Why read this? : We explore how to source and analyse data to make your e-Commerce forecast more accurate. Learn the impact forecast accuracy has on your marketing, supply and finance plans. Plus, we walk through an example from our first ever e-Commerce forecast. Read this to learn how to improve your e-Commerce forecast.
Any good e-Commerce expert will tell you, e-Commerce needs many different types of functional expertise to work together.
That includes supply chain and finance, who have to shift from their usual focus on efficiency.
That’s what traditional retail distribution channels do. Your products go on pallets in lorries from your warehouse to the retailer’s warehouse. They then ship them out to their stores with all the other products they sell.

Delivery costs are spread over truckloads of products, not individual deliveries. Customers then travel to the store to pick up the product. So there’s no last mile challenges. Customers do that themselves. It’s all very efficient from a supply and finance point of view.
E-Commerce deliveries are inefficient
Not the case in e-Commerce. Orders don’t go to a central location, they go to multiple different locations i.e. customer doorsteps. So, you manage the last mile. There’s more returns and damages. More work for your customer service team. Delivery cost per product is high. All very inefficient.
So, supply chain and finance people often put up barriers. You have to convince them there’s a sales opportunity based on what online shoppers want. That’s where the e-Commerce forecast comes in. It shows how much you expect to sell. This helps supply and finance people understand the opportunity and plan for it. It tells them what’s coming.
Forecast versus target
Though they start from the same place in the e-Commerce planning process, the e-Commerce forecast and the e-Commerce target are in fact different measures. They have different purposes.
The e-Commerce forecast is a prediction. How much you think you’ll sell. The e-Commerce target is a commitment. It’s your performance objective. Not the same thing.
Your forecast changes as you learn more about the market. But your target stays the same until it’s time to set a new target.
An example. Say your monthly forecast and target is 400 units. But, let’s say 2 weeks in, you’ve only sold 100 units. Your new monthly forecast would change to 200 (based on the first 2 weeks run rate). But your target remains 400. You need to do something to close the gap between your forecast and target.
The e-Commerce forecast
For functions like supply chain and finance, the e-Commerce forecast is a more helpful measure than the target, because it helps them plan.
It’s a more accurate view of future sales because it updates as you get more data.
They need to know how much you expect to sell because their plans depend on the numbers of future sales.

The e-Commerce forecast gives them time to prepare and react. As it adjusts regularly, it always gives the latest view with the most up to date information.
(For more on the e-Commerce target, check out our managing an online store guide where we talk about its role as a motivation and reward mechanism).
The e-Commerce forecast and resource planning
At the start of your e-Commerce planning, your forecast helps you plan resources. As you start to sell, you use it to manage and adjust those resources.
You use it in your :-
- business case – to secure resources, especially budgets, by showing the return (in sales and profits) on investment.
- marketing plan – to set the business context and schedule resources in your activity calendar.
- profit and loss – to track your financial performance, and identify when variances occur.

- e-Commerce dashboard – to track financial and marketing Key Performance Indicators (KPIs).
- manufacturing plan – to plan how many materials to order to produce finished goods.
- people plan – to plan team work schedules, both for manufacturing and service led businesses.
Start the e-Commerce forecast by gathering data
Forecasts get more accurate the more they’re based on data. It’s possible to go on “gut feel”, but unlikely you’ll have an accurate e-Commerce forecast if you do. So, start by gathering data.
In particular you’re looking for trend data. Past and current trends which show how different activities and events impact what customers buy.
Use these to make informed estimates about what might happen in the next 12 months. (the usual length of most forecasts).

You assume that if an activity had an x% impact in the past, and nothing else has changed, it’ll have the same x% impact if you do it again. You also make assumptions about what competitors will do based on their past actions. Collect these assumptions and use them to build your forecast.
The more experience you have of the market, the better the data you’ll have for your forecast. That’s why established online sellers will have more accurate forecasts than newer online sellers.
They can analyse historic sales data to see the impact of different activities and events. They can use this to predict future sales.
Analysing sales - linear regression
The simplest forecast would look at historic sales and assume the trend will continue. For example, let’s say, 2 months ago, we sold 100 units. Last month, we sold 150 units. A simple forecast for this month would be 200 units to keep the trend going.
The more months of sales data you have, the more accurate this type of forecast. You can use a statistical technique called linear regression. This draws a best straight line through your historic data points.
You can then project this straight line into the future. It gives you a good overall direction for the year. But it’s less good for working out month to month variations. It also can’t take account of anything new that’ll happen in the next 12 months. It’s a good start point for your e-Commerce forecast, but ideally you’d do more analysis.
Analysing sales - Correlations
This deeper analysis usually involves activity data.
Advertising spend, for example. Visitors to your store website. Sign-up and loyalty rates from your CRM program. Competitor activity. Price discounting and sales promotions data.
You look at the trends of these data, and compare them to your sales trends.
Find strong correlations between activities and sales, and you know those activities influence sales.

You can then estimate more accurately how much each activity will influence your e-Commerce forecast.
For example, you work out that an x% price discount delivers a y% increase in e-Commerce sales. So you adjust your forecast by that amount each time you run that discount.
Analysing sales - econometric modelling
If you’ve got the budget, you can also run econometric modelling. This helps you get to a more accurate e-Commerce forecast.
It’s a highly specialised statistical analysis involving large amounts of data and advanced statistical techniques like cluster analysis. However, it takes time to gather the data. You need statistics experts to build the model. This isn’t cheap. It’s usually only bigger e-Commerce players who can afford this.
While it’s the most accurate predictor of sales, even this approach can’t guarantee 100% accuracy. Because of the time it takes to run, the data might be out of date by the time the model runs. Plus, it can’t account for new events which have no historical data e.g. a competitor launching a new product.
Manual review of the forecast - Bayesian
Whatever approach you use, you should still manually review the statistics driven forecast. Does it make sense? Do you know of any “new” events it doesn’t account for? Innovations, new campaigns and competitor activity, for example.
The statistical model is a guide to the future, but you can adjust it if you know things it doesn’t to make it more accurate. (This approach comes from an area called Bayesian statistics).
For example, what about seasonality? Do your sales change depending on whether it’s summer or winter? What about peak selling times like Easter, Black Friday and Christmas? Your statistical forecast doesn’t know when those occur. But you do. You can build that into the model.
What about competitors? You probably know how often they advertise or price discount, and when they’re likely to do it next. You can adjust your e-Commerce forecast with this knowledge to make it more accurate.
When you don't have historic sales data
Clearly, the better your data, the more accurate your e-Commerce forecast. But what if you don’t have a lot of data? You’re launching an innovation, or just starting in e-Commerce.
In those cases, you have to be more creative in how you find data. And you have to make bigger assumptions. Doing these means your forecast won’t be as accurate as an established seller. But it’ll still be better than pure guesswork.
Benchmark versus other categories and channels
If you already sell through other offline channels, you can use those sales to benchmark your online potential. You could use industry standards for the ratio of online to offline sales, e.g. the latest NAB online retail sales index puts Australian e-Commerce at 14.3% of the total retail trade. (though bear in mind the e-Commerce boost from Covid-19 lockdowns. See our article on online alcohol sales, for example).
It’s unlikely you’ll go straight to 14% of your current sales in month 1. But assuming you might get there in month 14, an e-Commerce forecast of 1% of your current sales in month 1, 2% in month 2, and so on isn’t an unreasonable place to start.
(Note, with this approach you’d also want to factor in how many of these sales were incremental – new customers versus cannibalised sales – existing customers switching from offline to online).
Online customer data
If you’re adding e-Commerce to an existing website, you can use your digital data to tell you how many potential customers you can reach.
You can also boost the number of visitors with digital media and social media.
You can then make some assumptions about how many of those shoppers will buy. This is called the conversion rate in e-Commerce.
A typical conversion rate would be around 2%.

That sounds low, but many people like to browse a site before they buy (see our D2C business model article for more on this).
To grow your sales, you need to either increase visitors, or increase the conversion rate.
Reasonable assumptions in your e-Commerce forecast
No matter how much data you pull together, you still have to make assumptions about what’s going to happen. You want these to be as reasonable as you can make them.
All your assumptions should have a rationale behind them, so you can explain to others how you made that assumption.
It’s reasonable to use past history or benchmarks. Any time you can use real numbers, that also usually works well.

Assumptions will always have an element of subjectivity and opinion. But using real numbers makes the assumptions sound more objective and fact-based. People seem to have more confidence in a forecast when it’s based on real numbers.
It can help to to involve the rest of the team when you make your forecast, especially the supply and finance teams. This helps them understand the thinking, and appreciate that forecast accuracy takes time to build.
It’s important your e-Commerce forecast is believable. It will slow you down if people in your team don’t believe you’ll hit those levels of sales.
Conservative numbers are more believable
In general, it’s usually better to be conservative in your e-Commerce forecasts. Lower numbers feel more believable and are easier to manage. It’s much better to under-forecast and over-sell.
If you over-sell, you run out of stock. But you can use this to make your products seem more popular. People assume if a product’s out of stock, it’s selling well. This comes from the psychological concept of scarcity. (see our behavioural science article for more on this). It makes your products seem more desirable.
But if you under-sell, that’s tougher to manage. You end up sitting on stock, with cash tied up in the warehouse. If your products have a shelf life, you may have to write them off. People lose confidence in what you’re selling.
Over-forecasting puts pressure on the supply chain team (who need to find warehouse space to store the products) and the finance team (who worry about cash flow and returns). Given they’re already not fans of e-Commerce, best to try to avoid this.
An example from our first e-Commerce forecast
To finish, we thought we’d share some lessons from our first ever e-Commerce forecast. (see our articles on e-commerce learnings and barriers to e-Commerce for more on this).
It was a brand new store with no sales history. That meant many heated discussions and different opinions about the assumptions for the forecast.
On one side we had digital deniers who thought our forecast was too high. They didn’t believe customers would switch from other channels (mainly because they didn’t shop online themselves).

On the other side, we had the digital evangelists. Senior leaders who urged us to “aim big”, to “go for it”. They told us a modest e-Commerce forecast wouldn’t excite people in the business. We listened obviously. But for us “exciting people in the business” wasn’t the goal.
(see also our article on the three monkeys of e-Commerce. The deniers were conservative Cool Blues and the evangelists optimistic Sunshine Yellows).
In the end, we went with a middle ground e-Commerce forecast. We thought of it as a Goldilocks number. Not so high that people wouldn’t believe it. But not so low it wouldn’t interest people. We forecast monthly sales would grow to $200k/month by Month 12, with an overall year 1 forecast of around $1m.
Results from that experience
So, how accurate was our e-Commerce forecast?
The short answer is it was horribly inaccurate month by month. Much slower in the first 3 months than expected. Then a bit of an uplift in the next 3 months. And then we had unexpectedly fast growth in months 6 to 12. Overall, for the year, not that far off. But the accuracy month by month, was awful.
The slow start was down to an unexpected delay in advertising support from the brand team as we launched. Though we picked up some sales, these were modest. Actual sales were only about 10% of forecast for the first 3 months.
We used those quiet first 3 months to make improvements. We boosted the design of the store website. We bedded in the order to delivery system. And we got the customer service team used to dealing with e-Commerce enquiries.
The advertising support eventually came at the end of month 3. We also added new products to the range. By month 6, we were at 80% of our initial forecast. Awareness levels started to rise. Store visits increased rapidly.
The shape of the sales chart was similar to the shape of the product life cycle. Lower at the start. But with a definite tipping point when rapid growth took off.
Sales more or less doubled month on month for the rest of the year to take us past the original forecast. With 12 months of data, our Year 2 e-Commerce forecast was much more accurate.

Conclusion - the e-Commerce forecast
Your e-Commerce forecast sets out how much you expect to sell online over the next 12 months.
For businesses already selling online, the more you know the better your forecast.
Use statistical techniques like linear regression, correlation and cluster analysis to analyse historic sales and marketing data, and make e-Commerce forecasts more accurate.
However, you should also apply your own understanding of market dynamics (a Bayesian approach) to improve accuracy.

Businesses new to selling online need to be more creative in how and where they source data. Benchmark against industry standards and other related data (e.g. website visits). Forecast accuracy will start low but improve over time.
E-Commerce forecasts combine data gathering, analysis, and assumption making. They’re your most honest view of what’s going to happen. That view helps you organise your marketing, supply and finance plans in a much better way.
Check out our online store business model and managing an online store guides for more on this. Or, get in touch if you need help with your own e-Commerce forecast.
Photo Credits
Man looking at ceiling : Photo by Anton Danilov on Unsplash
Doorstep delivery (adapted) : Photo by MealPro on Unsplash
Dollar lights : Photo by Chronis Yan on Unsplash
Marketing Dashboard : Photo by Carlos Muza on Unsplash
Question Mark on Tree : Photo by Evan Dennis on Unsplash
Sale : Photo by Justin Lim on Unsplash
Google Analytics : Photo by Edho Pratama on Unsplash
Three people pointing at laptop : Photo by John Schnobrich on Unsplash
Thumb up (edited) : Photo by Markus Spiske on Unsplash