I’ve been reading quite a few books on economic incentives, monitoring, and enterprise change management — so goals vs outcomes have been at the top of my mind. The other thing that I’ve been thinking a lot about is how the incentives structures change between private for-profit organizations and public sector organizations.
In many ways, running a private organization is effectively much easier, because most decisions can be made through an economic lens and the success of a particular product or service is quite obvious because the market will direct money towards good ideas and away from bad ones.
If you create a terrible product that consumers hate, this will be clearly shown in the profit and loss statement at the end of the quarter or year.
There is no way to hide — unless you do an Enron.
And so, there is a natural tendency to be efficient, because managers at all levels in a private organization tend to measure their outcomes based on their profit and loss.
One of my private-sector customers runs a chain of pizza restaurants that revenue tens of millions a year. Each manager of each restaurant has a natural incentive to ensure that they do not spend money in a wasteful manner, because this would reduce the profit and loss of their restaurant, and likely directly affect their future earnings and career prospects.
When it comes to running a public sector organization, here things become more complex. How do we get a clear measure of success? And this is not just on a per-program basis, but even across programs?
This second point of cross-program evaluation is perhaps even more important than just monitoring individual programs for effectiveness because there is always a scarcity of resources.
These resources are money, management attention time, donors, and even the amount of physical work-hours available to the organization as a whole (and how much it can sustainably grow in a given year to cope with demand).
Based on my discussions with a significant number of individuals who work at public organizations, I don’t think this way of thinking is used. Each program has its own definition of success and impact, and this is not necessarily tied to how many resources are spent to achieve those results. Resources that could be spent on other initiatives.
I can give a trivial example here.
Let’s imagine that we have a web platform where users can log in to view data. Let’s imagine that it received 71,000 unique visitors last year.
I am not sure what to make of that. Is that a success? Is that a failure?
It’s difficult to know — because we don’t know the effort involved to build the platform Also, we don’t know if there was a goal set before the platform was built to get a certain number of visitors per year.
Hypothetically, let’s imagine that the platform costs $10,000,000 to build and $1,000,000/year to maintain.
This would result in $11m of cost for the first year, and this may not even include other secondary costs like staff time. If we had 71,000 visitors in the first year, this would result that each visit costing the organization $154.
But again, this really is not that useful, unless we can compare it to the value that is generated by each successful outcome. Perhaps, on average, each visitor to the platform receives $1,000 of value, which means that spending $154 to provide that value is good — and the initiative can be said to be a success.
And this gets to the point of why running a private for-profit organization is easier. The buyers of your products and services decide for themselves if what you are selling provides more value than the price that you are selling it at.
These types of free economic transactions are, by their very nature, win-win. You are selling something at a higher value than what you think it is worth (otherwise you wouldn’t bother selling it), while the buyer is buying something at a lower value than what it is worth to them (otherwise they wouldn’t make the purchase). Everyone gets a good deal based on what is valuable to them.
So, we’re starting to see some of the building blocks of how we could map out a standardized model to understand value transfer in the public sector, without having to charge end-users or run an organization on a for-profit model
So what’s required?
A Cost Model.
This is an accurate description of the total setup and running costs of any program, and ideally, this includes supplementary costs like staff time, and perhaps even a share of operational costs.
I’m not sure, at this stage, what to think about capital expenditures vs operational expenditures, but this is something that we can dig into later.
What is interesting here is that we would not need an extremely granular level of data. For instance, knowing the precise date of a specific transaction is not necessary, just which month it occurred in would be enough. Likely, various types of transactions could be grouped instead of split into individual transactions, which would make the cost model easier to follow.
A Transaction Model.
This is an understanding of how and when value is delivered by the program. In some cases, especially in the “offline” world, this might be rather straightforward. The World Food Programme might be able to count the precise number of food aid packages that successfully arrive in a country, or perhaps even the number of calories!
For digital products, this can be more difficult to measure. This is because there is the tendency to measure vanity metrics instead of metrics that truly measure value.
For instance, “sign-ups” is often something that is measured, when really what is more important is daily/weekly/monthly active users, often abbreviated to DAU/WAU/MAU. Or perhaps, the costs involved to successfully onboard a user in a given cohort.
These are typically the metrics that good venture capitalists ask from startups to cut through the bullshit and quickly understand if they should or should not invest in a startup.
Even better, if we can purely focus on the action that provides the value. For instance, an eCommerce website may focus on the number of transactions, and the value of each transaction.
So there are two things that we need to measure within the transaction model:
- The number of transactions.
- The average $ value per transaction.
The second can be difficult to measure, especially for the type of services provided where end users are not directly paying for the service. What’s the value of posting an Instagram photo, or of one scroll of the Facebook newsfeed?
The other problem is that often products or services, one successful, can actually cost significantly less than the value that they provide. A good example of this is your typical household fridge. This may cost a family in a developed nation $500. But, you would have to pay that family significantly more than $500 for them to give up using a fridge at home. And so, the ability to use a fridge is worth a lot more than what they paid for it initially.
Putting This All Together.
So we if merge these two models, we can reach something that we can call the “Cost-Per-Transaction” model. So, we can measure the cost required to create one transaction or one unit of value.
Once we have this data, we can start to do some really interesting things such as:
- Making predictions on the future cost-per-transaction of upcoming or potential programs, and making this the default model for conversations with donors, partners, and collegues.
- Comparing and ranking the efficiency of programs, and improving the allocation of scarce resources such as money and staff time.
- Finding out where the dead fish are. If a program is not submitting many transactions, perhaps there is something fundamentally wrong.
- Conversly, scaling successful programs to furhter help reduce the cost-per-transactions.
- Making changes to programs and being able to measure the (hopefully!) improved efficiency.
This is also very interesting with regards to Digital Transformation. It can help make the case for larger and faster investments in digital. If one measures the cost-per-transaction across channels, so for undigitized services as well as currently digitalized services, you can predict with a decent amount of accuracy how money (and time!) can be saved.
How Could This Work?
So how could this work in reality?
There is a big risk that if this is not managed properly it becomes full of “massaged” numbers that make everyone look good, instead of being a true representation of reality.
This is important because it is only by being able to understand reality that we can then build hypotheses and experiments to make positive change.
So let’s tackle how transaction data could be handled.
The first obvious step is that manual data entry should be banned. The only way to get data to a system is via real-time (or asynchronously in the case where technology has to work offline for periods of time) via some type of API. This means that the system gets the notice of the transaction precisely when the transaction happens.
In this case, we can then be sure that the number of transactions is accurate because the incentives of the people working on the programs will of course be to inflate the number as much as possible if we allow manual data entry. This is because a larger number of transactions results in a lower cost-per-transaction, which makes the program looks more efficient and successful than it really is.
However, this is not the only way that the system could be misused. A clear way to abuse things is to have the wrong type of transactions in the first place. Again, there is a strong incentive for anyone “in” the program to claim that transactions are happening much earlier in the funnel than what reflects reality.
So this would result in garbage data in, and garbage data out.
An example of this from the private sector would be measuring eCommerce sign-ups instead of actual purchases. As an eCommerce CEO, you don’t directly care about sign-ups, you care about how sales you are getting. The sign-up numbers are likely to be correlated to your sales, but that is not always the case. If you have enough signed-up customers, you are far better off working on customer loyalty and retention vs trying strategies to get even more sign-ups.
The for-profit incentives in the private sector will automatically solve this problem. Or, at least, the market and economic realities will eventually harshly punish those that ignore it — think of Adam Neuman and the WeWork fiasco.
So, how do we solve this problem for non-profit organizations? This isn’t straightforward at all, but it comes down to someone who is outside of the program, and not directly impacted by the outcome, to negotiate with those who are inside the program on what the right transactions to measure are.
Perhaps, once this initiative has been running for a long enough time, there could be standard transaction implementation based on different types of digital products or even programs, and teams are limited to choosing between those unless they have a very innovative program that goes outside of the typical boundaries.
The next challenge is to then accurately record the costs. Obviously, the best solution would be an automatic integration into the financial system to record and categorize all related program costs, but this is likely to raise a significant amount of complexities.
Often, digital initiatives are only a part of a wider program, and so it would be unfair to place all the program costs inside the cost model, and this would create artificially higher transaction costs.
And so, a manual interface to add the cost inputs would be required. Of course, this runs into the same type of problems described above if we offered a manual interface for transactions, but I believe that the scope for fudging numbers is a lot less.
One can conceivably imagine scenarios where transactions are miscounted by one or several orders of magnitude, but it is unlikely that costs could be underreported by a factor of ten or a hundred, without this raising red flags.
Because the main driver of efficiency is the number of transactions, a mistake or deliberate under-reporting of costs by 30-50% would not be as damaging to the quality of the data as one might imagine.
We could also assume that while the data is not perfect across all projects, the relative data would still be quite accurate, even if everyone is incentivized to under-report costs where possible.
The final solution to this problem would of course to either have the data independently audited, or to have an external party in charge of adding the data in the first place.
So, if we imagine a future where these, and various other, challenges are solved, what do we end up with?
This would be a data platform that could be used to analyze and deep-dive into individual programs, as well as compare and aggregate data between programs. Interestingly, the grouping of programs by country, initiative and organizational goal.
The final point to consider is whether we should add a “value” to each transaction. Again, the incentive here would be to inflate the value that each transaction provides to end-users.
Perhaps a better way would not estimate this at all, but to let donors, managers, and evaluators decide if the value created by each transaction is more than the cost.
A lot more work is required to see if this idea is practical, especially in large multi-national organizations. However, the picture of perfect is promising — this would create a new language and way of measuring and understanding impact.
It is not a panacea, but would help to significantly improve efficiency and operations across any organization where profit is not the motive, including NGOs, Governments, charities, and perhaps even educational institutions.
The real test if this idea works would be to try it, but also to abstract any system that is developed to track it and make it a Digital Public good, as well as a commercial SaaS (Software-as-a-Service) platform.
After all, if the idea works, then there will be a demand for it to be used, and this almost takes the whole cost-per-transaction model and applies it on itself. In fact, it is likely that within an organization, the cost of running the platform and the number of transactions should itself be showcased within the system!
That is always the ultimate test for an enterprise digital product. If it’s used by the internal captive audience that has no other choice, then it is difficult to know if it is truly successful. If it is used by external parties who have a choice — then this gives you a much clearer definition of success.