Uniting Data Fiefdoms Under Your Banner

Aligning incentives to destroy data silos

One unique challenge of running a Data & AI function is that every other part of the business

• has their own data,

• figured out ‘a way’ to work with their own data, and

• is incentivised to think locally (department), rather than globally (company).

When you start a new Data gig as a functional leader, you are likely to be looking at:

• a company that existed for n years

• most people have been doing their jobs for a while

• and they are doing just fine

• because the company is making x much revenue

And then, in your first week, when you declare

• our data quality is poor

• our data infrastructure is brittle

• our data culture is non-existent

• we need to and we will! fix all of the above in [timeframe]

your peers go, hmm this person sounds like trouble.

How should you navigate this?🤔

Robbing RevOps to Pay Finance

Let’s go back to the first part—that data existed before you, but not as a function. Sales, marketing, finance etc. have their spreadsheets when the company had <5 customers, and they still use it when they have >500 customers. They may not like it themselves, but it gets the job done. Then you come in and say, we need you to switch to platform integrations, because APIs!

If you think about, the Data function is one of the only business functions that doesn’t generate any proprietary data. But, unlike the rest of the business, you have* access to ALL THE DATA.

Which is nice, but also has drawbacks. For instance, you might have all the data, but surely you don’t have the context for the majority of it—i.e. you are not a domain expert in every part of the business.

Whereas other parts of the business has the opposite condition: they only have access to a fragment of all data, but they intimately know that data.

When departments are disincentivised to think company-wide, they essentially force the Data function to rob Peter to pay Paul—you allocate resources to fix a data issue now, while creating a data issue somewhere else in the near future.

Build a Single Customer View

As data proclamations go, few arouse and confuse as much as the Single Customer View (also known as a 360 view). This is a holistic dataset that captures the end-to-end customer journey.

In a typical SaaS company, it starts from the point of first contact (e.g. hand-raising), covers the whole sales funnel, includes in parallel product activity (i.e. behavioural data) and customer support/enablement, and captures the fateful decision to churn or renew. The totality of the customer journey is captured—both the commercial and the product sides.

This can be a true game-changer for your company. Before a SCV-like asset, senior management resorts to gut feeling/loudest voice in the room in situations like the following:

• An increase in ARR numbers! Cause for celebration. What did we do differently last month?

• We increased marketing spend

• We released several product features

• [everything else in the world that could have influenced the outcome]

• Who gets the credit?

The reality is, you don’t know who, if anyone, should get the credit. Perhaps it’s a cyclical sales trend. Maybe you just got lucky.

The thing is, you cannot do attribution if you don’t have access to the full picture.

Who has access to all data? You do!

Kuvira, the Great Uniter, announcing her intention to unite all data silos in the Earth Kingdom, whether they like it or not.

The first step is to sell the idea to your middle management peers + senior management. The latter will probably focus on the potential ChatGPT applications of it; which you can safely ignore for now (but don’t forget to silently nod). Focus on the positive ROI aspects of SCV—the ability to change course with speed and agility, understanding the causes of things—these all move the needle.

With your peers, you will find a warmer reception. However, there will be detractors even there—if you are heralding to open the Pandora’s box of attribution, certain departments that have zero-sum incentives won’t like it (e.g. who gets the commission—marketing or sales?). Luckily, they tend to be the minority. Plus, has anyone managed to fix attribution in marketing, anyway?😅

Continue selling your peers the big picture. Sure, perhaps you are doing fine locally, but we are not thriving globally. It’s hard to argue using a counterfactual (i.e. where would we be if we had good data quality), so it’s better to remind them the cost of not doing anything. Data quality issues tend to be exponential in scope—usually, various other reports depend on a single data asset. By not choosing to cooperate, they are holding the whole company back. This might be a good time to escalate it to their exec; however do it in the open—don’t go behind your peers’ backs.

Last but not least, create a metric that captures the change. Don’t forget to give it an important-sounding backronym that you can put in your resume! In a previous job, I named one such variable

Iteratively Marginal, Product-Attributable Credit Total—IMPACT

🤯No one could remember what it stood for, but it was a useful measure to track alongside investor metrics. Product managers hate it, unless when they love it!