Field sales
Put your sales team on the right doors with territory plans and visit lists ranked by Avra.
Field sales comes down to one decision, made every week for every rep: which accounts to visit, and in what order. This page shows how to frame that decision as a prediction over your sales history, how scores feed the visit list and the route, and how to measure the change.
The territory problem
Field sales is expensive. A rep can visit only a few accounts a day, and the wrong accounts cost days of selling time.
Traditional territory plans use static lists: “visit every account above R$1M in revenue in the South region.” Revenue doesn’t say who’s ready to buy.
Plan around what predicts a sale
Instead of geography and firmographics alone, plan territories around:
| Factor | The question |
|---|---|
| Propensity to buy | Which accounts look like the ones that bought before? |
| Growth | Who’s expanding and likely to need more? |
| Competitive exposure | Which accounts are connected to your competitors’ customers? |
| Warm paths | Where do your existing customers connect to a prospect? |
The data
Avra RFM learns what “ready to buy” looks like from your won and lost deals, account lists, and visit history. A typical schema ties each account to its territory, the visits it received and the reps who made them, and the deals it closed. The schema below is illustrative; see Relational data for how the RFM reads yours.
Deals give you the outcome, and visits record the effort that went into each one. Every table also carries its snapshot time, so a backtest sees only what your CRM knew on each date. See Time and history.
| Data | Purpose |
|---|---|
| Sales outcomes | Won and lost deals, deal size, and time to close per account |
| Account list | CNPJs of target accounts, with territory assignments |
| CRM activity | Visit history, engagement, and pipeline stage |
For Brazilian companies, Avra GFM embeddings can add context from outside your data, such as a company’s owners and sector, to your own models.
The prediction problem
Each question names who is scored, the outcome, and the window:
| Question | Who is scored | Outcome | Window |
|---|---|---|---|
| Who buys | Prospect accounts in the territory | A deal won | 60 days |
| Who stays | Accounts that just bought | Still active | 90 days |
| Who’s ready for more | Existing customers | An expansion deal won | 90 days |
A target that counts only sales sends reps to accounts that buy and then go quiet. Adding “still active” to the outcome sends them to accounts that stick. See What you can predict.
A ranked visit list
An illustrative daily list for one rep. The score comes from Avra; the notes are what the rep knows from your CRM:
| Rank | Account | Score | Rep’s notes |
|---|---|---|---|
| 1 | TechFlow Ltda | 94 | Growing fast, connected to three of your customers |
| 2 | Indústria Beta | 89 | Expanding, no competitive exposure |
| 3 | Comércio Delta | 82 | Strong fit, new decision-maker |
| 4 | Serviços Gama | 71 | Stable, routine check-in |
| 5 | Distribuidora Omega | 68 | Slight decline, watch for churn |
The rep works the list top-down, so the strongest opportunities get attention first.
Ranking plus routing
Distance and eligibility still shape the route. The score decides which accounts come first:
Score every account in the territory, keep the top 20 by score among those that are active and reachable, and optimize the route through them. The rep visits the highest-ranked accounts on an efficient path, and each visit’s outcome goes back into your CRM, where it becomes history for later predictions.
How it’s evaluated
Before reps change their routes, the model is backtested out of time on your sales history and compared with random and firmographic ordering. In the field, compare scored routes with unranked ones over the same period. See Evaluation.
| Metric | What it tells you |
|---|---|
| Lift at the top decile | How much better the top-ranked accounts perform than random or firmographic ordering |
| Precision at your visit capacity | Of the accounts a rep can visit in a week, how many buy |
| Visits per closed deal | Conversion from scored visits compared with unranked visits |
| Revenue per rep | The territory-level effect of ranked visits |
| Time from first visit to close | Whether ranked accounts close faster |
| Deal size, field compared with inside sales | Whether field time goes to the accounts worth a visit |
Integration
| Moment | Path | How it’s used |
|---|---|---|
| The weekly territory plan | Batch inference | Score every account in each territory and load the ranked list into route planning |
| A rep adds an account | Live predictions | Score it on the spot |
| Prospects like your customers | Embeddings | Find prospects that resemble your customers by structure, for Brazilian companies |