The chief data officer bringing machine learning to Coronation Street

Machine learn­ing and Coro­na­tion Street. The com­bi­na­tion sounds like some­thing from a hare-brained scheme that you might over­hear Steve McDon­ald, Weatherfield’s res­i­dent dodgy deal­er, dream­ing up at The Rovers Return. But not only is it real; it’s a point­er to the future of broad­cast­ing – and we can prob­a­bly expect con­tex­tu­al adver­tis­ing to find its way into the ad breaks served up by any com­mer­cial TV chan­nel.

Hav­ing start­ed in 1960, Cor­rie is the world’s longest-run­ning TV soap, so the prospect of an AI-dri­ven Weath­er­field rep­re­sents a curi­ous jux­ta­po­si­tion of tra­di­tion and moder­ni­ty. Not with­out rea­son might cyn­ics sug­gest that such a mod­el would her­ald a mar­ket­ing-dri­ven era in which com­put­ers are in cre­ative con­trol and on a sin­gu­lar mis­sion to mon­e­tise at all costs.

But, accord­ing to ITV’s group chief data and AI offi­cer, San­jee­van Bala, that won’t be the case – at least on his net­work. He acknowl­edges that there is a debate about “whether machine algo­rithms are going to decide what we watch or not. Speak­ing from ITV’s per­spec­tive, it’s def­i­nite­ly not the direc­tion we’re tak­ing.”

Hav­ing worked as a con­sul­tant and strate­gist across a range of indus­tries, includ­ing a four-year stint in mar­ket­ing with Tesco Club­card pio­neer Dunnhum­by, Bala even­tu­al­ly set­tled into broad­cast­ing in 2011, becom­ing head of plan­ning and ana­lyt­ics at Chan­nel 4. He was respon­si­ble for a huge growth in the amount of first-par­ty data received by the net­work through view­er reg­is­tra­tions. 

He joined ITV in May 2020 – and not a moment too soon, as the UK’s first Covid lock­down was spurring a huge surge in tele­vi­sion watch­ing. He has since over­seen a range of data-led pilot projects, includ­ing a dri­ve to obtain more infor­ma­tion on view­ers via reg­is­tra­tions, just as he had achieved at Chan­nel 4. This first-par­ty data is set to be cru­cial in bring­ing con­tex­tu­al adver­tis­ing to life in a way that’s effec­tive but sub­tle enough to avoid sul­ly­ing the char­ac­ter of some of the nation’s clas­sic pro­grammes.

Bala recalls that ITV under­stood “very quick­ly” that it had to retain the best ele­ments of its pro­gram­ming – the char­ac­ter of the shows, the cre­ative com­po­nents and the human sto­ries – and com­bine these with the lat­est tech. While art and sci­ence can seem like polar oppo­sites – as was the case when he arranged one dis­as­trous meet­ing between a pro­gram­mer and a data sci­en­tist in his pre­vi­ous job – it doesn’t have to be that way.

“We often talk about this aug­ment­ed world where it’s the best of both,” he says. “It’s about how we cre­ate tools and under­stand­ing and com­bine these with edi­to­r­i­al, com­mer­cial and dis­tri­b­u­tion.”

Here’s how it works in the case of Coro­na­tion Street: inter­nal com­mer­cial team ITV AdLabs, cre­at­ed in 2021, has built machine learn­ing sys­tems that, in effect, watch and lis­ten to episodes. Their algo­rithms will then recog­nise con­tex­tu­al­ly rel­e­vant moments. So if, say, food fea­tures promi­nent­ly in a par­tic­u­lar episode, then rel­e­vant adverts could more eas­i­ly be served up – by a food-box sub­scrip­tion ser­vice, per­haps.

“From an adver­tis­ing per­spec­tive, the met­rics are real­ly strong,” says Bala, who holds AdLabs’ work as a “great exam­ple” of where ITV is bring­ing togeth­er art, sci­ence and cre­ative sto­ry­telling. “The thing that grates with con­sumers the most is when adver­tis­ing isn’t rel­e­vant,” he argues.

This is just one exam­ple, of course. Under Bala’s guid­ance, ITV is tak­ing an “iter­a­tive” approach where it demon­strates the val­ue of data to one organ­i­sa­tion­al func­tion and then uses its suc­cess­es as a way to make oth­ers more recep­tive to the con­cept. For instance, soon after he joined, his team start­ed tar­get­ing new view­ers on social media with pro­mos for shows – peo­ple who may not tra­di­tion­al­ly have tuned into ITV at all – and then showed the results of this ini­tia­tive to oth­er depart­ments.

The trick to gain­ing their sup­port for the new approach, Bala reck­ons, is to avoid talk­ing about tech wher­ev­er pos­si­ble. Busi­ness­es that focus too much on tech­nol­o­gy risk falling into a trap where they “end up with a faster horse” but fail to secure “real trans­for­ma­tion”, he says.

When Bala joined ITV, for exam­ple, the broad­cast­er already had in place tech­nol­o­gy com­po­nents such as Cro­cus, an ana­lyt­ics tool devel­oped in asso­ci­a­tion with Google Cloud Plat­form. But what it still need­ed, he felt, was an artic­u­la­tion of the group-wide mis­sion, which required him to explain the desired out­comes in the lan­guage of the peo­ple he was talk­ing to.

“At our core, we broad­cast­ers are cre­ative pro­duc­er-dis­trib­u­tors,” Bala says. “Our val­ue chain is about com­mis­sion­ing, acquir­ing and dis­trib­ut­ing con­tent. It’s then about pro­mot­ing that con­tent through mar­ket­ing and, last­ly, mon­etis­ing the con­tent through adver­tis­ing. When you talk to busi­ness lead­ers in each of those areas, you start to have con­ver­sa­tions about how data might inform what you com­mis­sion or acquire, or ask our­selves, say, what if we could use data to think about how to dis­trib­ute con­tent dif­fer­ent­ly?”

Over the next year to 18 months, ITV hopes to expand data ini­tia­tives into its mar­ket­ing, prod­uct and com­mer­cial teams, includ­ing the wider roll-out of con­tex­tu­al adver­tis­ing. With this trans­for­ma­tion already under way, the net­work has some bold ideas for what’s next. Last year, it invest­ed £2.5m in Metavi­sion, enabling it to expand its intel­lec­tu­al prop­er­ty into the meta­verse. In July it made its first for­ay into this vir­tu­al world by launch­ing a ver­sion of its gameshow The Void in the Fort­nite Cre­ative sand­box video game, for exam­ple.

To realise such ideas, Bala and his team are focus­ing on “fix­ing the last mile first”, which means ensur­ing that both busi­ness lead­ers and end users under­stand the val­ue in any projects that are under way, and then con­nect­ing every­thing up to the out­come, rather than wor­ry­ing about all the tech­ni­cal details from the get-go.

Sim­i­lar­ly, he feels that organ­i­sa­tions occa­sion­al­ly con­cen­trate too close­ly on achiev­ing data lit­er­a­cy as opposed to build­ing busi­ness lit­er­a­cy. “We avoid the term ‘data lit­er­a­cy’, because I think it can be slight­ly con­tentious, with its impli­ca­tion of illit­er­a­cy,” Bala says. “We do talk a lot about how we can upskill our col­leagues and improve their capa­bil­i­ties. And we’re mak­ing sure that we do it one team at a time.”

This has to be a two-way process, he stress­es. Data pro­fes­sion­als must under­stand how an organ­i­sa­tion oper­ates, what its cul­ture is, how infor­ma­tion is used and what choic­es can be made. “This is about how you mesh all those things togeth­er to then roll out a more rel­e­vant pro­gramme.”