Intelligence. Through partnerships with internet service providers and web browser companies, it worked by anonymously compiling the internet searches made by millions of internet users in Australia, America and the UK. Logging in to the software allowed you to mine the data, using keywords you were interested in. You could also look at which searches were driving traffic to a specific website, or which websites were gaining traffic for a specific search.
As I talked to Hitwise’s account management team, I realised that most of our clients were using the tool in rather mundane ways: someone from the customer insight department at a bank might take a screenshot of the twenty most popular searches for mortgages and paste them into a monthly PowerPoint report; someone from marketing might use data on a competitor’s website as evidence that they needed more budget to spend on Google ads. It was while I was mulling over other possibilities for the data that I was introduced to Steve Johnston and Liam McGee, the data scientists I mentioned in Chapter One. Their company Kaiasm had partnered with Hitwise to offer a service they called ‘demand taxonomy’ to Experian’s clients in the retail sector. It involved comparing which searches were driving traffic to the client’s e-commerce website with all the searches happening on Google for products that the client stocked. The idea was simple – to find gaps between demand and supply that the client could capitalise on – but it took a lot of data and analytical smarts to execute it.
Steve gave me a real-life example of a demand taxonomy he had built for the DIY brand Screwfix. By analysing internet search data from Hitwise and comparing it to product listings on Screwfix’s website, Steve could see all kinds of commercial opportunities that Screwfix had been oblivious to. What would you call a tool that uses compressed air or gas to drive nails into wood? Screwfix called it a ‘nailer’ and listed it as such on their website. However, search data indicated that literally ten times as many people called the tool a ‘nail gun’. How about a lamp mounted on an exterior wall of a building, that illuminates when an infrared sensor is triggered by the approach of a person or animal? Screwfix categorised it as an ‘outdoor light’, but a large majority of people call it a ‘security light’. Lastly, if you were renovating your kitchen and had installed a new induction hob, granite worktop and double-door fridge freezer, what would you search for if you wanted to replace the strip light bulb on the ceiling? Search data suggests that most people would google ‘kitchen lighting’, yet Screwfix, thinking, not unreasonably, that downlights, spotlights and recessed LEDs can feature in lots of different places around the home, hadn’t listed them by room at all. By changing how they described and classified their existing products to align with what people were searching for, Screwfix was able to win much more traffic from Google and increase sales as a result.
I was inspired by this story, and as I got the train to play squash one Friday, my head was spinning with ideas. My opponent that evening was a friend who ran the digital marketing department at the Open University. It seemed obvious that education was an ideal sector to try big search data analysis. I lost the match, but in the pub afterwards I sold a six-month taxonomy project on further and higher education demand.
Little did I know that at the same time, venture capitalists were cooking up an even more ambitious scheme to use search data. One day, an account manager at Hitwise received a call from someone at Forward Internet Group, casually enquiring how much it would cost to license Hitwise’s search data for the past ten years; because the storage costs of making such a large data set available online would be very high, Hitwise only allowed users of Search Intelligence to access two years of search data. In putting together a quote for Forward Internet Group, the account manager eventually managed to find out from them what they were planning to do with the data. Their idea was to get their data scientists to mine it for the single best untapped opportunity for a new online retail business. The methodology would be similar to the one Steve and Liam’s team had used to show Screwfix and the Open University how to describe and classify their products and courses to gain more traffic. However, the important difference was that Forward Internet Group weren’t trying to drive more sales for a business they had already established; they wanted to start a business from scratch, which meant there was a far greater volume of search data to analyse.
The monumental scale and complexity of this challenge meant Forward Internet Group’s data scientists disappeared into a metaphorical bunker for months. Fuelled by coffee and Haribo, they would work long into the night. Colleagues started to notice their increasingly unkempt appearance and a wild look in their eyes: they were becoming obsessed. Eventually, the executive team lost patience and summoned the data scientists to the boardroom. At the head of the table sat Neil Hutchinson, the company’s founder and CEO, flanked by his most trusted advisers. ‘Let’s hear it,’ he began. ‘What have you found?’ There was an uncomfortable silence as the data scientists, incongruous in their hoodies and trainers, looked at each other. ‘Come on,’ said Hutchinson, ‘We asked you to mine the data to find the best opportunity for a new business – what is it?’ Finally, one of them found the courage to pipe up: ‘It’s parrot cages.’ ‘Parrot cages?’ Hutchinson repeated. ‘Are you winding me up?’
They weren’t. The search data had revealed that the biggest gap between demand and supply in the online retail market was for parrot cages. To his great credit, Hutchinson put aside his scepticism and committed himself to entering the parrot cages business. As a first step, Forward Internet Group set up a simple one-page website called Just Cages and spent a small amount advertising it in Google listings. It wasn’t possible to actually buy a cage from Just Cages at this point, but two weeks of clicks to the website showed that the data scientists were right – there really was unmet demand for parrot cages on the internet. Next, they found a so-called ‘dropshipper’ in India – a wholesaler who would send the parrot cages directly to the customer, meaning that Just Cages didn’t have the financial risk and practical inconvenience of holding stock. The final piece of the jigsaw was to build e-commerce functionality – the capability to list products, take payments and process orders online. To keep Just Cages as simple as possible, Forward Internet Group used Shopify, an off-the-shelf product for online stores.
The results were remarkable: within a year, Just Cages had turned over £3.2 million, and set up ten further e-commerce websites to sell kennels, hutches, aquariums and vivariums (environmentally controlled enclosures for pet reptiles). What’s more, the opportunities for new online stores turned out not to be limited to accommodation for pets. Just as Forward Internet Group’s data scientists were falling down their search data rabbit hole, the entrepreneurs Richard Tucker and Joe Murray were crafting a vision to change the world of home and garden retail, using the gaps between demand and supply that their analysis had uncovered. The company they founded, World-Stores, consisted of a multitude of niche websites with names like MattressesWorld, ShedsWorld and even TrampolineWorld – all designed to meet the consumer demand for unusual home and garden products that was being expressed through Google searches. By continually monitoring search data for emerging trends, WorldStores stayed one step ahead of its traditionally minded competitors. ‘One year there was a thing about trundle beds,’ recalls Tucker. ‘None of us had a clue what they were, but people were typing it into search. So we got them in stock and they sold like hell!’ (It’s a low wheeled bed that is stored under another bed, by the way.)
While Just Cages was a clever experiment that exceeded expectations, WorldStores was a serious business based entirely on analysing search data. By the time of its sale to the Dunelm Group in 2016, it listed over half a million home and garden products across its network of online stores, had annual sales of more than £100 million and employed 650 people. It’s a powerful example of the way that big data can enable consumer choice, economic growth, job creation and prosperity.
I remember sitting at my desk in the Experian office in spring 2012, reading news coverage of a recent venture capital investment in WorldStores, when my phone rang. It was my friend and marketing mentor, Ruth. ‘I won’t beat about the bush,’ she said. ‘I’ve called to ask you a favour.’ Ruth told me that one of her old colleagues from the strategy consultancy McKinsey was planning to set up a new business offering unusual types of insurance online. Immediately my ears pricked up. ‘Steven’s very smart commercially and he’s got a technical co-founder,’ Ruth continued, ‘but he needs digital marketing expertise. Would you meet him to see if there’s someone you could recommend?’
Although I was intrigued by the idea of applying internet search data analysis to the insurance market, I wasn’t looking for a new job. My work at Experian was varied, intellectually stimulating and came with perks including business trips to Paris and New York. Why would I swap that for the chaos and uncertainty of life at an early-stage start-up? I told Ruth I’d be happy to meet Steven for a coffee and offer some advice, but that was as far as I expected it to go.
However, as I began analysing the search data I’d downloaded from Hitwise in preparation for the meeting with Steven, my intrigue turned into excitement. To outsiders, big data analysis sounds dull and laborious, but it can be profoundly rewarding. It is the sort of activity you can completely immerse yourself in: distractions melt into the background and time seems to stand still, as all your mental resources become absorbed by it. Talk of being in a ‘flow state’ is more often associated with yoga, painting, or rock climbing, but it is there in data analysis, too. I sometimes liken it to swimming silently through a deep blue ocean, looking for pockets of phosphorescence. In the insurance search data I’d downloaded from Hitwise – this great outpouring from millions of minds – lights were glowing in every direction. There were so many opportunities that I couldn’t decide which ones to swim towards and examine more closely, so I decided to take a sort of underwater picture instead.
Because I was interested in the consumer demand that the current competitors in the market were missing – just like Forward Internet Group and WorldStores – I started by getting rid of the searches that were more than 80 per cent likely to result in a click; a lower percentage of clicks on a search would indicate that people couldn’t immediately see what they were looking for on the search results page. I then made a long list of ‘stop words’ – words and phrases that wouldn’t tell me about consumer demand for products. This included brand names of insurance companies like Direct Line and Aviva, synonyms for insurance like ‘policy’ and ‘cover’ and generic words about the process of buying and using insurance like ‘quote’ and ‘claim’. Once I had filtered these out, I dragged and dropped the data into an online visualisation tool called TagCrowd, and asked it to show me the words that appeared most frequently. This was the result:
‘Photographing’ big data: a word cloud visualisation of consumer demand for insurance, as expressed by internet searches.
Parts of this picture were reassuringly obvious: there was clearly plenty of demand for car insurance, travel insurance, home insurance and so on. More interesting, though, were clusters of lower-volume searches. Together, searches for words like ‘business’, ‘liability’, ‘professional’, ‘indemnity’, ‘market’, ‘traders’, ‘van’ and ‘taxi’ suggested a great opportunity to respond to insurance demand from small business owners, and perhaps explained why the specialist direct broker Simply Business seemed to be growing so quickly. Searches for words like ‘campervan’, ‘caravan’ and ‘motorhome’ on the one hand, and ‘young’, ‘drivers’, ‘black’, ‘box’ and ‘learner’ on the other implied that the opportunity in motor insurance might lie outside main-stream vehicles and motorists. And, of course, the early signs of the future success of Bought By Many were already apparent in words like ‘medical’, ‘conditions’, ‘pet’, ‘cat’, ‘dog’ and ‘horse’.
Rush hour had begun as I crossed London Bridge on my way to meet Steven, and masses of City workers were marching in the opposite direction towards the station. I held the line I was walking and noticed how easily the crowd parted in front of me. I realised that it was possible to go against the flow, and thought I might enjoy doing it.
I got to Pret, bought a coffee and texted Steven to tell him I’d arrived. I added that he might not recognise me, as I had grown a beard since the photo on my LinkedIn profile had been taken. I recognised him, though – tie-less in a grey suit, he looked serious, lean, energetic and restless. Ruth had told me he could be very direct, and in that first meeting he certainly was, at one point asking, ‘So why are you here?’ I replied that I wasn’t looking for a job, but that I had a lot of respect for Ruth and liked meeting interesting new people. I’d brought a printout of the search data analysis with me, and Steven was clearly impressed. He told me it was the first time someone had come to a meeting with him having properly thought about what Bought By Many could be rather than turning up expecting to be sold to, before spouting the conventional wisdom that millions of pounds would need to be spent on TV and above-the-line advertising to drive brand awareness if Bought By Many was going to be successful. He showed me a PowerPoint deck he had been using with potential investors and insurance partners, which asserted that the insurance market was ready for disruption and that Bought By Many was the company to do it. Insurance companies’ profit margins were under pressure from price comparison websites, and new regulations meant that insurers would be forced to hold more capital unless they diversified into new product lines and geographic areas. At the same time, group buying was gaining acceptance among consumers, thanks to Groupon and collective energy switching clubs. Steven told me a story about leaving Close Brothers, where he had worked after McKinsey. He had phoned AXA PPP, who provided the company health insurance scheme that he’d been covered by as an employee, and asked for a quote to continue the same cover as an individual. He knew Close Brothers had been paying just over £1,000 a year on his behalf, but now AXA wanted over four times that amount. Nothing about his or his family’s health had changed – it was simply an example of financial services companies giving preferential treatment to large corporations, and taking advantage of ordinary people to subsidise it. His vision for Bought By Many was to use the collective buying power of like-minded individuals to redress such power imbalances.
The following week I met Guy; in contrast to Steven, he was very laid-back and at first they seemed like unlikely business partners. Guy told me his side of the Bought By Many foundation story: after building some software to make it easier for his group of friends to split the expenses of their annual ski trip, he had started thinking about what else online groups might be capable of doing. He had been working at a bank where Steven was running the wealth management arm, and their ideas had collided. Guy had built tech businesses before, in San Francisco during the dotcom bubble and then in Europe, but they had always been structured as consultancies, which had made it hard to create equity value. There was not much of a start-up scene in London at that time, but he was convinced that the opening of the Google Campus in Shoreditch, along with new tax incentives for early-stage investors, meant that it was about to explode into life.
Here was a chance to push the boundaries of search data analysis and transform financial services, with people I’d instantly gelled with. My mind was made up. I accepted Steven’s offer to join Bought By Many as chief marketing officer, and agreed to resign from Experian as soon as the seed money from our investor was in the bank.
In Chapter One, I mentioned what we spent money on in the early days of Bought By Many: a printer, a coffee machine and rent for our miniscule office. Once we had sorted out these basics, my top priority was to build a ‘demand taxonomy’ for the insurance market. I haggled with Hitwise for a discounted licence, and hired Kaiasm to help crunch the data. My initial word cloud analysis had been based on 10,000 rows of search data from one twelve-week period in 2011. That might seem like a lot, but it was unlikely to be representative because of seasonality: people need different types of insurance at different times of year, the summer peak in demand for travel insurance for family holidays being one obvious example. It was also only a small sliver of all the data that was available, and like Forward Internet Group and WorldStores, I wanted the motherlode. I downloaded every search that had been made for insurance in the previous two years and sent it to Steve and Liam to compile and deduplicate. It turned out that in the UK during that period there had been 105 million searches for insurance, distributed across almost 850,000 distinct search expressions. Kaiasm’s team of data scientists used software they had built to count the number of web pages that contained each search expression in their title – that would be our measure of how well existing websites were meeting the demand people were articulating through their Google searches. Finally, Guy built an online tool to make it easier for us to query the database, and Liam created a visualisation of all the best opportunities, which we mounted on the office wall.
How did we use this magnum opus? At first, we tested the ideas we already had, including the schools rugby insurance and diabetes travel insurance examples I described in Chapter One. Next, we tested ideas from the insurance companies that Steven had started commercial discussions with, but they were all terrible. The insurers wanted to sell things like insurance for wine collections, kidnap-and-ransom insurance and indemnity insurance for planning consultants. The data showed that there was an abundance of supply for these products, but almost zero demand – the ideas all involved types of insurance that would be unlikely to result in many claims, and would therefore be highly profitable. Instead of responding to what potential customers actually wanted, the insurers preferred to develop products that met their own needs and then hope they could convince people to buy them.
The insurance demand taxonomy visualisation.
Having rejected the insurers’ ideas, we decided to start groups on the Bought By Many website for types of insurance where there was clear demand that insurance companies weren’t meeting. I would see if I could recruit fifty or a hundred people to a group, and then Steven would go and present insurers with this tangible evidence of consumer demand. Like the proverbial kid in the candy shop, I picked the quirkiest opportunities first: quad bike insurance, canal boat insurance, even stage hypnotist insurance. Sure enough, people quickly began to join, and from them I learned that:
• Insurers over-charge for quad bikes that are used for track racing but aren’t road-registered.
• There is only one dominant provider of insurance for narrowboats, and owners think more competition would ‘keep them honest’.
• It’s surprisingly hard for stage hypnotists to get the public liability insurance that’s mandatory for performing in many venues.
The trouble with these opportunities from a commercial perspective was that they weren’t very scalable. To meet these three niche demands would have required signing deals with three separate insurance companies, and Steven was discovering that closing deals could be a protracted process. I realised I needed to find clusters of related opportunities in the data – or, returning to my diving metaphor, to swim towards a bigger, brighter patch of phosphorescence.
That was how I discovered the golden opportunity that would shape Bought By Many’s corporate strategy for years to come. Imagine me, crazy-haired, looking up from my laptop for the first time in days, and announcing to Steven, Guy, Helen and our newest recruit David, ‘I’ve found something that’s going to change the face of financial services forever: Pug insurance.’ With that, I strode out of the room and went to make a celebratory cup of tea.
Pug Insurance
Why was I so excited about it? Firstly, because a lot of people seemed to be searching for pet insurance that was specifically for pugs, yet there wasn’t a single page on the web with ‘pug insurance’ in the title. Secondly, and just as importantly, it was clear that there were tangible reasons why pug owners weren’t satisfied with generic dog insurance. It’s worth going into them in some detail.
At the time, pugs were on-trend: statistics from the Kennel Club showed that they were close to becoming the UK’s most popular puppy. As a result, prices for pug puppies were skyrocketing, which in turn seemed to be leading to a spike in thefts. By quickly analysing data from DogLost.co.uk, a free service that owners can use to post notices about missing pooches, I could see that pugs had been the most-stolen breed in the previous twelve months, their small stature and friendly nature presumably also contributing to their ‘thieveability’. If you had a mainstream pet insurance policy at that time and your pug went missing, £1,000 was the maximum payout you would receive, but buying another pug puppy would cost you as much as £5,000. The theft cover offered by insurers was totally inadequate.
What’s more, many pet insurance policies also contained a blanket exclusion for so-called ‘breed-specific conditions’ – medical problems that particular breeds are known to be prone to. For the growing population of pug owners, this meant that vet treatment for encephalitis (an inflammation of the brain) or hip dysplasia (a problem with the hip socket that can result in dislocation) would not be covered. Facebook comment threads and discussions on online forums showed that pug owners took a dim view of this: what was the point of having insurance, if it wasn’t going to pay for the problems that would be most costly to treat? When it came to pugs, pet insurance products were behind the times – and Bought By Many had a great opportunity to bring them up to date.
Thirdly, I was excited because almost every other dog breed I checked showed the same pattern as pugs: there was an abundance of demand that insurance companies had failed to address. French bulldog insurance, Neapolitan mastiff insurance, Siberian husky insurance and, of course, German shepherd insurance – I had a list of more than seventy groups for Bought By Many to start, before it occurred to me that I could be looking at cats as well.
Once I’d explained my findings, the rest of the team quickly rallied behind them. While Helen and I got to work writing content and setting up multiple Facebook ad campaigns, Steven and David pitched the list of breed groups to the major players in pet insurance. There was a bidding war for pug insurance that was eventually won by Petplan, who agreed to increase the theft and loss cover on their product for members of our group. More Than agreed to offer a 20 per cent discount on dozens of breeds they liked the risk of or wanted to diversify into, while a smaller broker offered £25 cashback on a further ten breeds. While discounts and cashback offers didn’t tackle the shortcomings of the products, it at least left Bought By Many members with more money to deal with problems that insurance wouldn’t help with.
This was the platform from which we built the company as it is today. In 2017, three years later, we were fed up of waiting for our insurance partners to improve their pet insurance products so we designed our own, transforming Bought By Many from a distributor of third-party products into an online broker with the authority to take binding decisions on insurers’ behalf. Quantitative insights from analysing the demand taxonomy and qualitative insights from tens of thousands of interactions with pet owners on Facebook and Twitter told us what our products should look like. We set the maximum claim limit for theft and loss at £6,000. We did away with breed-specific exclusions and the age-old insurance industry practice of charging higher renewal prices to existing customers than to new customers. Recognising that for most people pets are part of the family, we included cover for travel, complementary therapies and the costs associated with the death of a pet as standard. The data told us that some people feel unhappy about paying for insurance they don’t claim on, so we invented a product that offers cashback for every year that no claim is made. The data told us that other people worry about the cost of pet insurance increasing over time, so we invented a product for puppies and kittens where the price is guaranteed for the pet’s lifetime. Some people hate paying the excess – the amount an insurer asks you to contribute to the cost of any claim – so we gave them the option to set it at zero. With annual sales of £141 million, Bought By Many now employs 220 people and protects more than 325,000 pets. No insurance company has higher customer satisfaction scores. It is the UK’s most trusted pet insurance brand, Europe’s fastest-growing insurance company and the world’s number one insurance provider for pet businesses and unusual pets.
It was search data – an effortless by-product of online life – that made this possible. The availability of large quantities of aggregated and anonymised internet searches was a necessary condition for Bought By Many to succeed – exactly as it was for Just Cages and WorldStores. If you played along with the experiment at the start of this chapter and still have a Google window with ‘I’ve just h’ in the search box, go back to it and hit return. Now look at the address bar in your browser; you’ll see a long string containing various parameters, like the search query you’ve just made (‘q=i%27ve+just+h’), where it originated (‘source=hp’, short for the Google ‘homepage’) and when exactly it happened (‘ei=’, followed by an encoded version of the time in microseconds). This data now exists somewhere on servers belonging to Google and your internet service provider. If you’re using a browser that isn’t Google Chrome or have browser extensions installed, it might also exist on servers belonging to the developers of those pieces of software. If you click on one of the search results, Google will send your query, ‘i%27ve+just+h’, to the destination website.
There are two possible futures for this data: it can be deleted or it can be used. If you’d prefer that it was deleted, you can use a privacy-preserving search engine like DuckDuckGo instead of Google, or a privacy-preserving browser like Brave – they will see to it that your search data vanishes almost as soon as it’s created. But I hope to persuade you that it’s better if search data is used to make something. With apologies to my old colleagues at Hitwise, they are like the manure collection and processing functions of the agribusiness we imagined in Chapter One. They compile search data from internet services providers and browser software and refine it into something useful: Search Intelligence. Just Cages, WorldStores and Bought By Many are like farmers, whose products depend on Search Intelligence for nourishment.
While human life sometimes improves in great leaps, most of the time it advances in tiny increments, like having more parrot cages to choose from and being able to find unusual types of pet insurance. Even infinitesimally small improvements can make a big difference to some people. Take Micah Carr-Hill, a food scientist from south London who insures his yellow Labrador, Chief, with Bought By Many. Chief is a trained autism service dog, who helps Micah’s son stay safe and increases his independence. Autism service dogs are expensive, and Chief cost £6,000. Losing him would have a serious impact on Micah’s family, which is why it was such a relief to find pet insurance that would cover the full cost of buying another dog.
Small improvements can have a ripple effect. In 2014, a researcher at Boston Consulting Group was working on a report about innovation in the insurance industry for Ping An, China’s largest insurer. She’d read an article about Bought By Many in Wired magazine, and contacted Steven for an interview. A few months later, one of her colleagues got in touch to say that the chairman of Ping An’s general insurance division was bringing his leadership team to London and wanted to meet Bought By Many. Steven found the thought hilarious: Ping An had 1.5 million employees, over 400 million customers and annual sales of more than $100 billion. At 115 storeys and 562 metres, its soon-to-be-completed HQ in Shenzhen would be the fourth tallest building in the world. Meanwhile, Bought By Many had eight employees shoehorned into a six-person room on a Clerkenwell backstreet.
We welcomed the chairman, Mr Sun, and his team to the largest of the shared meeting spaces in our serviced office building, a low-ceilinged basement room with a hot pink carpet. We had been asked to set up a video link so more Ping An colleagues could join the meeting remotely, but as there were no video conference facilities in the room, Guy had Blu-Tacked a webcam to the whiteboard. I sensed we were all feeling slightly hysterical by this point, aware that we were winging it and that the meeting would end in either triumph or disaster. After much hand-shaking and an elaborate exchange of business cards, Steven began his introduction. After a few minutes, Mr Sun noticed that neither Steven, Guy nor I were wearing ties, and respectfully removed his own. Noticing that Mr Sun was now tie-less, the other male members of the Ping An team did the same. I was next to present, with an overview of Bought By Many’s data-driven approach to product development and marketing. During a brief pause when I was a few slides in, Mr Sun got up and strolled to the hospitality trolley at the back of the room. He picked up the hot water flask, and I watched in horror as he poured it into a glass tumbler which he must have assumed was heatproof. The awful memory of our investor John plunging through the floor, latte in hand, flashed before my eyes. Incredibly, Mr Sun’s glass remained intact. However, no sooner had he sat down, than several of his colleagues began getting up to help themselves to their own glasses of hot water. I couldn’t look.
Somehow, we got to the end of our presentations without any breakages or scaldings. There had been a lot of questions from the Ping An team, and now Mr Sun was ready to tell us what he thought. He said that he agreed with Boston Consulting Group that Bought By Many had an innovative approach to insurance, and that Ping An was committed to partnering with smaller companies at the cutting edge of financial services technology – would we work with them to design and launch a new portfolio of travel insurance products for Chinese consumers?
We weren’t sure how to respond. It was a hugely flattering offer, but our agenda was at that time about scaling up customer acquisition for our partners in the UK. Conventional start-up wisdom tells you not to get distracted by speculative opportunities; you should have a laser-like focus on your goals. We turned to our investors for advice, who were sceptical; they warned us that Ping An could copy our intellectual property, and that something as simple as getting an invoice paid was fraught with difficulty when doing business in China. The more we thought about it, the more reasons there were to say no. We would have to find Mandarin-speaking staff with expertise in insurance and digital marketing. There were Chinese equivalents of Google and Facebook – Baidu and WeChat – that we would have to learn about from scratch. But the prospect of shaping how insurance would develop in the world’s most populous country had got hold of us, and it wouldn’t let us go. We said yes.
Like so many other industries, insurance in China has evolved at breakneck speed over the past decade, and as a result, it has some quirks. When we started our project with Ping An in the summer of 2015, much of the growth in the market was coming from products that seemed bizarre from a European perspective. During the previous year’s Football World Cup, one Chinese insurer had sold a policy known as ‘watching football drinking too much’ insurance, covering fans for self-inflicted alcohol poisoning. There were products offering small payouts for minor misfortunes like a sleepless night, an outbreak of acne or a cut to a finger while preparing a meal. In the travel insurance category, one of Ping An’s competitors launched a policy that offered holidaymakers compensation if views of Beijing, Shanghai or Xi’an were obscured by smog. The Chinese insurance regulator would later ban policies where there was no risk of financial loss to the customer, but at the time Ping An wanted new products that would stand out. One of their ideas was a policy aimed at adventurous eaters, which would pay out in the event of food poisoning. Could we find evidence of demand by analysing data from Chinese search engines?
The first step was to build a Chinese version of the data-base of anonymised internet searches that had revealed pug insurance as a golden opportunity in the UK. Hitwise didn’t operate in China, and as Google is blocked there, compiling a taxonomy from autocomplete suggestions was impossible. Luckily, Steve Johnston knew of a UK-based agency that specialised in search marketing in China. They had access to Chinese keyword planners – the software tools that search engines offer to advertisers to help them work out which search terms to target. The agency painstakingly copy-pasted volume data for every insurance-related search term into a spread-sheet, with Steve’s team adding data about how many times each term appeared in the title of a page on the web. It was slow, labour-intensive and costly, which meant we could only scratch the surface of what Chinese customers were expressing through their internet searches.
Nevertheless, we gained some powerful insights. It turned out that Ping An’s idea of ‘food poisoning insurance’ was not as left-field as it had first seemed; there was evidence of people searching for insurance that would cover farm stays, which are popular among Chinese foodies and carry some risk of stomach upsets. The data also told us where Chinese consumers were considering travelling to: surprisingly, there were many more searches for travel insurance to African countries than North American ones, and three times as many for ‘Nepal travel insurance’ as ‘UK travel insurance’. There was a lot of unmet demand for travel cancellation insurance, suggesting that consumers weren’t satisfied with what was available. Finally, we found demand for travel insurance to cover hiking trips to mountainous areas of China, explained by the fact that many existing policies excluded altitude sickness.
We flew to Shenzhen to present our findings. In the UK, we were used to having to wait months for our partners to mull over an opportunity, but Ping An took a decision immediately. Within two weeks, they had launched seven new travel insurance products based on our insight.
The ripple effect quickly spread beyond China, and coverage of our work with Ping An brought us a slew of enquiries. Insurers wanted to know if search data analysis could help them with product development in Switzerland, Poland, Korea and Australia. Still in a single-room office, we created Bought By Many International and took on projects involving mobile phone insurance in Mexico, life insurance in Canada and car insurance in Italy. It didn’t seem to matter what combination of insurance and country we tried; our approach just worked. Using search data to address the problems of UK pug owners has made insurance better for people all over the world.