Nothing is more demoralising for a team than to constantly be chasing impossible targets. And on the odd occasion the target is met? The next target gets pushed even further out of sight.
Worse, bad forecasts will leave the business with too much or too little stock, unhappy customers and in extreme situations, can kill the business entirely.
You want your ad budget to be working as hard as possible too. Nobody wants to waste cash pushing for an impossible target. But you don’t want to leave a ton of opportunity on the table either (nothing bugs me more than a case study showing a blended ROAS of 30! Opportunity lost!)
So… how much should you spend? The million dollar question.
Here’s how to think about the four basic parts of a good forecast – and important questions you’ll need to ask – that will help you create an effective digital marketing forecast.
All of which will ensure your marketing team can constantly strive to be ever more strategic and not constantly reacting to today’s numbers (more on that another day!)
Those 4 parts are:
1 What is a forecast and why create one for your digital advertising? What’s your ideal outcome and which KPIs should you consider? What are you trying to predict (& why)?
2 What do you need to consider before you start? What assumptions will you make? How the 80/20 rule can be your friend. What about competitors & external market conditions?
3 How to actually build a great forecast – rules of thumb. How detailed should it be?
4 Reforecasting and comparing results. Getting better over time, with continuous improvement.
Part 1 : What is a forecast and why create one for your digital advertising?
A forecast (in this context) is simply a model of the future that helps you to predict what’s most likely to happen (so you can order the right amount of stock, get staff levels right, move budgets to the best performing channels etc).
The first step in creating an effective digital marketing forecast is to determine what you’re trying to predict and why. For example, do you want to forecast daily website sessions, weekly ad spend or monthly revenue?
Choose one KPI for now so that everyone involved in creating the forecast understands what you’re trying to predict and why. Sounds obvious, but without agreement upfront there will be confusion about what information needs collecting and how to interpret the results.
Generally you’ll want this to be a shareable spreadsheet so that you can collaborate on assumptions (please write them all down) & update your data as you learn more from actual results.
Once you know what you’re predicting, you can gather the historical data that’s going to help and add your assumptions about how the future will differ.
For example if you know that last month you invested $100k in Google Shopping ads to make $500k in revenue, and next month you’re aiming to double your budget. The maths of diminishing returns (we’ll skip the formulae for now!) will suggest you’re likely to get ~40% more rev (a total of ~$700k).
Knowing in advance that your ‘return on ad spend’ (aka ROAS) will drop from 5.0 to 3.5 might encourage you to look for other channels in which to invest.
And even if you hate spreadsheets, once you’ve got a model setup, it’ll be easy (& fast) to change for future months – it’s worth the struggle the first time you do this!
Part 2 : What do you need to consider before you start?
Before we can start building a forecast model, we need to consider some things:
What assumptions are we going to make about our marketing data? For example, if I’m forecasting my PPC campaigns’ performance over the next few months but don’t know the exact dates for future sale periods, how do I deal with that uncertainty?
How far in advance do you need to forecast? Sometimes a manager just wants to know what’s likely to happen next week. The board might want a year or more. It’s all possible (error margins grow with time of course), but knowing your timeframe will dictate many of the assumptions you need to make.
No forecast is ever going to be perfect. It’s a model of the future. So don’t bash your head against the wall aiming for 100% perfection. An 80/20 approach will suffice. i.e what’s the 20% you can do effectively and efficiently that will get you 80% of the ‘right’ answer?
For example at WebSavvy we’ll often predict the likely spend for each ‘bucket’ of ad spend for the coming month and the likely revenue for each of those buckets.
We therefore need to make assumptions about the expected daily spend for each bucket, how sales might affect that, are there any holidays in the month ahead? No single day will be perfectly accurate, but the overall outcome for the entire month will often be very close to the actual.
You’re aiming for useful, not perfect!
You’ll also want to consider stock levels (if the top seller is about to go out of stock, that’s going to affect things).
What are competitors’ doing? Do they have a sale running? How do their prices compare to yours?
And of course the macro environment will have an impact too. If mortgage rates are over 7% and consumer sentiment is the lowest it’s been in 14 years, your ROI is likely to be impacted! So any assumptions that have gone into your model need to be clearly stated.
Part 3: How to actually build a great forecast
As an agency that is focused on investing our client budgets as effectively as possible, we typically start with the amount we believe we can profitably spend (per bucket) & get to an expected revenue figure from there.
However this is clearly at odds with most retailers, who will start with an expected (desired!) amount of revenue and then ask ‘how much do we need to spend to hit that target rev?’
Over the years we’ve adapted our ‘spend-first’ approach to better mesh with our clients’ goals, so we’ll adjust forecast spend (for each of those buckets) to get as close as possible to the desired target revenue, while still giving us a reasonable chance of hitting targets & delighting our clients.
Again the 80/20 rule is your friend here.
Often we see businesses treating multiple different channels as one big ‘lump’ of spend – so when they double the budget, the expectation is that revenue will double too (or more!)
Sadly that’s not typically the case, but by breaking each platform into 3-6 ‘buckets’ we can build a much more
accurate useful model.
The buckets we typically use for Google are: Brand search, Shopping (Performance Max), Non Brand Search & Display/Video. For Facebook we’ll have buckets based on the traffic ‘temperature’ (from cold to hot) plus a bucket for existing customers (if it makes sense given the client’s product mix).
It’s a double-edged sword of course! Too many buckets and you’ll drown in assumptions and spreadsheet tabs… too few and you risk the prediction being wildly inaccurate.
The fun is finding that Goldilocks amount!
Part 4: Reforecasting & comparing results.
Continuous improvement is key. It’s a never-ending process, but we’ve gotten much better at forecasting over time. Our internal reporting, spreadsheet templates, Google scripts and other tools have helped streamline the process, but nothing beats experience and an open conversation with your client!
Of course making sure to compare a forecast against actual spend is critical to keep improving and (ideally) getting more accurate each month. This ‘backtesting’ helps us catch any errors or assumptions that slipped through the cracks during the initial estimate so we can fix them before doing any damage!
So continue to reforecast and fine tune your model. How often? Again, the 80/20 rule is your friend here.
If you’re building a monthly forecast, you don’t need to look at actuals and re-run the model every day. Weekly should be fine. Monthly might be ok. It’ll depend on the actions you’re likely to take as a result of a new forecast & how much those will impact your business goals.
Predictive analytics are your friend. It’s no longer enough to focus on historical data, the rear-view mirror, and report on “what happened”.
All retailers need confidence in the future and while nobody can perfectly predict what’s going to happen, we can keep trying to get as close as possible!
Our goal is always to identify the most effective channels and allocate budget accordingly to exceed our clients’ goals.
If you’d like help with your forecasting, get in touch – just knowing your current spend and results in 5-10 ‘buckets’ of spend is enough to predict the likely impact of a 50% increase or decrease in budget.
You might be spending in the wrong places, let’s see if we can make that existing budget work a bit harder!