🧩🖊️🌳  Dear Agency Founder

A Guide to Forecasting for Small Agencies

👀 View the example forecast here

Two forecasts

The type of forecasting you do and the data you use depends on the questions you want it to answer.

Too simple and you won't trust it. Too complicated and it's hard to adjust and understand.

I think there are two forecasts worth making.

Forecast 1: The next three months

For this forecast, use as real numbers as possible. We want to estimate how much profit we're going to make in that period.

We should be able to see where the money's coming from: existing clients or pipeline. Costs should be predictable over that shorter time scale too. In three months, unless you're actively hiring right now, you're not changing your team shape much. And if you are, you probably know by how much.

What's good about this forecast?

You understand business performance right now in actual terms. You put down revenue numbers you believe you can hit, then check if you hit them.

You start to learn the difference between what you think is going to happen and what actually happens.

You can either find the reasons it didn't happen and improve them (i.e., utilisation was low because we didn't make clear how much time people should be spending on different projects), or find something you can't improve and just factor it in next time (e.g., we calculated based on utilisation that's too high because people are waiting for client feedback).

Forecast 2: The next financial year

This forecast is more of your model. Instead of exact numbers, we're going to use a budget.

You need to set that budget using your best prediction:

This is the thing that makes sure your plans for growth are viable, that they lead to more profit, and that they aren't overly risky.

If you've got a revenue target, you can move it around and use the model to check what the minimum you need to hit to break even is. You can see what salaries you can afford, what bonuses and pay rises you can afford, and what you can take out yourself.

Modelling the different parts

Both forecasts need the same building blocks. Here's how to think about each one:

Known revenue

Your ongoing projects and new projects you are almost certain of. The amounts probably come from proposals and agreements with current and soon-to-be-current clients. What you're really forecasting is when the money arrives.

Tip: Whatever pricing model you use (hourly, value-based, fixed cost) the money will always arrive later than you think. But this is our first pass, so you can be optimistic. One of the great things about a forecast is it shows you how good you are at keeping revenue flowing as you expect.

The model needs to account for how you earn the revenue.

If it's time and materials, the model needs a sense of how much resource you have to assign to each project, and therefore how much money can be brought in each month.

If you're value-based, you still need a sense of capacity. How are you going to model how quickly you can achieve that value and bill for it?

You might be fully automated and run with AI agents, but you do have a bottleneck. That bottleneck should be represented.

Pipeline revenue

We split this into two sections: active pipeline and future pipeline.

Active pipeline includes the inquiries you currently have at different stages, from first contact through to proposal and contracting.

You may choose to factor really late-stage work (where it's been agreed and it's just sign-offs) into committed revenue. That's fine. It all depends on the risk of that stage.

As you grow and meet businesses large enough to have internal procurement and legal teams, the contracting stage comes with a degree of risk. You only need to be burnt once to treat everything before signature as still a risk.

But there's pipeline that definitely fits here. First leads in, stuff where you're in a proposal and up against other agencies, etc.

Modelling pipeline revenue

The simplest way is to write down all the deals, their value, and your best guess of how that revenue might spread out.

Make your best guesses and put it down onto your forecast.

Weighting your pipeline

You aren't going to win everything. Your model might include a weighting: a percentage of certainty.

You can be instinct-backed. Just ask yourself each time what you think the chances are. You'll always be more optimistic than reality because you're an entrepreneur and you're cursed with thinking things are going to go great. So counterbalance your weighting with a little pessimism.

For more sophistication, be data-backed. How many leads do you get in a month? How many convert? That's your lead-in percentage. How many proposals do you send and how many convert? That's your proposal stage percentage.

Choose what level of modelling you need and tie it back to how risky the prediction is. Let that drive how instinctive or data-backed you are.

Future pipeline

How far into the future do you know where your work is coming from? Not a whole year. Deals that will be in your pipeline in 6 months time aren’t in the market yet.

So we need a way to forecast how much revenue we’ll make off work that doesn’t even exist yet.

At this point it gets easy to play spreadsheet millionaire and give yourself some record high nonexistent revenue months. And look, part of the point here is that we set a target to grow and the model lets us see how much we’ll have to invest in wages and marketing to get that many leads and deliver that much work.

For future pipeline you again have levels of sophistication. Most of what you’ll be modelling is revenue growth and then working out the business that supports that profitably.

You could take your year on year revenue growth and then extend it another year, or you could make a plan. How much are you going to invest in growth and what return do you expect?

And now you have the problem of deciding between whether you are modelling what you want to happen or what you think will happen.

Welcome to forecasting.

Once you have your model you’ll be able to run all the scenarios you want and it will tell you whether to worry or start planning that retirement.

Cost of delivery

For a simple P&L, this is what you're spending on freelancers and third parties to deliver work.

The rule: if you're forecasting revenue for anything, you need to forecast the cost of delivering that revenue.

If you've got a big pipeline project that needs a development agency, put their costs in. If you're weighting that project at 50%, put 50% of the costs in.

Alternatively, from those big pipeline projects, only forecast your gross profit. If you think someone's going to start paying you 10K a month but you're going to spend 3K on ads, you could just use the 7K.

Salaries and staff costs

Worth separating salaries out because they'll potentially scale with pipeline work if you're planning on growing. They're a little harder to model than the rest.

When we think of staff costs, we think of the salary. Divide it by 12 and that's how much someone costs. But there are other costs: taxes, pensions, other commitments.

Another staff member will raise your overheads. Software, bonuses, pay rises, hiring costs.

And then there’s your compensation.

If you're doing that through payroll or dividends, your model will be slightly different. But for the sake of forecasting, include at least the minimum you want to take home each month.

General overheads

These normally don't shift month to month unless there's something coming, like a new office.

The rest of your overheads are probably fairly stable. Take an average of what you're spending, unless you're planning a big change.

The place things may fluctuate is in investments you make into things like sales and marketing.

The best strategy here is to set a budget and track over time what you are spending against it. That way you know the business performance with that spend.

How to run your business with a forecast

Reviewing and adjusting the forecast should be a ritual. Every month you should get together and track how things turned out against it.

This will force valuable conversations about how close you are to executing your plan. What isn't happening and what tactics need to change.

You can also adjust it to see what's likely to happen given what you've learned since you made the forecast. You can keep that adjustment and start tracking against it or maybe you want to create a scenario for discussion. What happens if that client leaves? Can you afford 2 new hires?

Forecasting is a muscle you train, how you use it will be unique to your agency. As a founder it's an invaluable skill.

Even if it's just highlighting how bad our predictions really are.