4Finance

Data driven company, it is a well-known buzzword, since 2016. Many marketers and agencies talk about it, but still it is not so easy to implement that and make decisions based on data driven attribution results. In our organization, in marketing we still watch two attributions: last click non-direct and data driven. This data-driven is tailored for us and by us, we still work and improve it by adding more events to consider on the customer's path.

We are lucky, we have a lot of 1st party data that we are able to link with all our sales events and with their previous paths. We sell online, however signals about sale’s events come from the backend system.

Data driven marketing promise is “to reach the right people, at the right time, with the right offer and through the right channel”. Sounds really great, but after 2 years of implementing this approach I can write that:

• Yes, we synchronized our 1st party segments with Facebook, Google Ads and even with programmatic DSP platforms. Moreover, recently we did it with a marketing automation tool, so we are able to send emails based on data driven rules.

• Yes, we build lookalike audiences out of 1st party segments to gain more new clients and exclude current clients from targeting (not always, but we do).

• Yes, we have data driven attribution (bespoke Markov chains solution) reports about sales events per each digital (and non-digital) channel. We have even calculated CPAs and they are available online on a daily basis in unified Looker charts. (This part is really fresh and we still work on minimizing discrepancies in reported channel’s costs).

• Yes, we started to consider data driven results with our wider econometric modeling, although big Covid-19 turbulences crash all this kind of predictions.

but,

• We still struggle with proper judgement of our effects in our channels.

• With our financial status and due to some of our restricted products, we often struggle with data synchronization, especially with Google and Facebook environments.

• We still buy many channels with classic targeting segments like gender and age like in old fashioned offline TV.

• We optimize our efforts not only toward sales, but also registrations, traffic or brand awareness only (depending on country and local marketing manager’s beliefs).

• In some markets, we still buy digital classic displays (like one day presence on a portal's homepage) or classic bulk emailing.

The implementation of a data-driven approach is a long-term project and needs a lot of patience and constant support from the top management. We are fighting with the most powerful enemy: with human’s habits and previous experience.

Data-driven implementation steps:

1. Configure and integrate your 1st party data with digital buying environments (Facebook, Google Ads,dsp platform, sms sender, email sender, telemarketing tool, web cms, etc.)

2. Unify channel grouping in Google Analytics and build rules for proper incoming traffic descriptions (standardized utms, etc.)

3. Enrich tracking with impressions and click tracking

There we have to use Google Campaign Manager if we want to have a chance to bind it with Google Analytics traffic. So, it is worth having your own Google Analytics 360, Google Marketing Platform including Display and Video 360, Campaign Manager for traffic all possible to track events.

We can use another solution like Adobe, but it will be much more expensive and we have to decide on one traffic and impressions / click tracking solution.

4. If we are using a lot of internal channels like emails, sms or telemarketing, then it would be great to inform our Google Analytics 360 about contacts like email sent, sms sent, call done. We can easily do that with Google Measurement protocol and most of sms / emali / telemarketing tools already have webhook functionalities to inform google about these events (combined with client-ID).

5. Build data driven attribution based on Markov chains, consider internal channels events (email, sms, call). To improve data quality install Google Ads Data Hub which will allow us to consider also non-converting paths.

Be aware of some model limitations (can’t consider impressions from Facebook, from Affiliate channels, from organic traffic, etc.)

We can build it with external agency’s help or we can use our own BI department.

"The Implementation Of A Data-Driven Approach Is A Long-Term Project And Needs A Lot Of Patience And Constant Support From The Top Management"

The place we can merge/link/bind these data is Google Marketing Cloud with their Big Query. It is a very flexible and perspective environment, with proper data structure we will be able to build our own machine learning algorithms for many different purposes, like predicting client’s behavior, their willingness to buy or sensitivity for discounts.

6. Build media investment cost data storage with daily updates

Again easy to write, much harder to implement for many markets (it took me several months and still improving).

7. Build and share unified reporting with last click non direct and data driven attribution results with major KPIs like CPA per channel, blended CPA will be the same for both attributions, we will have a different picture for particular channels, in general all well tracked display channels will look much better in Data driven attribution.

8. Maintain constantly, start using that and make decisions based on that. Watch on global ROAS.

Where (which channels) can we use 1st party data

1. Programmatic display DSP

Here we need to feed our DSP-DMP from the web site level with our clientID to build a pair cookie+clientID. It will let DSP to match signals about conversions with generated clicks and bought impressions.

2. Google Ads (paid search brand, non-brand, display ads (GDN and YouTube), gmail)

Here we can synchronize our predefined segments via Google API into customer match segments

3. Paid Social (Facebook)

Here we can synchronize our predefined segments via Facebook API into custom audiences

4. own / crm channels (emailing, sms, telemarketing)

There it is obligated to use marketing automation tools and build scenarios according to the client's behavior. Easy to write, but it takes a lot of time, efforts and money to implement it properly.

These integrations need to be done between our CRM 1st party data and ad buying machines (FB, Google, DSP, email sender, sms sender) environments.

We can do it with our IT departments or we can configure (also with IT support) marketing automation tools like Sales Force or Sales Manago (same functionalities, but much cheaper).

Marketing automation tools should become the heart of our data-driven marketing approach. They allow us to constantly update and synchronize segments with buying environments (FB, Google, DSP, SMS, email, telemarketing) and manage scenarios with using channels in proper and the most accurate order (it might be different for different client’s segments).

This implementation must be strongly supported by top management and we need to involve many departments (in our case: crm, collections, marketing, product development, it).

We should not forget that a data-driven approach is much wider than data-driven attribution.

It contains also:

1. A/B testing and multivariate testing

We often forget what we wanted to test and do not provide really equal conditions for both tested options.

2. Website / app personalization / Usability testing

In case of global CMS management, it might be pretty huge challenge, so again we need to have top management support

3. sales abandonment analysis

Analyzing that is the very first step, in our case it might be our acceptance rate, but also competitor’s behavior.

4. Copy optimization

This we do with prospecting advertisements; we use machine learning in Facebook or Google Ads machine learning dynamic creatives and also in internal email communication where we play with email’s subjects and bodies.

5. Online surveys/Customer feedback

This is what we do often. We know general stats, but we plan to start linking those opinions with user’s profiles to be able to communicate better with the clients.

6. Competitor benchmarking

There we are working on data integration with monitoring tools like Similar Web and Senuto (one of the best SEO supporting tools). It is crucial to know competitor’s activity to be able to quickly react with accurate level of channel’s media investments.

Summarizing, we did a lot to implement data driven attribution and we finally have constantly reported results with this approach. We started to analyze and watch at the same time two digital attributions (LCND and DDA). We constantly (for some channels automatically using API) report marketing costs per each channel, so we can focus on analyzing data instead of collecting. We still work on improvements like adding competitor’s behavior (even with some built-in alerts). We also can’t never forget about creativity and real client’s needs, so we need to keep some space for experiments and uncertainty.