How do you forecast your website’s ad revenue? 2010/02/17 Samuel-f Parent Éducation, Measurement, Ad Spend ad spend, online advertising, ad revenue, website, advertising revenues, forecast

This post is a follow up to “How do you forecast your website’s traffic?” and “Nuancing website traffic forecasting”

If you’re going to forecast your website’s growth in reach and traffic, you might as well go one step further and forecast what it’ll bring in with online advertising. FYI, I’m director of Internet at Corus Quebec, a Quebec province based radio broadcaster, and it’s my job to maintain online content, development and sales. It’s also my job to set budget including ad revenues and appropriately adjust monthly revenue forecasts.

We don’t have any sales optimisation tool being a rather small operation, yet that does not diminish the importance of that metric. And contrary to radio, we don’t have decades of history to trend on, or daily audience measurement, nor do we have monthly ad spend figures for a specific market to compare ourselves with competition and estimate where our share should be in the coming months.

Our competition online is much wider than other radio websites. It’s every other website that targets the same audiences we do, be they television, magazine, print or stand-alone websites.

So, how do you forecast ad revenue?

Here’s how I go about it:

• For starters, I’ve already got my individual website metrics (page views, reach, etc.) plotted out for as far back as possible (see my previous post)
• I plot out my past months’ (years if possible) ad revenue by source: local ad sales (direct), national ad sales (agency based), secondary and tertiary ad networks and affiliation ad sales, month by month.
• I calculate, for each month, the Revenue Per Thousand Unique Viewer for each sales source.
  • Traffic builds based on your UV base and advertisers (mostly national, but local as well) also buy based on your reach (UV) number. It is simpler and more logical for future planning purposes to bring the metric down to a Thousand Unique Viewer basis than it would be to a Thousand page views basis.
  • I know a lot of colleagues in the industry who adamantly calculate eCPMs (revenue per thousand page view) and that metric is good to have a good idea how much you’re making per thousand page – to see if you shouldn’t be increasing your rates, or selling more premium inventory than ad networks.
  • eCPMs however cannot be used to forecast the future –  certain months may generate higher page views per user than others while sales would remain stable (or increase), and your forecasts will be in error.
• I can then plot-out the average RPMuv (revenue per thousand* unique viewer) for a given month over the past few years for every sales force:  i.e. let’s take that February as an example on site X :
  • February 2009 = 125 000 unique visitors (always use the same source – doesn’t matter which);
    • $15,000 in net national sales (excluding the agency commission)  =  RPMvu national for Feb09 = $120
    • $8,000 in net local sales  =  RPMvu local for Feb09 = $64
  • February 2008 = 100 000 uv
    • $10,000 in net national sales (excluding the agency commission)  =  RPMvu national for Feb08 = $100
    • $7,000 in net local sales  =  RPMvu local for Feb08 = $70
  • February 2007 = 80 000 uv
    • $8,000 in net national sales (excluding the agency commission)  =  RPMvu national for Feb07 = $100
    • $6,000 in net local sales  =  RPMvu local for Feb07 = $75
• I automatically (in my Excel spreadsheet) calculate the average for all past months of February :
  • Local RPMvu = $70
  • National RPMvu = $107
• I then apply the RPMuv for next and all upcoming months per sales force:
  • let’s say my forecasted reach (UVs) for site February 2010 is 150,000.
  • that would mean next month’s sales should be in the range of : 150 x 70 + 150 x 107 = $26,550

This model (so far) works (for me) great when a certain number of variables remain stable :

• sales team turnover is either low or regular
• sales team training remains constant = they evolve at a regular pace
• client base remains relatively stable (i.e. not many major account moving from one sales force’s territory to another’s – or no more than usual new clients and client bankruptcies)
• economic situation remains stable

However, there are instances where you may want to adjust or nuance the numbers a bit :

• if any of the 4 points immediately above have not remained stable – be more than conservative in your adjustment.
• if your website is set to achieve dramatic growth or shrinkage in coming months due to known and agreed upon factors
  • + you have a good idea what that change may look like
  • + you have a good idea how your sales forces will react (how long it’ll take them to adjust)
  • + you have a good idea how your advertising clients will react and how quickly

There you have it. How do you forecast your online advertising revenues?

*really, I’m still stumped on the CPM (in English) meaning Cost Per Thousand. Where does the M come in? Makes sense in French where it’s “Coût par mille”. So I kept the same “M” for thousand in my RPMuv metric.

** Further, in a recent discussion with a colleague at Kijiji, I learned they also forecast with this method.