We exchanged emails for this interview and it has been slightly condensed and edited.
Each of you please tell us your story on how you arrived at Chalice. And of course, ladies first.
Ali: I spent a decade in Big Tech – Google and Snap – because Adam wouldn’t date someone he worked with (we met at Razorfish, he told me I needed a new job if he were to date me, which is how I ended up at Google). So it’s ironic that Adam, now my husband, stealthily recruited me to be his COO by low-key asking me to help with a 5 year revenue plan, workshop the pitch deck, etc. Before I knew it, I was so invested in Chalice that I was COO.
One of the reasons I became so captivated with the idea for Chalice is what I saw inside the walls of Big Tech – the power of AI to transform the advertising industry. Chalice is part of the democratization of that – advertisers can rewrite the rules of advertising AI according to their own business rules.
Adam: I was an agency media exec and became aware of in-house teams customizing their approach to digital display and getting better results than any agency could deliver. It was frustrating how hard it was for agencies to take even the initial step toward customization, which is to store all the ad delivery logs and make them accessible to the advertisers. The book “Competing in the Age of AI,” which I read in proof form in late 2019, inspired me to try and bring Chalice to life. The book makes a strong case that the ability to use data and build predictive models is becoming mission-critical for business success.
Ken: I have spent most of my career with at least one toe (and sometimes both feet) in digital advertising. Most recently, I had taken a job as a partner in a consulting firm where my practice was focused on use of data and analytics. As of last March, the practice was suddenly shut down and I had to scramble a bit to make ends meet. As part of my scramble, I started doing some independent consulting and a mutual acquaintance…YOU Marc Goldberg put Adam and I in touch. Adam laid out his vision for Chalice and I liked his approach, so I joined on, first as a consultant and then as a co-founder. Interesting fact, we have never met face to face.
So Adam does not date co-workers (good strategy) but marries them. Ken I helped you meet them and you joined, that is great! So what is Chalice?
Ali: Chalice is a company that puts the power of AI directly in the hands of advertisers. When advertisers can write the rules on how their ads reach consumers, the connection is much richer. Mostly, because advertisers can write those rules to their true business outcomes instead of silly marketing metrics.
Background on the name. Adam kept coming to me with ideas that sounded so Ad Tech 2015. I encouraged him to think of his own roots – he started his career as a music journalist, mostly in hip hop. I told him that was much cooler than anything related to ad tech and suggested he think of a way to connect the two. Then we learned Ken was a sound engineer who toured with hip hop and rock groups, before getting his PhD. We liked how the Chalice is a hip hop icon, but also a nod to advertisers always chasing “the holy grail”. We’d say, “Choose wisely.”
Hip Hop! We will talk about it at the end, one of my favorite topics. Oh Algorithms. A previous interview joked that a lot of folks want to get into “custom algorithms and still don’t know how to log into their DSP. Can you tell us some of the moving parts to your approach?
Ali: They should be happy because they’ll barely need to log in, once their custom algo is running. Seriously, the algorithm is going to do 80% of the work that humans normally have to fiddle with in a DSP. That should be exciting, not threatening to marketers – they can focus on strategy and setting direction (including the outcomes their algorithms should be directed to) and leave the machine learning to triangulate bids, reports and optimizations.
Adam: While of course some advertisers are more advanced than others, everyone including us has a lot to learn. So it can be misleading to equate tech chops with readiness. What makes one ready is willingness to find out what the most advanced technology can do for your business results. Some display newcomers are ready. Some practitioners who were advanced in 2018 are afraid of what’s coming. Their first experience of custom algorithms will be competing with them, losing badly to outfits they considered technologically hopeless. By making the most advanced techniques immediately available on a SaaS basis, Chalice is changing the game.
In any other field, it’d be entirely non-controversial to say: “If you license a platform to execute millions of transactions per day, you should capture, store and use all that transaction data to make predictions and improve your rate of return.” That’s table stakes for competing in the age of AI. It’s only in advertising that the investment required to ante up was treated like a big risk. And now that the investment required is absolutely miniscule, compared to total ad budgets (under 3% in most cases), it’s hard to say anyone not yet testing their own algorithms has any claim to being advanced. As Ali said, you don’t need to log into anything! You need a vague understanding of what inevitably happens in competitive markets after big data becomes accessible.
Ken: Our approach on developing custom algorithms is automated until deployment into a DSP. We start with optimizing prices based on bid stream data and ad quality metrics (and in some cases we have been able to use your MMI quality metrics!). We call this supply modeling as it is not based on anything that we uniquely know about a user. After we get a supply model in place, we enrich clients conversion data with first- or second-party data, and/or a third party data source that provides a variety of individual level and household level factors including demographics, visits to physical stores, political affiliation, etc. Typically, we will build models using logistic regression, random forest, and deep learning techniques. The results from the three models are then scored at the user level, assessed for accuracy, and then the best performing model is pushed to the clients DSP of choice. The exact method of activating a model in a DSP depends on the capabilities of the DSP and we have spent a fair bit of effort figuring out the best way to work within the constraints of a given DSP.
Algorithms require data and with so much data coming from bots from data centers, how does that impact your data?
Adam: We avoid the biggest bad-signal problem by optimizing to real business performance, like incremental customer acquisition, instead of generic KPIs like last-touch tag fires. Optimizing to the bottom line makes it relatively easy for us to notice and correct when one of our model features isn’t consistent. For example, if lift is coming from reach to young women and our targeting data fails, we’ll know something’s wrong because lift will go down!
There’s a lot of human detective work to be done when predictions fail — Did we lose accuracy or did consumer behavior change? It’s definitely best to start with reliable partners, and to cast a suspicious eye on sources you haven’t used before.
It can be tough to get all cylinders firing on a custom algorithm, but remember most advertisers aren’t even trying to optimize for incremental lift or lifetime value. I watched a dozen retailers go out of business while their “performance media” team was claiming 5:1 ROI or more. All the way to Chapter 11! Data quality doesn’t get much worse than that. (In case it’s not clear, I’m talking about the Google Marketing Platform.)
Adam I remember speaking to you prior to the Presidential election and saw you posted something on Linkedin that both alluded to you having a sense of the outcome. Can you elaborate?
Adam: Yes, in our work for both the Ossoff and Biden campaigns, we had significant data that gave us confidence in the outcome. For Biden it was motivation among women in the Southeast suburbs (I thought Biden was going to win North Carolina as well as Georgia). For Ossoff, it came through as lift among under 25s who had not voted in November, if exposed to the ad we were running. Not huge, but enough to make a difference. And overall, it was like 10% intent to vote for Ossoff with zero exposures and 40% with only one. That creative really hit the spot! We love paying off the promise of a great ad in any case, but especially that one.
What are some of the insights you can provide to candidates and their teams from the data?
Adam: If you conduct good research on a message that should move the needle among voters, it’s definitely a mistake to hand that off to Facebook with some demo settings. It’s an unforced error in the last mile. Ad delivery must be specific to what you’re trying to accomplish, just like creative. Facebook will always over deliver to your base, because the entire platform is optimized for engagement. Changing people’s minds is a different endeavor, and Facebook is obviously, famously indifferent toward it.
How did you get into political space?
Ali: We’re really not a political firm, so in some ways it was a happy accident. We’d developed a solution for incremental brand lift that we were talking to brands about using for persuasion campaigns – for example, getting someone to try a new product or reconsider a brand – and realized it would be very powerful in a political campaign. We pitched and emailed wide, and kept hearing we were too late in the cycle and too “break the mold” generally. We figured we’d have to wait until the midterms, then someone on the Biden side heard about us and it captured his imagination. This was in early October. We turned the whole thing around in under a week. It was the best kind of madness.
Is political advertising going to be more difficult with cookies and privacy issues?
Adam: We don’t mind when things get more difficult, as long as they’re fairly competitive.
2020 was a fun year to start a company, how do you see 21 shaping up?
Adam: I’ve been wondering if we’ll ever sort of miss 2020. Maybe when the subway is crowded again?
Ali: Everyone seemed in shock in 2020, and the shockwaves seemed to just crest over and over. So while major disruption was happening, I think most businesses were paralyzed a bit by the shock, even if the need to change was the clearest it’s been.
We’re already seeing that businesses are shifting. Integrating all that 2020 taught them into a total rethink. Our convos have turned from “this is really compelling” to “how soon do I need to start to get a first mover advantage?”. I bet it’s happening across business in all areas, not just advertising. When people see they can live in a new reality, and adapt to it quickly (and endure the pains of transition), stuff starts to get a lot more wide open.
Talk about how your story is resonating “out spend outsmart?
Ali: Businesses have seen digital ad budgets grow and grow, and many aren’t seeing the returns grow at the same pace. But digital ad companies keep claiming better and better results. Something isn’t adding up. Savvy businesses are starting to realize that a lot of what advertising is getting credited with driving is not incremental to the business, just taking credit for what was coming to the business through other avenues. At the core of that problem – platform algorithms have been rewarded for getting credit for non-incremental stuff (like showing you an ad for something you just purchased). The only way around that is to retrain the algorithm to incremental lift.
Adam: It speaks to the mindset of brand-side marketers trying to innovate. They look at the agency-of-record service model, with its linear process, as antiquated. They’re looking for something else. The media assembly line goes wrong at Step One: “Let’s choose our audience.” Clients ask us to identify audiences they might want to build creative for.
Are you funded? Are you looking for more?
Adam: We bootstrapped Year One and are now doing a small Seed round.
Ali: it’s a no-brainer investment opportunity, if anyone wants to throw down.
Are you remote or in an office now?
Adam: Me and Ali are at home in Brooklyn with two toddlers.
Ken: At home in Atlanta, but looking forward to be able to get back in an office.
Assuming everything gets better. Ally and Adam, where are you going first.
Adam: I want to take our kids to Japan. I think they’d find it extremely fun. Also, I drank so much coffee this year, I feel like I should see Ethiopia.
Ali: I would like to NOT take our kids to Jamaica. Or maybe, take them for 3 days then send them home so we can have 5 days of no noise except the ocean.
I love that Adam wants to take our kids to a country where most spaces are smaller and noisier than our Brooklyn apartment. It’s what makes him a great dad, he’s tireless. I guess that’s what makes him a great CEO, too.
Ken where are you going? You can say to Brooklyn and stay in their apartment when they are gone in Japan or Jamaica. I will go drinking wichita!
I am a foodie and we have some great food experiences in Atlanta. I have a list of places that I want to get back to including Ticonderoga Club for drinks and Gunshow for dinner. And then Bachinallia for another dinner. And Watchman. The list is pretty long. My wife and I also host fairly regular dinner parties where we are pretty daring in our menu. I am looking forward to being able to host folks again. In terms of locations, I have owed my eldest a trip to France. I am ready to pay up, but he is now suggesting New Orleans. I would bet that NOLA is our first stop, once we can travel. And I owe you a drink for sure.
So Adam was a hip hop music journalist? Top 3 of all time?
Adam: MF Doom, RZA, Public Enemy. I interviewed all 3. (The news of Doom’s death ruined my New Year’s, as I expect was the case for all his fans.) Ken has more street cred than me, though — he was a soundman on the Raising Hell tour. Chalice is not about a salary, it’s all about reality.
Ken: For those I worked with, Billy Joel was the best. PE gets the nod. Oh, and the Goo Goo Dolls. I recorded part of their first album.
Hey Ali, we know your top three are Adam, Kid 1 Kid 2, no particular order. Adam, RIP Zev Love X (Doom).