Field sales reps spend the first minutes of every day deciding who to see, and they decide badly. This is the spec for a feature that answers the question for them: here are your five accounts today, in order, and here's why.
This document specifies the feature. Its experiment and metrics plan defines how we'd prove it works before a full rollout. Together they're the two halves of owning a feature: decide what to build, then decide how you'll know you were right.
Next Best Visit is a ranked list of accounts a field sales rep should visit today, generated each morning from data the product already holds. Each entry shows a transparent reason for its rank. The goal is to move reps from ad hoc route decisions toward a defensible, data-informed order of visits, without taking away their judgment or their ability to override.
A field sales rep manages more accounts than they can visit in a day, and the choice of who to see is left entirely to memory and habit. In practice that means three predictable failures. High-value accounts get visited on the same comfortable cadence as low-value ones. At-risk accounts, the ones going quiet before they churn, get noticed only after they've gone. And reps burn windshield time criss-crossing a territory because the order of visits was never planned around geography.
The data to make a better decision already exists in the product: last-contact dates, account value, open opportunities, and location. It just isn't assembled into an answer at the moment the rep needs one. The rep is asked to be their own analyst, first thing in the morning, in a parking lot.
Every field sales product competes on helping reps sell more in the same number of hours. Route and prioritization decisions are the largest uninstrumented lever left: they happen daily, they compound, and today they happen with no product support at all. This is a high-frequency decision the product is currently silent on.
| Goals | Non-goals (for v1) |
|---|---|
| Give each rep a ranked, explained list of who to visit today. | Full route optimization or turn-by-turn navigation. We order accounts; we don't draw the driving route. |
| Require zero new data entry to produce value on day one. | A predictive churn or win-probability model. v1 uses transparent, rule-based signals, not machine learning. |
| Keep the rep in control: every rank is explained and overridable. | Manager-assigned lists or top-down routing. This is a rep-facing assist, not a compliance tool. |
| Make the ranking logic legible enough that a rep trusts it. | Cross-rep or team-level optimization. v1 operates within one rep's book of business. |
The primary user is the same territory rep the rest of the product serves: in the field six to eight hours a day, moderate tech comfort, motivated to sell rather than to administer software. The use case is narrow and repeated. The rep opens the app at the start of the day, or between two stops, and wants a fast answer to "who's worth my time next?" The feature has to be readable in seconds and has to survive being ignored: a rep who overrides it every day should still find it harmless, not in the way.
Rank is produced by a transparent, weighted score over signals the product already stores. Transparency is a requirement, not a nicety: a rep who can't see why an account is ranked first won't trust the list, and a list reps don't trust is a list they don't use.
| Signal | What it captures | Direction |
|---|---|---|
| Account value | Revenue or tier of the account | Higher value ranks higher |
| Days since last contact | Recency of the last logged visit or touch | Longer gap ranks higher, up to a cap |
| Open opportunity | Whether there's an active deal in progress | Open deal ranks higher |
| At-risk signal | A previously active account that has gone quiet | Going quiet ranks higher |
| Proximity | Distance from the rep's current or planned location | Closer ranks higher, as a tie-breaker |
Each entry in the list shows its single strongest reason in plain language, for example "High value, not visited in 34 days" or "Open deal, going quiet." Proximity acts mainly as a tie-breaker so the list doesn't send a rep across the territory and back, but it never outranks a genuinely urgent account. The weights are configurable so they can be tuned during the experiment rather than guessed at once and frozen.
Prioritized so that if scope has to be cut, it's cut from the bottom.
The ranked five are the first thing the rep sees, each with its one-line reason. No configuration, no empty state to fill in.
Tapping an account opens it ready to log a visit. The feature hands off to the primary action instead of being a dead-end screen.
Dismiss anything that doesn't fit today's plan. The list reorders and remembers the choice for the day, so it never nags.
| Risk | Mitigation |
|---|---|
| Reps distrust an opaque ranking and ignore the list. | Every rank carries a visible, plain-language reason. Transparency is a P0 requirement, not a later polish. |
| Data quality is uneven: some accounts lack value or contact data. | The model degrades gracefully, ranking on whatever signals exist and flagging low-confidence lists rather than faking precision. |
| The list feels like surveillance or top-down control. | It is strictly rep-facing and fully overridable. No manager assignment, no compliance framing, in v1. |
| Default weights are wrong and the order is unhelpful. | Weights are configurable and validated through the experiment plan before a full rollout, not shipped as a guess. |
Stated here briefly; specified in full, with an experiment to test them, in the companion plan. The primary success metric is an increase in the share of visits that go to high-value or at-risk accounts, without a drop in total visits logged. In other words, reps visiting better, not just differently, and not fewer.
This PRD was written as a portfolio piece. The problem space, user, and signals are drawn from six years of direct experience in field sales technology. Any figures are modeled estimates meant to illustrate how I'd frame and scope a feature, not reported company data.