Convenie by Colin Edwin

1. 97.7% Accurate


Olivia leant back on the chair.


And shot a couple hoops. Or, in reality, a dustbin in the corner of the room.


In the London Head Office of Convenie, a mid-sized grocery chain that prided itself on its small-sized store portfolio, Olivia sat waiting for her latest machine learning algorithm to finish a test run.


Two years before, Convenie had launched a loyalty card called, rather unremarkably: the ConvenieCard. It gave customers the feeling that their place of shop gave them something back. But, that wasn’t the real value to any retailer with a loyalty card. The real value came from all the behavioural data it provided on customers: which customers bought what, when, where and how often, and the ability to directly market products to customers via email and post.


However data is nothing without insight and so a team was needed to separate the signal, of which there was very little, from the noise, of which there was a lot. So within two years, Convenie built a team of six analysts (most nabbed from a rival grocer) and six second hand laptops too heavy to carry around.


‘What you up to this weekend Olivia?’ asked Reuben. Reuben was a junior analyst, but a capable one who’d done exactly the same job in a different office.

‘Ermmmm.’ She paused, for a few seconds. ‘Dunno, actually. Maybe watch the new season of Dark Reflection. You?’

‘Friend’s birthday party tonight, down The Dog. Hangover tomorrow. Get stoned Sunday. Pull a sickie Monday. Back to work Tuesday.’


While Convenie had made leaps and bounds in its data analytics capabilities, the insights the company gleaned were years behind what the ‘Big Four’ grocery chains knew and although its analysts were good, they weren’t the best in the industry. The poor data capabilities was also why Convenie had stopped bleeding market share but also wasn’t gaining any either. The company was a zombie for its shareholders and a stubborn puzzle for its management team.

Processing complete. Model complete. Model fit calculations complete.

Olivia glanced at the computer screen mid-way through shooting another scrunched up paper in the bin.

‘Model fit is 98.2% accurate. Jesus. Now I’ve overfit it. Yesterday it was 0.4%, now it’s 98.2%!’

Why, one might ask, is a model that is so accurate in predicting something – a problem?

Because it almost always means it is very good at predicting the slice of data it has been fed, but terrible at predicting everything outside of it. If you gave a model three days of sunshine, it would tell itself that there is always sunshine and hey, tomorrow, it might be 100% accurate. But test it on new data, say next week, it might only be 60% accurate. Next month, 50% accurate. In November, 10% accurate.

‘Olivia, you’ve been tweaking that decision tree for over a month now. Bell is going to want to see results. The guy looked like he was going to have an aneurysm yesterday, upstairs must be using a cattle prod on him after the AGM.’ Simon Bell was the Head of Analytics. Upstairs were the men in suits, who answered to the faceless mass of shareholders who may or may not have worn suits.

Olivia knew what Reuben was saying was true. A re-structure was on the horizon and one month of wasted of time would not effectively land with management.

So she put the model through another testing phase, this time using 12 rather than 4 weeks of data.


Back to shooting hoops.

Processing complete. Model complete. Model fit calculations complete.

97.7% accurate. Olivia paused. Hoops was over.

Now, with 104 weeks of data.


Processing complete. Model complete. Model fit calculations complete.

97.4% accurate. Olivia turned to Reuben.

‘Holy shit,’ she whispered with eyes wide open.

2. The Process

Olivia had stumbled onto a goldmine that no other grocery chain had yet found. What she found, was as follows.

When a customer in Convenie bought a packet of Smokey Convenie Potato Chips, 89% of those customers would buy a tub of Convenie Brain-Freeze Free Frozen Yoghurt either during the same shopping trip or the trip after. Of those 89% of customers who bought a tub of Convenie Brain-Freeze Free Frozen Yoghurt, 93% of them would go onto buy Serial Pain Killer Capsules during their next trip. And of those 93% of customers who bought Serial Pain Killer Capsules, 76% bought ‘Save my Stomach’ Sodium Bi-Carb Sachets. And of those 76% who bought ‘Save my Stomach’ Sodium Bi-Carb Sachets, within four weeks 29% will take out private medical insurance with Convenie’s subsidiary company Live Free Don’t Die Ltd.

Combining different products together, or ‘product bundling’, wasn’t new to the grocery business. Two of the ‘Big Four’ grocery chains had been positioning beer near the baby care aisle for years: for Dads who wanted to avoid emasculation when buying nappies.

But what made Olivia’s insight so different was that it demonstrated a chain of not two related products, but five completely different products, including a financial product which was notoriously lucrative. If one product was missing, the chain fell apart and the prediction failed to come to fruition: success of the private medical insurance depended just as much on the sodium bi-carb sachets as the painkillers depended on the yogurt.

Different theories of what maintained the chain emerged within the business. The food technicians believed the high salt and fat content in the potato chips drove a propensity to purchase something cold refreshing and sweet yet perceived to be healthier than ice cream. The marketing psychologists hypothesised that the chain was built on the product names which unconsciously drove customers to make the subsequent purchase in the chain.

To senior management, the theories were academic. It was the holy grail of business insight: valuable, deeply hidden and almost impossible to replicate. And only Convenie had it.

3. Exceptional Growth

‘Why are we putting medicine next the freezer aisle?’ one the shelf stackers asked.

‘Orders from senior management.’ replied the store manager, in a disapproving tone.

‘Figures… Asshats.’

12 months later, every Convenie store achieved exceptional year on year growth that City commentators put down to ‘strong company fundamentals’, which was another way of saying: ‘we have no idea, but they must be doing something right.’ Each store had been redesigned to maximise Olvia’s sequence, including leaflets for the medical insurance, were always readily available.

Every time a customer purchased one product they received an e-voucher to receive a discount on the next product. Marketing messages were pushed on a continuous basis.



Soon enough a nickname for the chain of purchases was adopted in the business: The Process. And customers who went through it, were Processed.

Within 24 months, the Big 4 was now the Big 5 and Olivia was Head of Analytics. Simon Bell got promoted to Chief Technology Officer, despite not knowing the difference between Unix and a Eunuch.

A recession hit.

Not as bad as The Great Depression of 1929, but worse than the Great Recession of 2007. Consumer confidence fell, unemployment tripled, wages froze and homes were repossessed. But even though retailers’ sales fell off a cliff: amazingly, The Process stayed strong. In the face of severe economic pressures, customers spent more than ever at Convenie.

4. The Lion or The Dog?

‘We’ve re-shuffled the store layout again and created an additional 2% multiplier for Processing,’ reported Reuben.

‘Based on the simulations, next week we will gain $12m incremental from existing customers and $4m from customers switching from competitors.’ Everything was stated in US Dollars since Convenie’s parent company was based in New York.

‘You know I hate it when people call it that…. Processing,‘ Olivia said in a serious but not-too-serious tone.

‘I’ll stop saying it when you stop misspelling it.’ Olivia’s dyslexia was a vulnerability Reuben never failed to take advantage of.

‘What you up to this weekend?’ asked Reuben.

‘Nothing really. You?’

‘I’ve got drinks down The Lion after work today.’

‘The Lion, where’s that place?’

‘Used to be ‘The Dog’. But they made the place smart and now it’s called The Lion. Want to come? You look like you could do with a drink.’

‘Aw thanks!’ Olivia said with heavy dose of sarcasm and then paused for a moment. ‘But actually, yes, thanks, that would be great.’ Due to workload, it had been a long time since Olivia had done something remotely described as social.

At 5pm, Reuben knocked on Olivia’s office door (she had her own office now) and they walked out of the tall, shiny building towards the smaller, less shiny – but shinier than it used to be – ‘The Lion’. Reuben turned to Olivia.

‘I’ve never noticed that Convenie on the corner.’

‘It’s a new store, opened last month. Part of the new store roll out in the South.’ Olivia replied with an air of superiority that she was still getting used to.

‘When was the last time you went into one Olivia?’

‘I’m not sure. Once last year I think. Shit, no. It might have been longer, now that I think about it.’

‘Same. Shall we go and see our insights in action?’ Reuben said in a half joking way. ‘I’ve even got a ConvenieCard which I’ve never used!’

Olivia flipped a mental coin. Tails. ‘Okay, sure.’

The two of them walked into the store; the shininess rivaled that of Convenie’s new Head Office where they’d walked from. Nothing seemed that different to how they remembered it years ago.

But after a few moments, here and there, they noticed differences that the everyday shopper wouldn’t. Olivia stepped over leaflets scattered on the floor down the healthcare aisle, where customers had tried to pull one from the stand but three or four followed with it. Reuben noted the potato chips near the front of store, with a freezer of yogurt next to it.

Leaflets on the floor and a freezer next to potato chips. Monuments to their greatest achievement.

Then, they began to notice something that the data didn’t show…

The customers walking around with baskets filled with the all too familiar products were clearly less affluent than the average member of society. In fact, they weren’t just in lower socio-economic groups, a phenomena Olivia had known for some time from the data: the customers being Processed were actually poor.

Their clothes were worn, the dirty hands digging out change from wallets and purses, and financial hardship written in the creases of their faces. Two or three looked (and to Reuben, smelled) like they were sleeping rough. Some had visible signs of health issues and those picking up the medicine and browsing the leaflets had a particular paleness that suggested a nutritional deficiency of sorts.

While ConvenieCard captured data on socio-economic group, it didn’t capture true poverty. If it tried, most applicants wouldn’t answer it accurately on the loyalty card applications and Convenie would be facing a PR nightmare for offending customers.

‘Olivia, look at these people. They’ve got nothing and they’re ALL buying products from The Process. These people actually look sick.

‘I told you, I hate that name – The Process,’ Olivia scorned.

Reuben could feel the blood filling his cheeks. ‘Fuck your name,’ he snapped back. He’d grown up on a council estate on the poorer east side of the city and he projected his family and friends onto the grim circumstances of The Processed customers.

‘Maybe it’s just the profile of this store.’ Olivia proposed.

‘This store is in one of the most affluent areas of the city! This is the fucking business district Olivia. Everyone in here buying this shit is either on minimum wage or they don’t work at all. See, look! That woman cleans the toilets in our office and look how much of that frozen yogurt shit she’s buying – she’s got a basket in each hand!’

Sure enough, Reuben was right. The situation was far worse in other stores. One could have mistaken the typical Convenie store to be a Foodbank if the decor were less colorful and price tags removed.

‘You’re exaggerating. This is getting silly now.’

Olivia’s words quickened.

‘So what if people are buying those products. We’re not making them do it. It’s what they want. It just shows how good the products are. It’s our job to look at the data. Not to say what’s right or wrong. Fuck this, thanks for putting a downer on the evening.’

With that, Olivia turned her back on Reuben, the store and The Processed.

5. Overcoming Saturation

Olivia’s weekend was consumed by thoughts and emotions that ranged from indifference to depression. Are we really making the whole nation poor and sick, she asked herself. There were attributes in the loyalty card data which now had a revelatory meaning in her mind: ‘low social grade’, ‘large families’, ‘credit challenged’, ‘high school education’ – all of which had a significant correlation with the model she’d constructed.

But she concluded that nothing could be done even if they were.

She couldn’t go to law enforcement; nothing here was illegal, it was business. Sharing the insight to competitors wouldn’t have any impact; the beauty of The Process was that it was difficult to replicate, and some analysts in the Convenie team had already been headhunted by competitors and little difference had it made. Olivia could leak it to the press, but in a culture that fornicated over Big Data there was a good chance that it would just be celebrated in an article for Harvard Business Review.

Olivia had created a monster not made of flesh and bone, but one constructed from a few lines of code. A society destroying specter.

It was Monday morning and Olivia crawled into work after a sleepless weekend. Reuben wasn’t in. ‘Probably pulling a sickie,’ she thought to herself. Then all of a sudden she felt very unhappy as she played the movie back of his reaction at the store.

She started the typical morning procedure of going through 100 odd emails. At the top was one from Simon:


Finance say growth stepped down from 29% to 24% this quarter. Need to continue to maximise returns on Process… new data shows key segments have reached saturation. Limited revenue growth… need to calibrate for KIDS – lock them in. Put it on the roadmap for next quarter.


Simon Bell

Chief Technology Officer | Convenie | 0330 123 4721

Olivia went from unhappy to disgusted to outraged. Convenie had maxed out vulnerable adults and so they were now trying to tap into a new segment – their children. Simon wanted Olivia to refit The Process algorithm to lock them in; a lifetime of frozen yogurt and stomach melting painkillers.

But, there was nothing Olivia could do. Would a journalist care? Probably not; Convenie’s advertising spend had tripled over the past 12 months, most of which went to the free press that employed these journalists.

6. (Sic)

Reuben slouched on the sofa, a joint in his mouth and a video game controller in his hands. He still stewed over the Convenie store: people who had almost nothing and because of clever data analytics, now had even less.

He emailed the editor of an online blog popular in the analytics community. Reuben thought would be interested in the story and she was, but not in the way he expected:

I LOOOOVE IT! Can we interview the analyst who designed it? Sounds like a candidate for our algo of the month.

It was a paradox. Those who understood Olivia’s creation appreciated the ingenuity of it but couldn’t comprehend the detrimental human impact. All those trapped by The Process wouldn’t understand the evil genius behind it.

The Process was maintained by the most traditional of social structures: the class system.

Despondent, Reuben picked himself up from the sofa and walked to the local corner shop – Star Stores – one of Convenie’s competitors.

Browsing lunch in the Food to Go aisle, he heard his phone ping that signaled an incoming email.

That editor, chasing me for Olivia’s contact details no doubt, he thought.


‘Oh fuck off would you,’ he said to himself – out loud. A couple of fellow shoppers gave Reuben the kind of look someone might give to a talkative vagrant from the street.

He paused. Something didn’t read correctly. How did the direct marketing guys manage to misspell Exclusive. All messaging, as Reuben knew, had to go through multiple quality control checks before being dissemnated out to ConvenieCard customers.

He pressed the email icon.


Finance say growth stepped down from 29% to 24% this quarter. Need to continue to maximise returns on Process… new data shows key segments have reached saturation. Limited revenue growth… need to calibrate for KIDS – lock them in. Put it on the roadmap for next quarter.


Simon Bell

Chief Technology Officer | Convenie | 0330 123 4721

Reuben stood there opened mouth.The open mouth turned into a large grin.


Reuben glowed. With a mental fist pump and a skip in his step, he continued looking down the Food to Go aisle for lunch. The red tag caught his eye.

Sweet, a meal deal on the sandwiches!

Colin lives in the sleepy village of Sawbridgeworth, on the Essex/Hertfordshire border outside London. He works in data and analytics, which informs some of the themes he writes about. Colin hadn’t written fiction since he was a teenager – but decided to rekindle his love for short story writing three months ago.

%d bloggers like this: