When your customer owes you money, marketing changes its metric

Everyone understands what that changes in theory, and almost nobody applies it in marketing. A sale is not revenue, it is a receivable. An acquired customer is not an asset until they have repaid. And the only question finance cares about is not how many customers were signed this month, it is how many of them are still paying three months later.

The diagnosis that turned everything over

One number was put in front of the market directors that autumn. The company's unit economics assumed an average collection rate of 80 percent over the life of the contract. The reality: three months after the sale, close to half of customers were already below that threshold. And half the sales agents were delivering exactly those customers.

Nothing in the commercial system addressed the distortion. Agents were paid on volume. They were doing precisely what they were paid to do.

Marketing was rewarding the wrong behaviour

There was a recognition programme for top sales agents. It rewarded the number of sales in the month, nothing else. I looked at the actual distribution: across 1,245 agents measured over one month, 1,016 made fewer than ten sales. Leadership wanted a target of thirty sales per agent per month. Thirty-five agents out of 1,245 reached it, 2.8 percent of the sales force.

I said no, with that number. Then I proposed something else: twenty sales for the agent, thirty for the shop supervisor, and in both cases a new condition, a collection rate above 80 percent on customers at three months. The target of thirty did not disappear, it moved up one level. That was not a refusal, it was a counter-proposal, and it was implemented.

The month one document served both to decorate and to police

The switch is dated. Until July, the award nomination file held one performance column, the number of sales. From August, two things appear in that same file. A three-month collection rate column, sitting next to the number of new customers. And a second list, beside the winners, naming the agents with high volume and poor collection.

One agent appears there with 146 new customers and a collection rate of 7.3 percent. Another with 123 customers and 67.8 percent. Volume alone had stopped being good news.

On the customer side, the constraint dictated the reward

In parallel we built a customer loyalty programme, segmented on the cumulative collection rate of the last three months. Four bands. The written objective was to bring the middle bands to the payment habits of the top one, and it carried an explicit constraint: without touching cash flow.

That constraint produced everything else. No reward could be an object, because an object has to be paid for. So the rewards became services with near-zero marginal cost: a discount conditional on an upgrade, technical support already provisioned, a delivery, priority service in the shop. In one market the message went straight to the point, one day of free power if you pay on time. A day of power on a live contract does not leave the bank account.

And the negative side was written with the same precision: a higher minimum reactivation amount, service fees for slow payers, fewer days credited than days paid. In the shop, the benefits poster had a twin, the consequences poster. In consumer credit, refusing to carry the unpleasant half of the message makes the rest inoperative.

Three failures out of four

The customer programme was piloted on four cohorts. Only one lifted collections, and it was the smallest and most qualified: a little over 1,200 customers on the same product, collection rate from 79 to 93 percent, average daily collection up 17 percent.

The other three failed, each for a different reason, and each one identifiable. On the first, the problem was deliverability: message delivery fell from 84 percent to 26 percent in four weeks, and across the period, on average, 48 percent of customers received the message and 37 percent paid. On the second, the problem was targeting: the cohort labelled in the top band was in fact in the bottom one, its prior payments put it below 50 percent collection. We sent a reward message to customers in arrears.

On the third the result is ambiguous, and the reading is what counts. Monthly collection falls 45 percent. But the average amount per payer goes from $0.65 to $1.25, a rise of 92 percent. Fewer people pay, and those who pay, pay considerably more. That is not a success, it is a shift, and it had to be said that way.

Phase 2 contradicted phase 1

The pilot ran on selected cohorts. In May and June the programme was extended to the whole customer base, and the result went the other way. On one market, collections fell 45 percent. On another, 6 percent. The third was not measurable, for want of clean data.

One positive effect survives the extension, and it is the most instructive thing in the file: a rise of 1.7 percent on the lowest band, the worst payers, and only on the constraint messages. Not on the reward messages.

In other words, what moved the line was not the promise of a benefit. It was the stated consequence. A customer who pays badly does not react to a gift, they react to what happens if they do not pay. That is counter-intuitive for a marketer, and it is written in the numbers.

I put the decision on paper: pause the programme for four weeks to isolate seasonality, rewrite the messages for the middle bands, stop contacting the top band which needed no prompting at all, and restart on one market only. I stopped a system I had designed, on the strength of my own numbers. It is the decision from that period I am happiest about, and the least flattering.

Version 1 of the report said the opposite of version 2

This is the part of the file I talk about most readily. The first version of the pilot report announced, on one market, a 40 percent rise in payments, and concluded the strategy had worked. On checking the baseline, the rise turned out to be a 40 percent fall. The collection rate had actually gone from 51 percent to 36 percent.

The corrected version is the one that was signed and kept. A number in a report is only worth what its baseline is worth, and a baseline is checked before circulation, not after.

What execution taught, and no plan says

The programme stopped in one market because the shops were out of stock, in another because cash was tight, in a third because two commission systems collided. None of those is a marketing reason. They were my problem all the same.

The criterion bit too hard. Over one quarter, only four winners across twenty-seven possible slots. A programme that rewards almost nobody stops being a programme.

And the most useful lesson: the agent was now assessed on whether their customers repaid, but the script we handed them said nothing about payment regularity, contract length, or what happens when you fall behind. It sold access. Its central line, you only pay for what you use, actually planted the opposite of what we were asking for. Changing the assessment criterion without changing the sales pitch is asking someone to win a game while leaving them the old rules.

What this is worth elsewhere

This reasoning is not limited to prepaid energy. It applies to any business whose customer carries a payment obligation over time: instalment sales, microfinance, recurring-premium insurance, subscription with commitment, and mobile money the moment there is credit, float, or an agent commission.

In those businesses marketing has a natural metric, and it is neither reach nor awareness. It is the collection rate, the amount collected, the average amount per payer, the number of payers. Those are lines finance already reads. Marketing that speaks that language stops being a cost centre in the budget conversation.

And the question to ask before launching anything fits on one line: what are you rewarding today, and is that really the behaviour you want?