How to plan customer activities? (part 1)

From traditional top-down targeting & call planning…to AI-powered “Next Best Recommended Action”

This is Part One of our blog series on AI-powered “Next Best Action” tools for field forces.

If you’re sufficiently old, you might remember a time when field forces activities were all about reach and frequency.
planning top down

Reach more potential customers, visit them at a higher frequency, and sales follow, as the thinking went.  This “carpet bombing” was controversial, especially in the pharmaceutical industry.  Command & control hierarchical organizations, top-down targeting & call planning, rigid objectives, and incentives: that’s how it used to work.
Not anymore…



For at least ten yeaplanning ideal worldrs now, the trend in pharmaceutical sales (as in other industries in a digitized world with expected personalization), has been to move away from a traditional command & control model and to implement more bottom-up, collaborative and customized sales models. At least this is the (meaningful) intention. Indeed, in an ideal world, the local plans would start from what the reps are already doing well and suggest only a few high priority recommendations.  In practice, it’s quite hard to go away from the traditional model while ensuring efficiency.

That empowerment trend is now conflicting with another powerful wave that threatens to take the industry in a completely different direction. I’m talking, of course, of Artificial Intelligence (AI) and its applications.

picture AI 1

Within KARETIS, we have developed strong expertise on those topics, and we’ve looked for a solution that reconciles effective people’s empowerment with insights generated from data and technology.


We recently worked with a large bio-pharmaceutical organization on their yearly targeting and call planning exercise for a specialty brand.

  • The professionals involved in the patient journey and the prescription decisions were a very diverse crowd. Their sales force had to work with diverse specialties, both in retail and at the hospital. Local healthcare networks and Professionals communities could be very different from one region to the next.
  • For historical reasons, their brand was a market leader in its main indication, but the competition was much fiercer in some geographies, resulting in a wide range of sales dynamics.

In that situation, only the sales reps had sufficient knowledge of both the local environment and professionals inter-relationships to make a reasonable plan.

After a discussion with the management, we decided to skip the traditional top-down approach and let the sales reps decide their targeting & territory plan objectives on their own.
That might seem like an overly permissive approach:  usually, the management proposes a draft targeting & call plan, and let the sales force make some improvements, within negotiated boundaries.
But a key question remained: how to help the reps to make the right decisions?

 That is where AI enters the game…

Well, for one thing, don’t be demagogic and keep in mind that reach and frequency on segments still works. We often do ROI analysis on sales force promotional efforts, and the algorithms always confirm the correlations: visit more customers, more frequently, you’ll get more sales.
But this has limits. Something recommended at a national level can be very sub-optimal locally.
The complicated dynamics of today’s pharmaceutical brands have forced us to upgrade our arsenal of traditional statistics with more trendy machine learning approaches (a.k.a. “AI”).

With more data and more powerful ways to make sense of it, it is tempting to use machines to tell us what to do when things become complicated.

picture lost found

That is, after all, what we all do when driving or walking in cities. Who stills spend much time looking at maps and plotting itineraries? We have our GPS to do it for us. We know that it can be an imperfect tool, and sometimes the route recommendations are not the best. However, it’s good enough most of the time, so we trust it. And we stay free to choose our road or select options: pass by this place, … it will refresh and adapt its recommendations to the route and choices we made.

Nobody wants to be obliged to comply with instructions, but life is complicated enough as it is. Any proven help is desired and a source of efficiency.

Taking inspiration from the GPS, it’s natural to see pharmaceutical companies, as in other industries, exploring the use of AI for “Next Best Action” recommendations.

The idea is simple: a GPS gets 2 data points from each user: where you are now and where you are going. Based on that, it calculates the optimal route to reach the destination. Similarly, you can create an AI that looks at the current situation – in terms of sales (or any other target metric) and past promotional activities – and, based on your objectives, calculate the optimal activities to reach the objectives. Technically, it’s an upgrade on the old reach-frequency analysis. Instead of recommending call targets once a year or once a cycle, at the national level, the AIs do it at the territory level to adapt to the diverse local situations and refresh as often as you get new data.  It would be too complex and too time-consuming to analyze it manually. When used effectively, the tools help to increase the impact on the most valuable sales targets, and decrease unproductive activities, downtime, and/or traveling.

picture AI 2If you’re a techno-optimist manager, the above description might sound like a dream come true.

Many salespeople don’t see it this way, however. And it’s understandable. Think of it: if your boss told you that she’s rolling out a tool that tracks everything you do, that might scare you a little. If she then told you that the tool always looks over your shoulder and tells you exactly what you should do next, you might not like it at all.

Few people want to be nagged all day by a robot, no matter how right the machine is. When implemented thoughtlessly, “Next Best Action” risks taking away the sales reps’ agency, which decreases autonomy and motivation. People don’t want to become robots. Some efficiency gains might not be worth reduced engagement and increased turn-over.

If you know the KARETIS team, you might have guessed already that we’ve looked for a solution that reconciles people with data and technology. The good news is that we think we’ve found an approach that achieves the best of both worlds.

Wants to know how? Read our detailed experience in implementing AI-driven next best action tools in the Part Two of this blog series.
Interested to implement it?  Contact directly any partner of Karetis.