Economics and measurement
Designed backwards from value.
Start from what a change in behaviour is worth to you. Work out what can profitably be spent to cause it. Then measure what actually changed.
From spend to return
Follow the money all the way to value.
The question is not how many gigabytes were given out. It is whether the programme caused behaviour that is worth more than it cost.
Invested
What the programme spends on service, telecom and data.
Customers reached
The customers offered the plan.
Extra behaviour
The change against the control group.
More balances, transactions or salary
What that change looks like on your books.
Value to you
Priced with your own economics.
Return
Value against spend, measured, not assumed.
Scenario calculator
Test the economics with your own numbers.
Cost and benefit are kept apart. Every customer who qualifies is costed, but only the change against the control group counts as benefit. Start with the break-even point, then try a cautious and a stronger case.
Your scenario
Net result ₦1,500,000. Break-even needs 9.75% of the test group to change.
Cost: everything the programme spends
- Data delivered
- Data cost
- Service and operations
- Programme cost
- Per customer in the test group
Benefit: only the change against the control group
- Customers who changed
- Value of each change
- Benefit
- Net result
- Return for every 1 spent
- Break-even change
- Cost per extra outcome
How the sums work
- Weeks = months × 4.33
- Data delivered = customers × weekly qualifying share × weeks × GB per qualifying week
- Programme cost = data delivered × cost per GB + service and operations fee
- Customers who changed = customers × extra share who change against the control group
- Value of each change:
- Balances: extra balance × yearly value of deposits × months counted ÷ 12
- Transactions: extra transactions a month × value per transaction × months counted
- Salary: value of a salary account a month × months counted
- Benefit = customers who changed × value of each change
- Break-even change = programme cost ÷ value of each change
- Cost per extra outcome = programme cost ÷ customers who changed
Benefit counts only the difference between the test group and a matched control group, never everyone who qualified. Gigabytes given out, clicks and redemption rates are not counted as success.
Correlation is not impact
Some customers would have done it anyway.
Before-and-after numbers cannot separate a programme from everything else happening in the market. A matched control group can. The programme takes credit only for the difference.
Test group offered the plan
Matched control group not offered the plan
- Would have done it anyway. In the test group they still qualify, so they are part of the cost.
- The extra behaviour. Only this difference counts as the programme’s benefit.
- No change.
Set the baseline
Measure how the chosen group behaves before launch, so there is a clear starting point.
Form matched cohorts
Split similar customers into a test group, offered the plan, and a control group that is not.
Agree the period
Fix the evaluation period, the measures and the scaling criteria before the experiment starts.
Compare, then decide
Credit the programme only with the difference between the groups, then decide whether to scale.
What counts
Outcomes, not activity.
What we measure
- Extra balances
- Transaction frequency
- Salary accounts moved
- Cost per extra outcome
Not counted as success
- Gigabytes given out
- Clicks
- Redemption rates
Before a decision
What we would need from you.
A real business case uses your numbers, not ours. These are enough to build cautious, base and stronger cases for a first experiment.
- The size of the customer group you would start with
- How that group behaves today: balances, transaction frequency or salary credits
- What a change in that behaviour is worth to you, such as your margin on deposits
- Any incentives already running for the same customers
- The period your finance team is comfortable counting value over
What does a programme cost?
A starting programme is priced as one commitment that covers the Aura service, the telecom side and the data for the experiment. The amount depends on the size of the test and the behaviour, and is set out in your proposal.
How is the return assessed?
By following the chain from spend to value: the money reaches customers, some of them change behaviour, that shows up as more balances, transactions or salary accounts, those have a value to you, and that gives the return.
We agree the measures and the scaling criteria before the experiment starts, and build cautious, base and stronger cases on your own numbers.
Do customers who would have qualified anyway inflate the results?
They are part of the cost, which is why the comparison group matters. The programme only takes credit for the difference between the test group and the control group.
Build the business case on your numbers.
Share the behaviour and the customer group you have in mind. We will work through the economics of a first experiment with you.