Manaz Afzal
Retail & E-commerce

The week’s call list: trade customers about to reorder

Reorder calls for a growing wholesaler, tested on 47 weeks of real invoices against the CRM due-date rule.

63%of 100 weekly calls reach a customer about to order
19%for the due-date rule
68%of reorder value at the rule’s list size
31%for the rule

Executive summary

A weekly reorder score puts 63% of 100 calls on trade customers who order within a fortnight, against 19% for the CRM due-date rule.

63%of calls reach a customer about to order
19%due-date rule
68%of reorder value at the rule’s list size

Situation

Account managers call customers past their usual gap, and most calls land at the wrong time.

Approach

A reorder score each Monday from invoice history, tested on 47 weeks of real wholesale invoices.

Result

63% of calls on customers about to order, against 19%.

The problem

A wholesaler’s inside sales team calls trade customers from a list of accounts, usually alphabetical or by last year’s spend. A customer who usually reorders every three weeks and has gone five is the call that matters. Distributors that give reps a daily list of customers due to reorder report more orders per rep and revenue they can trace to the calls.

I measured how well a weekly ranking by each customer’s chance of ordering in the next fortnight puts the coming orders at the top of the list, against calling the biggest accounts first.

Team
Owner, account managers, the person who runs the CRM
Scale
A few thousand trade customers; a few hundred calls a week
Systems
HubSpot or Pipedrive tasks, Shopify B2B or Xero invoices, a phone and email sequence
Measure
Calls that reach a customer within 14 days of their order; order value reached per call

What I built

A reorder score refreshed every Monday from the invoice history. For each customer with two or more orders it reads the days since the last order against their own usual gap, how regular their gaps are, their order count over three and twelve months, order value and what they did in the same weeks last year, then ranks the chance of an order in the next fortnight. The top of the list becomes HubSpot call tasks for the week, with the customer’s usual gap and last order on the task.

InvoicesOrder history bycustomerScoreChance of an orderin the next 14 daysHubSpot tasksWeekly call listfor each managerOutcomeOrders and call notesfed back
The score reruns every Monday on the latest invoices.

Results

I tested on the UCI Online Retail II data: real invoices from a UK online giftware wholesaler, many of whose customers are wholesalers and shops, 33,107 order-days from 5,878 customers between December 2009 and December 2011. The score was fitted on weekly lists for 2010 and tested on the 47 Mondays of 2011, each covering about 3,166 active customers.

Score, 50 calls a week75%Score, 100 calls a week63%Score, 200 calls a week52%Score, 710 calls a week32%Due-date rule, 710 a week18%
The shorter the list, the higher the hit rate; every list size beats the due-date rule.
Due-date ruleReorder score50 calls a week75%100 calls a week63%200 calls a week52%
A weekly list of 100 calls ranked by the score reaches a customer who orders within a fortnight 63% of the time, against 19% for 100 customers from the due-date list. Data: UCI Online Retail II, Chen, 2019
Weekly call listCustomers calledOrder within 14 daysShare of reorders reachedShare of reorder value
Due-date rule71018%28%31%
Reorder score, same number71032%51%68%
Reorder score, top 10010063%14%31%

At the due-date rule’s own list size of about 710 customers a week, the score reaches 51% of the customers who reorder against 28%, and 68% of the order value against 31%, because it favours regular, larger accounts.

Data and method

Data
UCI Online Retail II (CC BY 4.0): every invoice line from a UK online giftware wholesaler, December 2009 to December 2011. Cancellations and lines without a customer removed; orders collapsed to one per customer per day. Customers appear only as numbers in the release.
Population
Each Monday, customers with at least two earlier orders and an order in the last twelve months
Target
Customer places an order in the 14 days from the Monday
Methods
Gradient-boosted classifier on recency, the customer’s own gap statistics, order counts over three and twelve months, order value, the same fortnight last year and the month; fitted on 2010 and tested on 2011
Comparison
Due-date rule: days since the last order between one and two times the customer’s median gap

Limitations

How I would take it live

Financial impact

Calls that land before an order

The gain is calls that reach a buyer while they are ordering. Call count, order value and the uplift a call adds are assumptions; hit rates are measured.

InputValueSource
Calls a week100
Average order valueA$450
Basket uplift when a call lands10%
Hit rate, score63%Measured
Hit rate, due-date rule19%Measured

Calculation

  1. Extra calls landing before an order a year2,141

A$96,327 a year in added basket value

A held-back half of the list measures the real uplift.

Implementation and rollout

Five weeks from invoices to a tested call list

  1. Week 1

    Export

    Two years of invoices from Xero or Shopify.

    Owner, Manaz

  2. Week 1

    Fit

    Score fitted; weekly call count agreed.

    Account managers, Manaz

  3. Weeks 2 to 5

    Test

    HubSpot tasks each Monday, half held back.

    Account managers

  4. After a month

    Adopt

    Orders and value per call compared.

    Owner

Tools

HubSpot or Pipedrive, Xero or Shopify B2B, Email sequence, Python, LightGBM

Team

  • Owner owns the target
  • Account managers make the calls
  • CRM admin tasks and fields
  • Manaz score and weekly refresh

The timeline is a plan for a typical engagement; it depends on data access and how many systems are involved.

Risks and mitigations

What could go wrong in production, and the control for each

RiskMitigation
Calling buyers who would order anywayHeld-back half measures the uplift.
New customersLeft to onboarding.
SeasonalitySame weeks last year in the score.
Over-callingOne call per customer per fortnight.

Key takeaways

Key takeaways

  • Due-date lists mostly reach the wrong week.
  • Each customer’s own rhythm matters.
  • Larger regular accounts rise to the top.
  • Test with a held-back half.