Data analytics for pets

How is Mars Petcare using data to help dogs and cats?

Dogs and cats can’t talk, so they can’t tell us when they start to feel ill or oth­er­wise may need health­care. At Mars Pet­care, we are using data and ana­lyt­ics to reveal new insights into pet health. By iden­ti­fy­ing changes in health as soon as pos­si­ble, we hope to reduce the impact that dis­ease can have on pets’ well­be­ing. We already have diag­nos­tic tools that can help pre­dict the onset of dis­ease, tech­nol­o­gy that can detect changes in pet behav­iour that could sig­nal a change in well­be­ing and we are using lead­ing-edge data sci­ence appli­ca­tions to enable a new era of pre­ven­tive pet care.

And it’s not lim­it­ed to health­care. There are sev­er­al areas of our busi­ness where we’re using data sci­ence, from opti­mis­ing sup­ply chain and hos­pi­tal oper­a­tions to per­son­alised mar­ket­ing and retail assort­ment.

We are posi­tioned to col­lect infor­ma­tion on the behav­iour, health and genet­ics of pets around the world through our nutri­tion, health and pet ser­vice busi­ness­es – and we want to use that priv­i­lege to be a force for good. Mars Pet­care has a num­ber of pet food brands, includ­ing Pedi­gree™ and Whiskas™, as well as Roy­al Canin®, our health nutri­tion brand. We’re also very proud of our glob­al net­work of more than 2,500 vet­eri­nary hos­pi­tals and diag­nos­tic labs that employ more than 65,000 com­pas­sion­ate vet­eri­nary pro­fes­sion­als.

We’ve got a DNA busi­ness called Wis­dom Pan­el™, while Whis­tle™ helps us under­stand pet health through activ­i­ty and behav­iour track­ing. Sci­en­tists at the Waltham Pet­care Sci­ence Insti­tute work to advance sci­ence to bet­ter under­stand pets’ needs and over­all well­be­ing. The scale and gran­u­lar­i­ty of this data across our busi­ness­es pro­vide a phe­nom­e­nal oppor­tu­ni­ty to gen­er­ate insights that will help enable pets to live hap­pi­er, health­i­er lives.

What is your latest breakthrough?

Pet Insight Project is a pio­neer­ing mis­sion to col­lect and inter­pret the health sto­ries of hun­dreds of thou­sands of dogs. It’s enabled by Whis­tle FIT, a smart, col­lar-worn device that cap­tures infor­ma­tion about a dog’s behav­iour – like a fit­ness track­er for dogs. It has a three-axis accelerom­e­ter to sense how much a pet moves up to 150 times a sec­ond.

Pet Insight Project invites cit­i­zen sci­en­tists to send in videos of their dogs doing cer­tain activ­i­ties, such as eat­ing, drink­ing and scratch­ing, whilst wear­ing the Whis­tle device. By pair­ing the data from the device with the video footage, we have built ML algo­rithms that can detect when dogs are exhibit­ing cer­tain behav­iours. We can now detect pre­vi­ous­ly unseen behav­iour pat­terns that could be ear­ly indi­ca­tors of changes in well­be­ing.

For exam­ple, an increase in the amount of time a dog spends drink­ing each day could be a sign of ill health, such as dia­betes or kid­ney dis­ease. We can noti­fy the pet own­er of these changes through the Whis­tle app, often before they have noticed the change in behav­iour them­selves.

How does your data help vets?

Pro­fes­sion wide, it’s esti­mat­ed that the anaes­thet­ic mor­tal­i­ty rate is twice as high for cats as it is for dogs. We applied Nat­ur­al Lan­guage Pro­cess­ing to unstruc­tured med­ical notes and processed data from 1 bil­lion raw med­ical notes, cre­at­ing 300,000 struc­tured anaes­the­sia records. Our research to try and bet­ter under­stand the risks asso­ci­at­ed with feline anaes­the­sia mor­tal­i­ty has iden­ti­fied six unique risk fac­tors. The next stage of this project is now in devel­op­ment — a pre­dic­tive algo­rithm that can pro­vide real-time feed­back to vet­eri­nary teams to help make anaes­thet­ic pro­ce­dures safer for cats.

Tell us about your work in AI and ML

Our AI and ML algo­rithms are writ­ten most­ly in-house, but we also part­ner with third-par­ties and aca­d­e­m­ic insti­tu­tions. Anoth­er great exam­ple is our appli­ca­tion of ML to X‑rays. There is a dearth of vet­eri­nary trained radi­ol­o­gists glob­al­ly, and a huge year on year increase in demand for radi­o­log­i­cal reads by vet­eri­nar­i­ans. Using 2.2 mil­lion X‑rays, we have built ML mod­els that can iden­ti­fy 42 find­ings in tho­rax and abdom­i­nal X‑rays in cats and dogs, with many more find­ings to come. We can now pro­vide a ser­vice to vet­eri­nary hos­pi­tals where ML can be used as an ini­tial assess­ment for an X‑ray.

We’re also using pre­dic­tive mod­els to opti­mise our retail assort­ment, so that we can pro­vide pet own­ers with more tai­lored prod­uct offer­ings.

How can data help in early disease detection?

Chron­ic kid­ney dis­ease (CKD) is the num­ber one cause of death for cats over five and affects 30% to 40% of all cats over the age of 10. Renal­Tech™ is an AI-dri­ven diag­nos­tic tool that can pre­dict whether a cat will devel­op CKD up to two years before tra­di­tion­al clin­i­cal diag­no­sis. It was devel­oped by data sci­en­tists at Waltham Pet­care Sci­ence Insti­tute using the med­ical data of more than 150,000 cats and 750,000 vet­eri­nary vis­its over 20 years. It’s avail­able to vet­eri­nar­i­ans through our diag­nos­tic lab Antech’s net­work. By hav­ing an ear­ly insight to the onset of dis­ease, pet own­ers and vet­eri­nar­i­ans can part­ner to cre­ate a proac­tive, per­son­alised care plan for the pet.

What does the future look like for pet care?

A com­mon use of AI in human health is the devel­op­ment of pre­dic­tive mod­els: where a com­put­er can spot very sub­tle pat­terns in data that are an ear­ly indi­ca­tor of a health con­di­tion or dis­ease. Because so many dis­eases have bet­ter out­comes if they are picked up and treat­ed ear­li­er, pre­dic­tive mod­els have the poten­tial to be trans­for­ma­tive for pet health­care, as in human health­care.

We believe that going for­ward, we’ll con­tin­ue to tap into the poten­tial of new tech­nolo­gies and tools to dri­ve pre­ven­tive and more per­son­alised care, which can allow us to work hard­er and smarter to improve the lives of our pets. Already sci­en­tists are using the pow­er of AI to not only under­stand how a pet’s genet­ics and behav­iour can share pow­er­ful insights about their over­all health, but to also help pre­dict dis­ease in pets. Indi­vid­u­alised nutri­tion­al sup­port may ulti­mate­ly help in the pre­ven­tion and treat­ment of cer­tain dis­eases. In the future, vet­eri­nar­i­ans and pet own­ers will be able to part­ner even bet­ter to keep pets healthy and hap­py.

What motivates your team?

Pets make our lives bet­ter, and we’re priv­i­leged to have the oppor­tu­ni­ty to use sci­ence and data to make their lives bet­ter too. By respon­si­bly har­ness­ing pet data, we have the oppor­tu­ni­ty to change the lives of mil­lions of pets and the pet care indus­try. What moti­vates and unites us is our pur­pose: A bet­ter world for pets.

To find out more vis­it mars.com/made-by-mars/petcare

Spon­sored by

Mars Pet­care