Before I worked in analytics, I was on the other side of the dashboard. As a sales representative, I covered more than 250 physicians across 19 geographic areas and spent a good part of my week logging contacts and events in a CRM. Later, as a business consultant, I worked with community pharmacies on profitability and operations. Today I lead teams that build the data products those same field and commercial teams rely on.
That journey taught me something simple: most commercial data problems are not technical problems. They are relevance problems.
1. Field teams don’t need more reports. They need fewer, better questions answered.
When you’re in the field, your time is split between travel, meetings and admin. A dashboard with 40 visuals is not a tool; it’s homework. What a rep or a key account manager actually needs is short:
- Who should I see next, and why?
- What changed in my territory since last month?
- Am I on track, and where am I losing ground?
Every report we build should map back to a question like these. If it doesn’t, it’s probably serving headquarters, not the field.
2. Data quality starts at the point of capture
CRM data is only as good as the moment it’s entered. I know from experience that when logging a visit feels like a compliance exercise, the data reflects that: rushed, incomplete, inconsistent.
The best way to improve data quality is not more validation rules. It’s to give something back. When reps see their own data turned into something useful for them – a clearer view of their accounts, less manual reporting – the quality of what they enter improves.
3. Combine sources, but explain them
Commercial analytics in pharma usually means stitching together CRM activity, sales data, wholesaler information and market data from providers like IQVIA. Each source has its own granularity, timing and blind spots.
The temptation is to merge everything into one “single source of truth” and move on. In practice, users trust data more when they understand its limits: when wholesaler data lags, what a market share figure does and doesn’t include, why two numbers don’t match. A short, honest explanation next to a metric does more for adoption than a perfect data model nobody understands.
4. Validate with local teams, not just with code
Automated checks catch broken pipelines. They don’t catch a number that is technically correct but makes no sense to someone who knows the market. Some of the most valuable data-quality work I’ve been part of was simply sitting with local teams and asking: does this look right to you?
That loop – data team builds, local team challenges, data team adjusts – is slower at first and much faster over time.
5. Start with the decision, not the data
The most useful question I’ve learned to ask at the start of any commercial analytics request is: what decision will this help someone make, and how often? A weekly territory-planning decision needs a very different product from a quarterly strategic review.
As pharma moves towards omnichannel engagement and AI-assisted recommendations, these basics matter even more. A next-best-action model built on poorly captured, poorly explained data will fail in exactly the same ways a bad dashboard does, only faster.
The tools change. The principle doesn’t: build for the person who has to act on the data.
Leave a Reply