There is more than 85% chance your data projects will fail but we figure out a way to solve this.

Recently McKinsey reported that over 85% data projects, genAI projects fail.

That's becuse most don't build it as a product.

If you are trying to build a product, you will have things like

- Objective to achieve
- Pain points solving and so on
- Target market,
- End results

All mapped out before even starting to built the product- but most data teams they just start tracking everything and think about business later.

Which results in data not being actionable, and bunch of unused tools and big bill in licenses.

So this is how it looks like building data product right way:

1. Identify decision-makers: Who will use this data to make decisions?

2. Map decision points: What specific choices will they make with this information?

3. Define success metrics: How will you know the data product is delivering value?

4. Set quality standards: What level of accuracy, freshness, and completeness is required?

5. Establish clear ownership: Who is responsible for each aspect of the data product?

Here is how real life example might look like this

Product Initiative: Churn-Risk Early-Warning System

Use Case:
Flag subscribers who are likely to cancel in the next 30 days so the retention team can intervene with a timely, data-backed offer.

Primary users:
Marketing Department → retention marketing roles people who plan save-offers & run A/B tests
Customer Success → account managers, reps who do proactive outrach
Finance → analysts who roll the churn forecast into LTV / revenue models

Key-metrics framework:
1️⃣ Engagement-gap index- drop in daily watch-hours for > 5 days vs 90-day baseline
2️⃣ Payment-friction flags (25 %) – hard / soft credit-card declines and auto-retry failures
3️⃣ Profile dormancy (15 %) – household has not opened app across any device for ≥ 7 days

Success criteria:
- ↓ 15 % relative churn in six months
- 30 % lift in win-back-offer conversion
- 95 % of at-risk users contacted within 24h of first alert

Now you have a clera roadmap of initiatives to work on

TL;DR

1. Start with problems, talk to executives
2. Think like a product manager for every data initiative
3. Prioritise according to complexity vs value
4. Constant feedback, gradual improvements
5. Build like legos, starts small build while proving the value

This doesn't matter if you are building dashboards, data initaitevs its for all.

Whats your take on this? You think the majority projects fail cause of this?