In this interview, we talk to Chris Cummings, CEO of Curiosity Media and Mike Miller, Director of Revenue and Analytics at Curiosity Media’s SpanishDict.com. Curiosity Media are the people behind Spanish learning websites, SpanishDict and Fluencia which collectively help more than 100 million people each year.
Hi Chris and Mike, I’ve been looking forward to this interview. SpanishDict was a header bidding early adopter and quickly developed sophisticated optimization strategies to maximize your monetary performance. Thanks for sharing your knowledge with our community. Let’s get started!
CH: Looking back on 2016, which monetization strategies worked well, and which didn’t? How would you change things going forward?
CM: The single biggest impact on our revenue in 2016 was expanding the use of header bidding. We went from just a couple partners to around 15 active header bidders, and as we added more demand to the site, prices went up. We saw that more competition is better for publishers; people have said that in the abstracts for years. With header bidding, we had a very strong bottom line impact in 2016. So that’s the most important initiative that we had for 2016.
In terms of what we tried and what didn’t work, we actually run experiments constantly on our site to try to determine what will add revenue while preserving a good user experience. And, a lot of those experiments didn’t work.
We found that while many of the header bidders that we added provide incremental value, a number of them we tested were removed because they did not provide any value.
They may have accounted for some revenue, but the corresponding slow down in the auction or slow down of the site may have negated that revenue increase.
We also tried a number of price flooring strategies with a lot of different partners. Some of them helped. Others didn’t. We tried moving people into Prebid and moving people out of Prebid. Most of the time it was better with them inside Prebid, but sometimes it wasn’t. So we still have a few people that are not in Prebid.
In the absence of these fast response times, we’ve experimented with different flooring strategies, and we’ve seen some good results from trying to set price floors that reflect what we think the impression will be worth. Every partner has different price flooring capabilities. Some you have to set it within the UI. Some you can do with real-time floors based on the API. And so we actually have to tailor the price flooring strategy to each partner based on what their capabilities are.
CH: Have you found a dropping off point with regards to latency based upon the number of partners you bring in?
CM: I haven’t thought about that in terms of too many partners or too few partners. We look at it more in terms of the incremental impact of any given partner. And so if adding one more partner causes latency to go up, and that last incremental partner is not worth it, then we would drop them. But we are not philosophical about the number of partners that we run. And the impact on the site can vary tremendously from partner to partner. Some partners load a lot of Javascript, gather a lot of custom data, and then only after doing that, submit their bid. Other partners are much faster and easier to work with. To Sovrn’s credit, you guys have provided a great API that makes it very easy to request a bid. It’s very lightweight. We can bundle the bid request into one request rather than having to send out individual requests for each adding on the page. And so requesting a bid from Sovrn is fast and easy.
For others, it’s more difficult. But to get back to the question of how many partners is the right amount or too many. Since the concern is about the impact on the page and the impact on the auction, each partner actually has very different profiles in terms of how much they impact the page, how much they impact the auction. So just adding one bad partner could be worse than adding 10 extremely good partners. No exaggeration.
CH: We talked about header bidding, it’s a been big growth opportunity for publishers in 2016. In order to keep a revenue growth trajectory, what other strategies are you going to use in 2017?
CM: I think that the number of companies that want to compete in the header is going to continue to grow. We’ve already seen some agencies and marketers want to get into the header directly. I think that there will continue to be more demand on the marketer side to be in the header on publishers, and that will drive growth.
I’m also hopeful that improvements in header bidding technology will allow the auctions to run faster, the bids to return quicker, and the ads to render quicker. And then that will lead to revenue gains as well. We’ve seen that if an auction takes a couple seconds to run and then it takes another couple seconds for the ad to load, by the time the ad’s rendered on the site, we could have lost 5% to 10% of our revenue because the ad was slow to load. I’d really like to gain that 5% to 10% of our revenue back this year by having ads that are fast.
CH: Do you think S2S is a technology that will compensate for that?
CM: I think there are great aspects of server-to-server integrations that can help reduce the page load and provide faster bid response times. But the coordination challenge of getting different players in the industry to integrate with each other is huge. And everybody is skeptical of everybody else’s server-to-server approach because no matter how many times they say it’s going to be fair and even, it’s not as equal as if you’re a header bidder. And at the same time, it’s actually not that bad to run header bidding in the publisher’s site if you have fast bid request and bid responses.
I would love to see server-to-server integrations, where everybody bids quickly. It lightens the page load, and we get all our partners on there. I just think that’s going to take a long time for it to materialize. And in the meantime, we don’t want to leave all the revenue on the table that could be had from just having better header bidding experience here and now.
CH: Let’s talk about PubNation since I heard from other publishers that you created the ad quality solution from frustrations you had with slow and bad ads impacting your users experience. What specific concerns do you see right now in terms of the improving the customer’s experience and reducing or eliminating the bad ads and slow ads?
CM: Yes, these problems are front and center for publishers. We continue to see video ads coming through banners. We see display ads that download hundreds of requests and take up 5 to 10 megabytes of bandwidth just to show the display ad. We see ads that redirect users or point them to scammy, fake news websites. Programmatic ads are a tremendous boon for publisher revenue, but they have created more ad quality problems than ever before.
One of the reasons why I think we’ve been able to expand so rapidly with programmatic and with header bidding is because I believe we have better ad quality controls than most of the industry. Those two go hand-in-hand. We could not have achieved our revenue growth if we did not have the ad quality control processes that we do. We’ve shut off partners because of ad quality problems or significantly reduce the amount of business we do with partners because of ad quality problems.
On the flip side, for some partners, we’ve done a lot more business because they have such good ad quality. Since we now have visibility and control over what’s appearing on the site, it makes us much more willing to try out that incremental header bidder because if they have ad quality problems, we’ll know about it. We can take them down or take action on them.
CH: If I’m a publisher that’s having these issues, what are two or three ways that I can alleviate these issues?
CC: Use PubNation is a great answer, but more generally, I think it’s very difficult to manage what you can’t measure. Whether you use PubNation or not, you need to get a tool in place that lets you measure ad quality. And there are a number of different types of ad quality problems that arise, things that slow down the page, redirect the users, play videos, and you need a way of identifying whether these problems are happening on the site. Once you can measure them, it’s much easier to take action on them. So that’s maybe the most important thing is measure it so you can manage it.
After measuring it, the next biggest challenge that we’ve seen is that for people that work with multiple sources of programmatic demand, it can be very challenging for them to know where an ad came from. There’s a lot of complexity in programmatic tags, and that complexity is only magnified once people move to header bidding. I think after measuring ad quality generally, the next step is to measure it specifically by partner so that you can attribute different problematic ads to their source.
The third thing is to have somebody who’s directly accountable for quality. I think right now, for a lot of publishers, it’s a game of finger pointing. The engineers are saying, “Oh, the site is slow,” and they point at the revenue team. Revenue team is saying, “We’re just trying to make money, and we have no visibility into what’s fast or slow. All we see are CPMs.” And so they feel like they have no power to address the quality problems. And I think that smart publishers are increasingly making somebody accountable for ad quality so that if there’s a problem, they know who to go to get resolution and to get help.
CH: Personally, I think it’s gonna get worse before it gets better. There are so many questionable sites and bad ads coming in from questionable sources, and programmatic is supposed to grow even more in the next couple of years.
Okay, one last question. At the Tech Forum last year, I talked to your team about their internal ad op processes. Could you recommend two to three things ad ops managers can do to streamline and improve their day-to-day chaos?
CM: Sure. I think number one is to automate reporting. The complexity of working with multiple programmatic partners and trying to pull reports about how much revenue, to answer you on a simple question, how much money did we make yesterday? For a lot of publishers, that’s like an hour-long analysis. And it’s so long that you just don’t do it. You do it on a weekly or monthly basis. We’ve seen over the last couple of years massive improvements in our ability to operate by changing that from an hour-long process to a four-second process. And for us, we’ve actually moved to running all of our programmatic through header bidding. For the most part, we don’t do any tag-based networks. And what that means is that when we gather information for most of our partners on what they’re bidding, we’re able to get real-time revenue about on the site. There are couple areas where we have to estimate or fill in some gaps, but by and large, we’re able to get a pretty good approximation of what’s happening in real time, and that’s been a game changer for us. When we run experiments, we can see what’s happening on the site. When we want to pull report on the revenue, it’s dead simple.
You can use a variety of tools to collect this data and report on this data. We built a way of storing the data in Amazon Redshift, which is a data warehouse. And then we used a business intelligence tool called Looker to run reports on top of it. We talked to a lot of publishers that have different ways of storing the data and reporting on the data. But I think that’s probably the most important thing is to automate the reporting, because again, if you can’t measure it, it’s difficult to manage it.
That’s been a huge time saver. But also, more importantly, it’s helped us drive more insights that helps us make more money. A lot of people talk about the cost of the reporting in terms of how long does it take for a person, an analyst to go pool the report and put together the data. And that’s only, you know, 20% of the value. Eighty percent of the value is in the insight you’re able to get by having real-time information and running experiments and seeing how it impacts revenue. So I think that’s probably number one thing to streamline operations.
Another one is that as we’ve added more partners. So if we run 15 header bidders, we probably work with 25, maybe 30 different partners over the course of the year. And keeping track of all the partners when we talk to them what the terms are, you know, payments, it’s just a lot to keep track of that many different vendors. Consequently, we’ve set up a database to keep track of everybody that we work with in all the key points of information for each of our vendors, and that’s been really helpful, too. We use a tool called SalesforceIQ.
The third is our emphasis on ad quality. Internally, we’ve had more emphasis on ad quality this year than ever before, and it’s been a huge enabler of expanding into that programmatic demand.
CH: Thank you Chris and Mike taking time to share you successes, processes and passion. I can’t wait for our community to learn from you in the upcoming AMA on February 1.
About Chris
Chris Cummings is the founder and CEO of Curiosity Media. He’s passionate about education, technology, entrepreneurship, and riding his electric skateboard to work.
You can follow him on Twitter: @christopherdc1.
About Mike
Mike Miller is the Director of Revenue and Analytics at SpanishDict.com. Before taking over this position, Mike was an engineer working on the company’s website and ad tech. In his current position, Mike works with a small team to manage all of the advertising on Spanishdict.com. All of the advertising revenue generated by the site comes from programmatic demand and the vast majority of that comes from the company’s 15 header bidding partners.
When I turned 30 this year I made a personal goal of learning how to play chess. Judging by the class I took, I might be the only person in the US older than 12 years old that is trying to do this.
