rock formation by ocean

EXAMINING TRANSPARENCY IN DIGITAL MARKETING

By: Jared Rodecker, VP, Advanced Analytics Solutions

 

What the world needs now is transparency, real transparency.
It’s the only thing that there’s just too little of ...

Jackie DeShannon, please forgive me for butchering your 55-year-old ballad. I couldn’t help myself. As a marketing professional, I find the recent calls for more transparency in our industry to be music to my ears.

Recent high-profile changes to Google’s cookie policy and other digital tracking disruptions are to “blame” for this transparency push. In 2020, true transparency requires the developer of machine learning algorithms — whether it be for customer targeting and personalization or automated media placement — to share explicit details with their clients and stakeholders. That data includes:

  • Mathematics and statistical techniques used to process the data.
  • Outputs that demonstrate the efficacy of the technique as it was applied.
  • Code and data that would allow the client or stakeholder to replicate the results of the algorithm.

In some cases, privacy considerations will not allow ML developers to provide 100% transparency, particularly when they are third parties. For internal marketing teams within your organization, there’s really no reason they can’t be wholly transparent and follow the same protocols that, say, ML developers in the Risk or Finance departments have to follow.

The era when third parties can simply claim that their methods are “proprietary” — and as such not subject to transparency — is quickly coming to an end for all but the most dominant players in the market, whose size and market influence can insulate them from the need to provide transparency to their clients (e.g., Google, Facebook).

Cookies Gone Stale

Google’s shift in policy has provoked many companies to examine their digital marketing practices and has brought a lot of awareness to how cookies work (or, more to the point, don’t work). Previously, there was probably too much false confidence that cookie-based methods were more accurate than they really were; more consumers were regularly clearing cookies, and the proliferation of cross-device media consumption has made cookies a less accurate way of tracking the same consumer over time than in years past.

Thinking about how they will adapt to this new policy has caused a lot of companies to reconsider how they were using cookies previously and realize they weren’t as good as they thought they were. Spurred by Google’s decision, companies are now thinking more critically about how they place media and target consumers. These are all good things.

The change will require marketers to work harder and think more critically about their approach, but in the end, it should incentivize innovation in a space where for far too long marketers would just assume cookies were near-perfect and would over-rely on them in their marketing efforts.

The Value of Explainability

There is momentum for data scientists to embrace the “explainability model” rather than continuing to rely on black-box algorithms. Granted, there are still some business problems, such as forecasting, for which black-box models will suffice. If pure forecasts of sales or lead volume can be shown to be highly accurate, those forecasts can benefit our clients (particularly their finance departments) even if dense, hard-to-explain methods are deployed (e.g., neural networks).

But for most marketing problems — and certainly the ones that impact high-cost/high-risk bets (e.g., producing new creative for a TV campaign, launching a new product, or entering a new market) — transparent justification for the methods used are essential. Black-box models will be palatable to clients only as long as they are working, but alarm bells will go off the moment they stop. Clients and stakeholders will face a serious crisis in confidence if a data scientist can’t adequately explain why the once-working model is now busted. Models with strong explainability are much more useful in understanding where the model is failing and can be invaluable sources of insight when things go off-track, which they almost always will.

Once clients get a taste of transparency being infused into the deliverables they get from their marketing agencies, they never go back to being satisfied with black-box marketing solutions. They might not always fully access the accompanying code and documentation that provide transparency, but simply having it in hand builds trust and strengthens the relationship with clients.

That trust comes with two benefits: more understanding and forgiveness when technical errors occur, and more appetite to let marketers innovate and steer their businesses in new directions that will benefit them. With the trust engendered by transparency, clients are more willing to let marketers take them on the journey toward innovation and growth.

Brands that embrace transparency themselves and require it of their agencies and vendors will be better positioned to navigate the growing importance of consumer privacy. They’ll also have a competitive advantage, both in how they can position their brand and in building cultures of innovation and continuous improvement among their data science teams.

The move toward transparency is a one-way street, and there’s no turning back now.

... No, not just for some but for everyone ...