Data & AI Risk Management and Ethics ♦ Gradient Institute Fellow ♦ 2022 Actuary of the Year
This landed in my inbox earlier. Sharing as I think it’s instructive for data science folks doing targeted marketing. I'm a very, very infrequent user of Uber. I downloaded the app in a panic in Adelaide in 2016, when I urgently needed a cab to the airport and couldn't find one. Since then I've had (checks records...) a total of 25 uber rides. In 8 years. Here I’m being offered 20% off my next 20 rides, before 1 April! So, what's the problem? First: it looks really silly, and looks like you don’t know anything about me. I’d probably react better to a discount off a single ride, or a discount off a second trip, as it better reflects my actual usage patterns. It would also be a lower value offer. A small-ish discount off 20 rides in a week just looks very daft and not well targeted. It tells me you didn’t even bother. Second – it won’t work. Will this change my chance of using uber this week? No - I'm a rare user at the best of times. Would any ad be successful? Realistically, only something of significant value might be enough to change my behaviour. 20% off is probably just waste – if I was already planning a trip it’s now cheaper, but it won’t convince me to take more. The lesson: some important questions that data science folks can ask themselves when designing targeted offers: - Is the audience well matched to the creative? Will it resonate well? - Is the creative likely to change the behaviour of the audience? If the answer to either is: ‘no’, for some obvious section of your audience that you can easily identify in the data you’re holding, then it probably needs refinement!
Stuart Nickols
5mo
Could it be that it worked on people just like you last week? ie You have "rationale" but Uber has data.
Ganesh Shenoy
5mo
Having worked in digital marketing, I can say a couple of things about this marketing offer that are not obvious at get go. Many times an offer is sent out to remind the potential customer about the brand. The customer doesn't take the offer? Good! But a customer who is offered 20% generous discount is likely to not unsubscribe, and in the future is likely to think of Uber before the local taxi. This one looks like the campaign that came out of a conscious decision to send out generous offers to customers who are the most unlikely to take them!
....and the counterpoint is, Chris Dolman does it feel good that they don't seem to be examining your personal useage patterns?
Stuart Nickols
5mo
Actually.... What if their ML has noticed a pattern whereby people like you get offended by inaccurate email marketing and share it onto LinkedIn, and it is working on second order impression benefits?
Zhang Jack
5mo
Hi Chris, long time no see. I suspect operationally the backend systems can only handle one or a few promotions at a time, but with AI it’s possible that targeted individualised promos won’t be too far away. Definitely something to watch out for.
Julia Lessing
5mo
I wonder how many Coasties regularly use Uber, let alone 20 in a week?!
Matt Francis
5mo
I think assuming that Data Science was involved in the targeting decision, even at a company like Uber, is brave! Otherwise the message is on point.
I think that’s part of my frustration with discount use/ML optimisation, in large part it’s probably going to the most marginal/least connected user and then due to a “tag” classified as a win because out of the sea of people who don’t engage that they have details for, they’ve created an engagement. As you say, it’s more likely that they’ve timed a “going to happen anyway” user, but success is measured on click through a to a base case. It’s about the Ill considered counterfactual right!
Honey Muir
5mo
This! I get these unpersonalised promo offers all the time. I don’t even have the app anymore and haven’t frequently used Uber since before Covid. The most recent offer was titled “let’s get you back on the road” & I thought - yes finally some click bait - make it about me!
Mini MBA in Marketing - Personalised Marketing Lead at IAG
5moWould have to respectfully disagree Chris. - this would have cost Uber literally $0 to email this offer to you. - whilst it doesn’t sound like you are the perfect target audience, the offer isn’t that bad or misplaced that it would make you go out of your way to not take an Uber. - at a bare minimum Uber and this offer just might be more likely to come to mind if you need a ride in the next week. In summary: I think one of the biggest traps we can fall into is limiting your audience too much, or spending huge amounts of additional effort trying to craft the perfect message for the perfect audience. I’d rather a 2% conversion on an audience of 100k vs 10% conversion on an audience of 5k.