How do you maximize the ROI … on your ROI analysis? Every cross-platform provider builds up their best-in-class data and analytics, but many marketers feel they’re falling short.
Big data, in and of itself, is nothing special. Everyone and anyone can collect and aggregate data. But that doesn’t mean it’s providing value.
You might be reading this blog post on your mobile phone. But that’s not all. Did you check your weather app this morning? How about election coverage?
Deterministic cross-device matching is important for accurate attribution, because the average consumer switches between his or her 4.5 devices regularly, interacting across both apps and mobile websites. The more granular and accurate the user data, the better your attribution partner can follow your users along the path to conversion.
Convertro's 100% deterministic cross-device solution lets you follow your customer’s path across their mobile phones, tablets, and computers as they interact with your brand at work, home, and on the go. The result is a more accurate, rich understanding of how your media works across multiple devices to influence a customer’s purchase decision, which can help you make smarter investments with your advertising spend. The breadth of data to which Convertro has access, and the way in which we identify users across devices is why The Forrester Wave™: Marketing Measurement And Optimization Solutions, Q4 2016 gave Convertro the highest possible scores in Data Acquisition, Cross-Channel Identity, and Granularity Resolution.
Based on this report, published last month, we have had a few questions from clients and prospects around our robust cross-device matching capabilities and how we do it. Here, we will focus on the mobile app ad to mobile web connection.
The average person uses over 27 apps per month and sees 40+ marketing and non-marketing touchpoints on their way to conversion. Advertisers have greatly shifted budgets to mobile: In 2016 mobile ad spending was projected to account for 60.4% of digital ad spending by eMarketer.
So users are inevitably seeing a lot of mobile ads within apps they use, and clicking out on some. But technological limitations and restrictions (eg: cookie blocking) put in place by certain mobile browsers make it difficult to link the ads a user interacts with in a 3rd-party mobile app to actions they take afterward in their mobile web browser; even if on the same device.
Since the mobile app and mobile web browser function as two separate environments and are tracked by different methods, the reconciliation process of identifying users across the entire path to purchase can be tricky. Therefore, without a workaround to circumvent these limitations, it can appear that the in-app activity and the mobile web browser activity were performed by different users. Without the ability to accurately track a user's transition from an ad in an app to your mobile website, it can be difficult to accurately optimize your mobile media campaigns.
Fortunately, Convertro has a solution to effectively track this transition and ensure that in-app ads receive proper credit for out-of-app conversions.
- We insert a redirect between the in-app ad and the mobile web landing page
- The redirect captures the user’s device ID (IDFA/GAID) from the ad in the 3rd-party app and creates 1st-party and 3rd-party cookies for the mobile web browser that contain a unique Convertro ID for the user
- This establishes the Convertro cookie as a "trusted” 3rd party cookie in the mobile web browser, which ensures they are not blocked by iOS Safari browsers.
- We now can link the click trail based on the user's device ID to the click trail tracked with the Convertro cookie ID in the browser
- The advertising device ID (IDFA/GAID) and the Convertro browser cookie are linked together and added to Convertro's co-op cross-device pool
Don't worry; our Marketing Tag Builder does all the work for you. The steps are outlined in our newly revamped Knowledge Base. Got questions? If you are not an existing client, please contact us at email@example.com.
The opt-in shared or co-op device pool contains device links from all participating Convertro clients. This gives participating clients access to a massive, combined collection of deterministic cross-device links, which also includes linked device data from AOL, Verizon, and Millennial Media.
Alternatively, clients can store their device links in a private device pool managed by Convertro. Private device pools contain only the device links created from clients' own customer data and are not enhanced with device data from AOL, Verizon, and Millennial Media.
This solution respects the user's "Limit Ad Tracking" preference set in the device's privacy settings, as received from the publisher. If set to "true", Convertro does not track the device ID and instead creates a random Convertro user ID.
Sources: Convertro data, Measurement Is A Digital Buyer's Best Friend", Forrester Research, Inc. 2016 and Nielsen, Q4 2015
This post was written by Maryam Motamedi, Senior Product Marketing Manager.
Top independent research firm gives Convertro highest scores for ‘technology platform’ and ‘customer reference adoption of unified measurement’
We’re excited to announce that our Unified Marketing Attribution Platform (UMAP) has been recognized by Forrester Research in their new report titled, “The Forrester Wave™: Marketing Measurement And Optimization Solutions, Q4 2016.” Market measurement and optimization solutions help marketers gain valuable insight and visibility into how their budgets perform across various campaigns. Convertro was identified in the report as one of the “10 most significant providers.”
The report helps to educate the market around the many challenges marketers face with attribution. Today, most marketers use two separate models - Marketing Mix Modeling (MMM) and Multi-Touch Attribution (MTA) - to understand the true return on advertising investments and informative decision-making. However, because the models are fundamentally different, and were developed using different data sets and methodologies, the results are most often inaccurate or biased when combined and cobbled together.
Recently, the industry has moved to unify these two models. Having a unified model that exists within a single framework enables marketers to more effectively execute and optimize their efforts. In addition, it provides a consistent and more accurate set of metrics, drives more informed decisions for the allocation of marketing budgets across channels, and helps brands and marketers understand and act on the entire customer journey to achieve key objectives.
Convertro’s UMAP held the highest possible scores in several key criteria that support a marketer’s strategy in connecting with its customers:
HIGHEST-RANKED IN TECHNOLOGY PLATFORM CRITERION
Among evaluated vendors, Convertro’s UMAP also received the highest possible scores in the following criteria:
- Data acquisition
- Cross-channel identity
- Granularity resolution
HIGHEST-RANKED IN CUSTOMER REFERENCE ADOPTION OF UNIFIED MEASUREMENT CRITERION
Convertro also received the highest possible scores in the following criteria:
- Client satisfaction: user interface
- Client satisfaction: data normalization
We believe these scores reflect our commitment to delivering the most robust solution in our Unified Marketing Activation Platform (UMAP). Released one year ago, UMAP is an entirely new platform that seamlessly integrates MMM and MTA and attributes both online and offline ROI all the way down at the user-level.
This is critical as we live in an increasingly digital world, and more and more types of information are available across all channels. Our platform is open and flexible to handle inputs from wherever they come, both now and later. With the industry’s top data acquisition tools available through one platform, we are able to identify and customize insights to each consumer’s ‘web of influence’ across all of their devices and channels, today and well into the future.
It’s also important to note that having all of this information down to the user-level can be intimidating. This is why we have heavily invested in creating an intuitive interface that easily visualizes data that does all of the heavy lifting – for rapid execution on digital insights. We think Forrester recognized these capabilities, with the highest score in the technology platform category.
Beyond technology, however, marketers need to understand how to interpret all of these new, more granular insights, which is why we have continued to invest in education and training for our clients resulting in top scores in this category. With the highest possible client satisfaction: user interface rating, we think it’s working.
With our UMAP product, even after just a single year, AOL is poised to handle not only the convergence of channels that we see today, but also those we have not yet imagined for the future. We are very pleased with our performance in Forrester’s evaluation and will continue to build on our offering for our customers.
By Amy Mitchell, General Manager and Head of Convertro
Advertising influences all consumers; it serves to increase brand awareness and affinity for both those who are in-market and everyone else who could become customers in the future. However, the effect of advertising is not constant over time.
The traditional concept of Adstock describes how the response to an advertisement changes over time. Adstock effect describes how initially as a consumer is exposed to new ads the response increases until later when the campaign ends and the response starts to wear off.
The influence that a single piece of advertising has on a person’s purchase behavior is not constant and diminishes over time. You can use adstock models to approximate the rate at which an ad’s influence diminishes over time, but adstock models are often based on preset assumptions rather than data-driven observations.
The illustration below visualizes adstock for a scenario where the effect of advertising follows a radioactive half-life decay i.e. as each day passes the effect is halved.
The problem with adstock is that it attempts to “abstract” reality instead of reflect it. This disconnect from real-world data can lead you to make sub-optimal decisions on how to spend your media and marketing budget.
Convertro’s Data-Driven Approach To Time Decay
Instead of making a single, one-size-fits-all model of time decay, Convertro creates new time decay models based on the actual response data collected from people who have been exposed to your ads. This feature represents the real-world aspects of how advertising works and hence makes results more business-relevant. Specifically, attribution of conversions is now dependent on “time since exposure”, so recent exposures could see more revenue attributed to them than not-so-recent ones.
Because Convertro knows who was exposed to an ad and how long after that exposure they converted, Convertro’s time decay models reflect the true rate at which the influence of your ads diminishes over time. This leads to better understanding of how your media touchpoints actually perform, which in turns leads to more accurate insights into how to optimize your media spend.
Converto’s time decay models are client-, event- and channel-specific
Convertro creates customized time decay models that are unique to your organization and to the data collected from your own customers. Your time decay models are further customized by:
Event type - A different time decay model is created for each event you are tracking with Convertro. We use the actual converting clicktrail data for an event type to create the time decay model for that event type.
Media channel - Convertro takes into account that different channels have different rates of decay, so it models time-decay for TV spots differently than for display ads or emails, etc. To calculate each channel’s rate of decay, we look at the “channel type” of touchpoints within converting clicktrails to determine how long each type of ad remained influential after exposure.
Compare the Convertro time decay models below to the adstock model shown earlier in this article. Because they are created from actual user clicktrails, the dropoffs and plateaus in the Convertro models reflect real-world customer behavior much more accurately than the artificially smooth curve of the adstock model. This difference enables Convertro to provide more precise attribution.
Time-decay models for display touchpoints contributing to “sale” events in February and March
Impact on attribution
The clicktrails below illustrate the impact that time decay has on attribution. In these examples, a customer is exposed to 3 display ads between August 2015 and February 2016, then makes a $90 purchase on 2/17/2016.
Before accounting for time decay, the display touchpoints, being otherwise equal, are attributed equal credit for the conversion ($90).
Attribution for a $90 sale, before accounting for time decay
When time decay is accounted for, however, the amount attributed to each touchpoint becomes dependent on “time since exposure”, as represented in the time-decay model below for the event type, channel, and time frame.
Time-decay model for display touchpoints contributing to February “sale” events
You’ll see that after accounting for time decay, Convertro adjusts the amount of credit given to each touchpoint according to the time-decay model above. In this case, revenue attributed to the touchpoints nearest and furthest from the conversion date have shifted dramatically to reflect the relative impact each touchpoint had on the conversion according to the model.
Attribution for a $90 sale, after accounting for time decay
As you can see, accurately accounting for time decay can make a huge impact on attribution, and by extension, optimization. By including data-driven time decay as part of its algorithmic attribution process, Converto also increases the accuracy of its reports, spend recommendations, and optimization insights.
This means you, the marketer, will understand the impact of recency of exposure on attribution and use those learnings to space your advertising campaigns to achieve maximum effect.
Want to know more? If you are already working with us, contact your Client Services Manager for more information. If you are new to attribution and or Convertro, please email firstname.lastname@example.org to get started.
This post was written by Vish Oza, Director of Product Management, Jonathan Boswell, Senior Technical Writer, and Maryam Motamedi, Senior Product Marketing Manager.
Mashable recently reported that “ad fraud could become the second biggest organized crime enterprise behind the drug trade”. The Association of National Advertisers (ANA) estimates that advertisers will lose $7.2 Billion to bots globally in 2016. Over half way through the year, and likely about $4 Billion of ad fraud later, there has been no shortage of articles about the matter and initiatives attempting to resolve the issue.
Yet over a year after the Association of National Advertisers (ANA) and the American Association of Advertising Agencies (4A’s) formed a joint task force to investigate ad fraud and transparency issues, the advertising community is still grappling with this complex and ever-evolving problem. Stephan Loerke, Chief Executive of World Federation of Advertisers, was quoted at Cannes Lions last month, saying “ad blocking, ad fraud, transparency, ad viewability – all these are pretty fundamental questions that we need to be addressing as an industry collectively...”
According to David Perez, co-founder and former CMO of Convertro, “fraud has been a concern since the advent of digital marketing... What is new is simply the fact that, unlike during the original Internet heyday, we now have the technology to identify and put a stop to fraud before it does too much damage.”
How A Good Attribution Solution Is Inherently Anti-Fraud
Attribution technology uses science to determine what media are driving purchases, so advertisers no longer need to guess or waste their budget on channels that aren’t performing. With AOL Convertro’s attribution solutions, any bot or fraudulent activity not explicitly filtered by our software is heavily penalized by algorithmic attribution, due to showing very little or no ROI for that particular source.
Proportional crediting is key. In proper attribution modeling, credit is assigned proportionally to each touchpoint according to its influence on the customer’s purchase or conversion decision. The goal of attribution is to determine which touchpoints are producing a positive result, and, by using the cost of each touchpoint, an advanced attribution system can then show which touchpoints are profitable. This means marketers are inherently steered-away from fraud in a good MTA model, as fraudulent sources are basically “kicked out” of the credit path, since they don’t contribute to the final conversion event. For example, banners stuffed below the fold won’t drive conversions, due to lack of viewability, making viewability an inherent part of our solution.
It’s not just the monitoring or analysis of the customer path that is useful; advertisers can use the Spend Recommendation tool within the dashboard to allocate funds away from any underperforming or unnecessary channels.
Many retargeting solutions can be over-credited in last-touch attribution, because of their very nature of targeting an existing “hot lead” which may or may not have needed the additional touchpoint to convert. Touchpoints that don’t actually drive conversions are automatically not credited for the sale, regardless of their position as first, middle or last touch. Advertisers leveraging Multi-Touch Attribution (MTA) can effectively drive ROI and optimize efficiently even if campaigns are not on a conversion-based pricing model.
Knowing the complexity of organizational navigation, AOL Convertro offers Change Management and strategic guidance as part of the solution. Our strategy team is composed of industry veterans who can provide counsel on all aspects of the process: data policies, implementation, and more. We enable scenario planning for both MTA and UMAP clients and set out a strategic roadmap to ensure success.
In addition to helping your organization adapt to these important changes, we build custom attribution models based on client specifications. If you would like to define and use a customized attribution model, you can let your assigned Client Services representative know.
This post was written by Sr. Product Marketing Manager, Maryam Motamedi.
Want to learn more about Convertro's solutions? Info@convertro.com
If you are a marketer, you are probably familiar with the 5Ws model -- Who, What, Why, Where and When. Some even add a sixth question (“the H”) -- How? Those who can address these questions well can drive great results. Attribution vendors who offer marketing mix modelling (MMM) and multi-touch-attribution (MTA) help marketers increase ROI by answering some of the “6Ws”:
- Where: place your message in the right media space - specific placements or keywords or shows that are highly effective.
- When: understand what times of day/days of week work best and optimize accordingly (more sophisticated vendors help you find the right timing to put your message in front of the individual, rather than using aggregates like time of day).
- What: test different creatives through the lens of ROI and optimize that message to drive consumers to particular actions: particular offers, high profit products,etc.
- How: use attribution to help focus media spend on tactics that yield great results and remove tactics that perform poorly.
- Why: the best attribution companies model the consumer choice process, displaying how much incremental revenue the company earned because of each marketing touchpoint a consumer experienced prior to purchasing.
However, even these great companies are missing the bigger picture -- the reason that we behave the way we do is usually because of who we are.
Think about it this way, if you are selling razor blades for men, what factor best explains the reason that a user purchased your product? Is it because he interacted with your cool video ad on YouTube or is it because he is male and needs razor blades? Do you think the ad would yield the same results if presented to a female?
Most MTA/MMM vendors miss the Who: what are the person’s characteristics that explain his/her behavior? Even more important, they then miss out on providing recommendations and activation options based on audience data. This is true the other way around as well. DMPs are great at understanding people and finding lookalikes at scale, but are missing out on understanding why these people are buying, where to find them, what message to put in front of them and when.
So, bringing the two together can answer all the questions that marketers should be asking, at scale. For example, you can collate the best performing audience segments by campaign ROI and then find more people like that through your buying system (DSP).
You might be saying to yourself, “This is great! Now I can do everything by joining these two separate systems.” Unfortunately, it’s not so easy. Joining two different systems; using DMP A and attribution vendor B has its challenges:
- ID syncing between the two systems will result in loss of scale - for example only being able to match 70% of the people from your DMP to your MMM/MTA vendor.
- Looking at audience vs. marketing effectiveness side-by-side requires deeper integration between the two.
This is why more and more providers are integrating attribution and data management into ONE system. A properly integrated combo provides a holistic view across audiences, content, inventory, timing as well as the effectiveness of your tactics. This enables you to then execute buys at greater scale.
Want to learn more about Marketing Mix Models + Multi-Touch Attribution best practices? Get in touch with one of our experts,email@example.com.
This post was written by VP of Product Management, Roi Lavan.
The promise of programmatic advertising technology is increased efficiency - both in dollars and time. The reality is more of a mixed blessing, as the proliferation of vendors, solutions, channels, devices, and challenges complicates almost every step of the campaign lifecycle.
Multi-touch attribution (MTA) is being used increasingly to provide a north-star of measurement - giving clarity as to the effectiveness of each ad a consumer sees. But what’s a marketer to do with this clarity? The work of continual optimization across paid social, paid search, display and video media can be mind-numbing. When you multiply that with creative optimization at a message level across these same channels, just the thought of another A|B test can lead shear depression.
To help, the individual systems of our programmatic stacks will optimize for us. They can move media spend from tactic to tactic, and creative to creative, based on various goals we set. However, a multi-channel marketer taking this approach now has siloed systems working toward different KPIs. This is like the different sections of an orchestra playing different pieces simultaneously. Without a single conductor leading all the players in the same piece, there’s no harmonious result.
Talk with your MTA provider about how their systems can act as a conductor. Here are some elements that MTA should be able to remote control other systems in your programmatic stack:
- Spend Allocation: If your MTA system provides fractional credit at the ad level, it should be able to direct small budget changes to DSPs and SEMs.
- Keyword Selection: Some keywords, while they don’t close a sale, are valuable in the mid-funnel portion of a consumer’s journey. Your MTA system should value each search keyword according to multiple factors, not just how frequently it closes a sale. By controlling your SEM tools, MTA can defund keywords that have the least value, while investing in high ROI keywords.
- Audience Selection: MTA systems that integrate directly with a DMP provide ROI insights for audience segments purchased as well as those who have performed well even without being activated. Certain audiences have high ROI. Other audience segments will be heavy repeat purchasers. Other segments may have a particularly short path to purchase. Use these insights to create new test buys that target these overlapping segments.
- Creative Messaging and Sequencing: Modern MTA solutions provide insights on creative ad performance: where high-performing messages fall in the path-to-purchase, which call to action yields the highest revenue per purchase. Links between MTA and ad servers or dynamic creative providers can automate the optimization of creative.
- Frequency Capping: By overseeing all channels, MTA systems will show you the optimal ad frequency for reaching a brand’s key goals. These optimizations can be transferred to ad servers and programmatic tools to reduce manual work.
How does this all happen?
MTA systems can be integrated with other parts of your programmatic stack in a few ways:
- Server to server connections: These advanced integrations allow fast sync of two systems in real-time or near-real-time.
- APIs: These interfaces allow one system to request or send information to another and have that info go to the right place. Depending on how sophisticated a system’s APIs are, one system can remote-control another system.
- Batch updates: These transfers of large data files can occur daily or hourly. Batch updates can contain user IDs, conversion data, and recommended optimizations/changes.
How do we avoid letting the machines take over?
Speak with your MTA provider about the business rules and governance you want to institute. Advanced MTA systems allow a variety of engagements between the attribution results and the activation commands:
- Manual: You may wish to review all MTA optimizations for a given channel first, then manually log into the DSP/SEM and change the items you approve.
- Approve automated changes: MTA solutions like Convertro will present a variety of options for optimization. You can review the options and click “ACT” on the items you wish to optimize. Convertro will then automatically make the adjustment in the programmatic tool for you.
- Fully automated changes: MTA, if properly integrated, can virtually take over the optimization of campaigns on a DSP or SEM, using its north-star view over every channel to ensure optimal performance in each channel.
Want to learn more about Marketing Mix Models + Multi-Touch Attribution best practices? Get in touch with one of our experts, firstname.lastname@example.org.
This post was written by Michael Lamb, Director of Product Marketing.
When I first heard the concept of “machine learning” my mind conjured up images of Robert Patrick repairing his body in Terminator 2, or the HAL’s famous line from 2001: A Space Odyssey. While these outcomes of artificial-intelligence-gone-awry still worry me from time to time, I’ve come to love what machine learning does for us in our modern age.
Machine learning is great at discovering anomalies in data - finding needles in haystacks - very quickly. This helps authorities uncover insider trading or fraudulent transactions on your credit cards. It helps Google identify spam and remove it from your email. Amazon uses machine learning to predict what other items you might find useful.
Machine learning is also very helpful for classification - bucketing things that are similar. Humans can do this too, but when the similarities aren’t obvious, machine learning picks up the slack. News aggregators use machine learning to comb through millions of articles to curate personal experiences on Google News, Flipboard, and other apps. Machine learning helps Kayak organize flight pricing and routes to deliver personalized results. Real-estate appraisers use machine learning to accurately determine housing values. It’s no wonder that Stanford’s most popular class is machine learning.
Recently, there has been an increase in press coverage on how machine learning impacts marketing. Good articles from CMO and Econsultancy help us marketers understand the role that machine learning can provide in building more relevant creative and automating marketing processes.
Here, though, I’d like to speak specifically about how machine learning can help marketers in calculating the true effectiveness of their marketing.
Marketing activity creates a deluge of data. For every $1,000 spent, you may generate anywhere from a million to over ten million data points crossing impressions, clicks, interactions, users, geo-demographics, psychographics, conversions, viewability, GRPs, subscribers, products preferences and configurations, content preferences, and revenue. Attribution providers seek to award credit for converting purchasers to individual marketing activities, so making sense of this deluge of data is critical.
Machine learning can be applied to attribution in two ways:
- Helping to Automatically Sort Marketing Efforts. Instead of manually determining which buckets to classify your marketing, a machine learning algorithm will crunch through all the possibilities and sort the marketing automatically. For example, when Marin executes a search ad which is clicked by the user, machine learning can associate that action with the appropriate marketing strategy used in the development of that ad.
- Removing Human Bias. Machine learning is critical to removing human bias from how we will weigh the results of that search ad against everything else - from print ads, to broadcast ads, to catalogs. Marketers who arbitrarily assign how much credit to give search activity vs social vs. display vs. TV are immediately biasing their results. It’s important to let the data speak for itself, and applying machine learning algorithms help do this more efficiently.
How does this all happen? As each generation of computer hardware and software advances, machine learning algorithms are written that exploit hardware advances like parallel processing Central Processing Units (CPUs) and Graphics Processing Units (GPUs) to perform a range of general statistical learning problems at scale. Advertising in today’s tumultuous media landscape requires near-real time instruments which can quickly identify changes across many, many variables.
Our modern day consumer has come to expect the benefits provided by machine learning. It helps us get answers more quickly, more accurately and with more personalization. CMOs and marketers are coming to know how to activate more personal ad experiences using machine learning, but I would encourage us not to stop there. Holistic omni-channel measurement and optimization can also benefit from these advanced data science techniques.
Want to learn more about Marketing Mix Models + Multi-Touch Attribution best practices? Get in touch with one of our experts, email@example.com.
This post was written by Michael Lamb, Director, Product Marketing.
Our economy is transitioning to becoming data-driven. Businesses are leveraging big data and machine learning to generate insights and make better decisions. There are a lot of awesome things you can do with big data. You can track disease outbreaks, recommend movies and songs, predict the outcome of the next election or optimize your marketing spend on the most highly performing channels. But before you can do all that, there is one basic thing you need to do…collect the data!
Here are four things you should be aware of when collecting data at scale:
1) Data Completeness
When collecting data at high scale, you will be using a distributed system of multiple servers, each collecting a part of the data. Later on, when you process the data, you want to be sure that when you process data for a certain period in time (i.e. all the data that came in between 8:00am and 12:00am), you have the complete dataset for that period. If one of your servers failed to deliver the data it collected, you will have “a hole” in your data.
You can avoid that situation by using service discovery. It can be a tool like Consul or even a simple database table. The service discovery will keep track of all running collection servers. This will allow you to go over the data you wish to process and make sure you don’t process it unless it has been delivered by all of the servers that are active at the time.
2) Handling “Bad” Data
The internet is a wild, wild place. Your collection servers could be probed by bots. The server that was just provisioned to you might have been serving some other system in the past and stale DNS entries could still be sending traffic your way. It could even be packet loss or just human error. You should expect that you may collect data that is not usable. Your system must not fail when such data arrives.
Now you should decide, what is the right place to filter out bad data? Filtering bad data on the collection layer is cost-efficient. You don’t have to pay for storing and processing it. There is only one caveat to that approach though--if you did not collect the data, it’s gone for good. At Convertro we have many times recovered usable data out of bad data that was collected. Sometimes the value of recovering data from a badly tagged marketing campaign is worth the extra cost of storing and filtering it further downstream.
When you collect thousands of incoming data streams, it’s very hard to notice that something may be wrong with one of them.
At Convertro we collect data from multiple sources for multiple clients. We want to be able to notice when there is a drop or spike in one of those thousands of streams. We run anomaly detection algorithms on the data we collect and compare traffic levels to trends observed on similar periods in the past. This allows us to detect issues very quickly and proactively help our clients fix them before they become significant.
4) Belated Data
When you break the barriers of cyberspace and wish to combine the data you collected online with data from the real world, you will realize that different streams of data arrive in different cadences.
At Convertro, data sets contains online ad-views, website visits, online conversions, TV spot logs, in-store purchase, mailed catalogues, weather and many other data points. Online data is collected in real time, but TV data or in-store purchase data could arrive in a delay of hours to weeks. When we get belated data, we learn of something that happened in the past and now we must update our calculations accordingly.
There are several ways to handle belated data:
In some use cases, data that arrives late is no longer relevant. If you have already performed irreversible actions based on the data you had at the time, you cannot change your actions and there is no point in adding the new data. An example for this could be a real time bidding algorithm. If it already made its bid, there is no point in recalculating it.
If you are storing raw data and analyzing it on the fly, you could simply add the new data. It will be picked up the next time someone analyzes it. This approach is possible when there is no high Service Level Agreement on data query time. If you wish to provide your clients with a highly responsive dashboard, you must pre-calculate your data and cannot use this method.
When the calculated metrics are additive (i.e. sum of conversions per day), you could update the aggregated metric with the new data that has been introduced.
You could also apply these methods non additive metrics like average by keeping track of the number of items that went into the aggregation.
If you require a complex metric such as ranking (i.e. the ‘n’ highest spending customers) you will have to rebuild your data set.
At Convertro we have built our system in a way that allows us to go back in time. When a feed of belated data arrives we can go back and “replay” all the data we collected since the earliest data point in that feed to our algorithms. By doing this we could reflect the new knowledge we just obtained to better understand what happened in the past and to better predict the outcome of our future actions. This ability is crucial if you wish to provide one model which accurately combines the online world with the real world out there. Otherwise, you will be modeling each data set separately and combine them using duct tape.
Want to learn more about Marketing Mix Models + Multi-Touch Attribution best practices? Get in touch with one of our experts, firstname.lastname@example.org.
This post was written by Iddo Rachlewski, VP of Research and Development, Convertro
For additional insights on MMM + MTA, check out the following resources:
Forrester Wave™: Cross-Channel Attribution Providers report recognizes Convertro as a leader with its interface, algorithmic approach and scenario planning capabilities.
Why Being Customer-Centric Means Being Data-Driven, research that reveals how multi-touch attribution is critical to understanding actual return on investment.
As someone who made his career in attribution and analytics I can tell you that no one really cares about MTA, MMM, MMM+MTA and/or Unified Models.
What everyone cares about is how to use the RESULTS from these methodologies to optimize marketing and gain true ROI as a result.
Attribution is a means to an end. The end goal is to be able to measure the effectiveness of marketing and then use that to activate marketing campaigns across offline and online channels, while considering all factors that are impacting a consumer during the purchase cycle. This activation can be done in many ways, but the most effective way is what we like to call “closed-loop activation”.
Closed-loop activation is a term that is used to describe the automation of optimizing marketing spend:
Closing the loop is important because it helps marketers to quickly understand what is working and what isn’t and take actions accordingly. When done automatically and at scale, this can be used as a competitive advantage that helps business grow fast leveraging technology instead of manual processes. Think about other fields such as logistics: Amazon managed to become one of the biggest retailers in the world by having a closed-loop inventory optimization mechanism supported by technology.
Analytics that stop at measuring or even attributing value to marketing but fall short on optimization and closing the loop will not be actionable nor fast, causing more work on the marketer side to infer the insights, run the campaigns, synchronize between different systems and finally do that at scale in real time. One of the key ingredients in speeding up the activation loop is incorporating audience management with advanced analytics. For example, almost all competitors in the multi-touch attribution space incorporate a DMP or have been acquired by a company with a DMP.
We also often forget that while media accounts for the lion share of dollars in advertising, creative messaging accounts for most of the effectiveness in driving consumer behavior changes. Smart advanced analytics systems must help advertisers close the loop on creative optimization and personalization.
Closed-loop optimization does not mean “passing back attribution value to 3rd party”. If you want a closed-loop system that works at scale you have to have:
- All the components within the system (measure, attribute AND optimize)
- Always-on system that decides based on the most current data
- Full transparency into the reasons behind decisions that were taken by the system
- Ability to remote-control programmatic buying systems
When passing back the results to a 3rd party you fail to achieve the transparency, in fact you don’t even know what decisions are made and why.
Some pitfalls of simply passing back attribution values to a 3rd party:
- No intelligence on availability of inventory.
- Data loss due to poor user matching percentages between platforms.
- Confusion for users of the 3rd party tools as to why specific optimizations are being recommended.
- Manual effort required to implement recommendations from the attribution platform.
Although passing back the attribution results to 3rd parties is a common way to “close the loop” today, we believe that the future is having ONE system that closes the loop for you. A simple example to such a capability is what Convertro does today with ONE by AOL: Video; budgets are shifted programmatically between different tactics based on the ROI goal and using the attribution results, as well as other factors such as inventory availability, pacing and more.
However marketers decide to implement closed-loop activation, it is important to think about the advantages to their scale, time-to-decision and transparency of decisions made by the system.
Traveling between cities and states with paper maps can sometimes be a grueling process. To go from point A to point B typically requires the use of two maps; interstate maps and city maps. The former would help determine the route between two states or cities. The latter would help figure out how to get to a specific location within the city. The use of two separate methods to solve for one overall problem presents a lot of challenges.
Now, with the advent of the GPS, drivers can plan out an entire route from point A to point B to the utmost level of detail that is constantly updated in real-time based on where the driver is, where they want to go and their individual preferences. The technology uses sets of maps that provide the driver with a combined look at the overall travel route and the level of detail in the city maps in a consistent way. While driving, the GPS constantly monitors the process to make sure the driver is on the best route, providing information on any obstacles or detours, to make sure that they reach their final destination in the most optimal way.
This comparison is a lot like how a marketing attribution platform should be operating for marketers. The job of marketing attribution platforms should be to help all decision makers, at all levels within a marketing organization take the right actions to achieve their business objectives.
Are you getting that from your provider? Here are the three things an attribution partner should be doing in order to provide you with a GPS, not a paper map:
- Use a single framework consisting of all marketing activities to measure their influence on the marketer’s goals. In the scenario above, using one paper map on its own is not sufficient enough to get the traveler to the right location and requires that the user consult both maps on their own and try to piece together the directions to come up with a combined route that’s optimal, while the GPS provides a consistent overall look that is updated in real-time.
In terms of an attribution platform, this allows one to create an actionable path specific to each decision maker (channel manager, VP or CMO) in the organization, but leads them to the same end goal regardless of their function and scope.
- Measure the impact of all controlled and uncontrolled factors on the ability to achieve business goals. Controlled and uncontrolled inputs vary from product trends, consumers’ tastes, consumers’ media consumption habits, consumers’ responses to messaging, and competitor responses, to less obvious factors such as weather, economy, and seasonality. These things are dynamic, meaning they’re constantly changing and require a model that can monitor inputs as they evolve to take the most updated and accurate information into account. The impact of these factors produces a different response from each consumer, so the model needs to be granular enough to measure those nuances. Just like how the GPS can offer alternative routes based on real-time, unforeseen traffic or construction changes, your attribution provider needs to be able to take into account all external factors impacting a campaign’s effectiveness in order to inform you on how best to optimize.
- Facilitate actions that lead marketers to their objective. It is no longer sufficient to have a plan and set it on auto-pilot. The actionability component of any attribution solution also needs to allow the marketer to bring in their unique perspective and evaluate/grade the impact of any optimization on its ability to meet marketing goals. Marketers need a balance of automation and oversight where they can continuously monitor the execution of the plans and adjust them when the need arises. Just like a paper map can’t provide the driver any guidance on the best optimal route, a good GPS will provide the driver with options and their impact which will lead them to a good path to reach their destination.
Written by Vishvesh Oza, Director of Product Management, Convertro
In today’s fragmented media landscape, the consumer’s path to purchase has evolved into a meandering trail of online and offline touchpoints, making it even more difficult for marketers to understand the true return on investment. This shift has caused the need for marketers to leverage measurement tools and rich data in order to make truly informed marketing decisions. Predictive models are the chassis and engine for data-driven decision making. When allocating marketing resources and executing marketing tactics, there must be a predictive model that links marketing actions to marketing objectives, sales--for example, so that marketers can gain full clarity over their efforts on a micro to macro level.
Here are three reasons that it’s essential for marketers to leverage a single, unified measurement model:
- The two measurement models in use today are fundamentally different. Today’s most competitive marketers use marketing mix models (MMM) and multi-touch attribution (MTA) models to inform strategic and tactical decisions respectively. Each model lives in isolation, uses different sets of data, and shares some features. Where the marketer subscribes to both models the two incompatible pieces are cobbled together either formally by an MMM+MTA provider or informally by marketers themselves.
- Processing different data results in different outcomes. Because the models are developed using different data, the results are most often biased. What’s the impact? Certain factors are accounted for in one measurement model but precluded from the other and given that that the models are substantively different in the way they link marketing actions to marketing objectives, you end up with results that are both conceptually and mathematically irreconcilable. The result is a ”house of cards” solution that jeopardizes the effectiveness of millions of media dollars.
The current solutions can’t work when forced together. Marketing-impact analysts tend to think about a “unified” model in two different ways, one that cobbles together separate but linked equations, and one that ensembles models as a weighted-blend which is determined by their ability to predict performance. These two perspectives are in stark contrast to thinking about it in a cohesive singular framework.
The issue with the first approach is that it doesn’t resolve the conflict between the two sources of information. This happens partly because an MMM model has no knowledge of how individual consumers make choices. Forcing separate equations together is taking the easy way out. It estimates the impact of aggregate data and user-level data with two different models rather than think about modeling how consumers truly make decisions.
The issue with the second approach is that it determines the blend of models based upon how well they fit observed behavior rather than how well they predict the causal impact of marketing activity. This trade-off sits squarely against the whole point of predictive analytics, which exists to measure the causal impact of marketing, and guide marketers to make financially efficient and effective decisions.
A new, more elegant solution is required.
A sophisticated unified measurement model will yield major marketing outcomes. In order for MMM and MTA to exist harmoniously they must be effectively unified in a single model. They must establish the same link between marketing actions to marketing objectives, and they must process the same data inputs, considerations, and variables.
Having a model that exists within a single framework allows the entire marketing organization, from the channel marketing managers to the CMO, to more effectively execute and optimize marketing efforts. A single model provides a consistent and more accurate set of metrics, better informs how to drive decisions and allocate marketing budget, and helps marketers influence the entire consumer journey and achieve key objectives.
Want to learn more about MMM+MTA best practices? Get in touch with one of our experts, email@example.com.
This post was written by Michael A. Cohen, Ph.D., Head of Data Science & Analytics, AOL Convertro