Why people aren’t using data - and how to (maybe) fix it

We all sit on tons of data with potentially significant insights. But we are too lazy to unlock it.

When we were building Growblocks, one thing we stumbled over all the time was this:

Almost all companies struggle to get value from their data. 

I guess you could call it “struggle to be data-driven”. But I feel this is an overused term by now. 

And don’t get me wrong, 100% of people struggle with data quality and trust issues. But I would argue those are symptoms rather than the root cause.

The much bigger issue was: even with all problems resolved and clear visibility across the GTM established, people struggled to interpret and use the data effectively.

What we ended up doing a lot was to run mini-QBRs as part of our onboarding process. Very often this delivered peak-endorphins for our customers. Feeling the depth of insight they had been sitting on for years without using it. 

In some cases, we were able to diagnose fundamental issues in the business - things the CRO or CEO would later tell us were in the back of their heads, but they couldn’t quite figure it out. 

What this tells me is really two things:

1. There are still too many obstacles for folks to work with data - and this is a hard skill gap. 

2. Analyzing data is like going to the gym. People know it's good for them but they never do it (and when they finally do, it hurts a lot).

What baffles me is that Business Intelligence was supposed to solve this issue. Having data folks work on the data stuff, and then serve it up to business folks who do the business stuff. 

But apparently, it didn’t pan out that way. 

What’s happening instead: Folks submit a ticket, wait days to not get the answer they need, at which point they can submit a new ticket or just give up. In effect, this is teaching folks to not ask questions, and not be data-driven.

This problem was referred to by Looker and Tom Tunguz as the “Breadline problem”. Meaning, that for people to get access to insights, they needed to stand in long lines until they would finally get it. 

And I think this problem has gotten a lot worse over the last few years:

1. We’re seeing more and more data being produced by all the SaaS tools we purchased. What used to be some pipeline and lead data is now 10x-ed. 

2. The rate of change in the business has gone up. And data folks are struggling/giving up on trying to keep up with it 

Those two trends are wiping out any progress we’ve made in recent years to get value out of the data we are producing. 

And looking at this, I think it is increasingly hard to imagine a future where we wake up one day and have a data-disciplined workforce fluent in SQL and Python, to take advantage of all the business critical insights they are sitting on. 

So at this point, we should ask: why are we producing all of that data to begin with if we won’t use it anyway? 

Can GenAI solve this for us, please?

As with many things these days, whenever folks run into a difficult problem, they point at GenAI and say/hope that this new technology could solve it. 

I’m not an expert at this. But GenAI still has some issues that make it hard to work with:

1. Can’t reason (how many “r”s in Strawberry?)

2. It’s inconsistent (different answers to the same question)

3. It hallucinates (it just makes up random stuff)

Those are not great features in a business application context. But let’s say all of that stuff gets solved. 

But even if we had the super AI, here is what still will be an issue:

1. Garbage in, garbage out

50% of data quality issues happen at the source, e.g. Tyler the Rep puts in the wrong pieces of information into the CRM. 

50% of data issues come from errors in manipulating the data. E.g. I know that Tyler always puts stuff in wrong, so let me download this report to Excel, mess around with it (aka fat finger the data), and present this to folks.

Data analysis AI won't solve this. 

2. No Skill gap, but will we trust it?

It makes perfect sense for AI to help bridge the skill gap for business folks. They don’t need to learn how to write SQL. 

But if we outsource this part to an AI, we have to trust that the AI is getting it right. If you think about other spaces, this is still not solved:

• AI for content still has a human (that knows that language) read the text and fix issues

• AI for code writing still has a human to read the code and check it 

This safeguard would not exist here. 

3. Cultural change is still pending. 

I wonder how long it will take for folks to actually use data. Many times data issues are really just brought up when you don’t like what you’re seeing. It’s like some US politicians cast doubt about fair elections … until they win that election. 

How long will it take with “perfect data” to overcome that? 

And can AI really help with that? 

Is there a path? 

I don’t know how long it’s gonna take but:

1. AI will be able to solve a bunch of these problems. Simply by way of fewer humans being in the loop and by less and less manual work hinging on humans (think data input) 

2. I think there will be a new generation growing up using and trusting AI tools a lot more. My mum still worries about using her credit card for online purchases. Things like that can change. 

3. I think Adam Smith and Darwin will fix this eventually, folks ignoring the competitive advantage data can give them will eventually be removed from the market by folks that do. Both on a company but also on a career scale. 

What I’m pretty convinced about though, buying Tableau or Looker won’t solve this issue for you. It feels like the right next step. But putting up some dashboards that no one looks at or uses for analysis won’t make you data-driven.