Archive for June, 2009

June 12th, 2009

Measuring web sales from offline advertising like TV, press and radio

If you are running press, radio, TV or DM activity to drive traffic to your web site and generate online sales, you may be wondering how to measure the relationship between offline media and online sales. This short piece will give you a simple guide to analysing whether your offline advertising is delivering web sales.

It’s worth stating at the outset that the job of relating offline media to online response is a complex area. Media channels like TV, press and radio don’t carry cookies so the tracking options available within the online sales funnel are simply not available when you start working with offline media.  Consequently,  we have to use other techniques that can give us an informed view about ROI from offline media in online environments.

The approach I am going to take you through is not 100% watertight, but it is a fraction of the cost of statistical modelling and will give you a reasonable idea of how to explore the efficiency of offline media in driving online sales.

Stage 1 - Visual observation of the data

  1. Obtain web log data running around 1 month prior and 1 month post your offline campaign activity
  2. These web logs should be daily level data showing both total unit sales and total sales value by day
  3. If possible obtain data for sales originated through both search engines and direct browser visits
  4. Enter into an excel spreadsheet
  5. Separate these two sources of data and undertake the following for each of the two sales data sets
  6. Add your daily offline media spend
  7. Chart the direct and search sales alongside the spend data
  8. Look at the data and see if there is any visual pattern in it. See of there are slight rises either during or lagged behind your offline campaign activity
  9. If there are any patterns which suggest an effect between your offline media and online sales proceed to the next stages.

Stage 2 -Sales analysis by day of week

  1. We need to ‘eliminate’ any day of week effect (for example, for many online companies Sunday and Monday are often their best sales days) so sort your data into the seven days of the week for the whole period
  2. You should now have seven mini data sets, one for each day of the week, each containing media spend (where applicable) and sales data.
  3. Calculate the average number of unit sales or sales revenue for each day of the week
  4. Now rank within each mini data set of days by daily unit sales or sales revenue
  5. Compare the sales for each day to the average of that day over the period
  6. If the advertising supported days are all above average, then your offline advertising is likely to be driving these online sales.

Stage 3  - Estimating the value of web sales driven by offline advertising

  1. For each of the days of the week, refer back to the average sales for each day
  2. Now for each of the days with web sales above the average, subtract those sales (or their sales value) from the average figure for the day of the week
  3. This is your incremental sales revenue
  4. Now compare your total incremental sales revenue to your offline advertising spend in the period
  5. You can now estimate sales and gross margin ROI
  6. For revenue ROI, compare the incremental sales value to the ad spend
  7. For gross margin ROI estimate your gross margin a percentage of sales and then compare the margin value to the advertising spend.

You will also need to factor in seasonality. Ideally, you would undertake the same exercise for the same period one year prior to the period you are analysing. If your sales in March are high, it may be the case that March is a good seasonal sales month. You need to account for this eventuality too.

If you’re an advertiser with large volumes of traffic and omnipresent advertising, then of course, things become more comlex. You will need to build a test marketing campaign structure with lots of media variation - i.e. different channels running at different times with different messages. This can then be seasonally adjusted and modelled using multiple regression to estimate the sales effect of each media channel being used.

Labels: , , , , ,
  • Facebook
  • del.icio.us
  • Digg
  • Technorati
June 9th, 2009

Online advertising works beyond the click

It’s an ongoing debate: just what influence does digital communication create beyond clicks? Well the short answer is a lot. It contributes the following:  subsequent search visits (product and brand terms),  subsequent direct site visits (over the short and long term), visits to retail premises in the case of retailers, visits to attractions in the case of leisure destinations and shifts in brand and product reputation in the case of branding messages and content.

Recent research from iProspect / Forrester (May 2009) supports this view. It reveals that of those who viewed online ads on an ad funded web site,  only 31% actually clicked, but a further 48% either searched for the product in a search engine or subsequently visited the site via a direct browser visit. A further 9% reported that they investigated further through social media or message boards.

forrester-click-behaviour-june09

Readers who run online campaigns will observe that few online campaigns generate click through rates as high as 31%, in fact, most display campaigns generate click rates of about 1% of that, i.e  0.31% or less.  If we factor down the other responses by a similar level, then we get to 0.27% performing a direct search and 0.21% visiting the advertising site directly through their browser.  Whilst these numbers may appear low, it does indicate that responses are many and varied and exceed the response counted as clicks alone.

I’d argue that when it comes to branding effects, such as awareness, attribution and considerations scores,  the numbers may be higher than the figures above suggest.  The problem is that we have not fully understood how to quantify these additional branding effects. There are products able to isolate groups people who are exposed to online communications and, via online surveys, compare their advertising and brand awareness to non-exposed groups, and these can reveal interesting short term results. See some of those here.

But often the changes in awareness and consideration build slowly over time, particularly in products which have to be advertised almost constantly in order to reach comparatively small groups of active buyers. Mobile network O2 springs to mind here.  Whilst much of its online display activity is designed to attract potential buyers to its online shop, there is no doubt that the constant presence of O2’s blue and white imagery on the UK’s top 250 or so web sites helps to maintain and reaffirm its credentials as a player with a big interest in the digital space.  Would we still see O2 that way of we had never seen its distinctive blue online display presence?

Labels: , , , , ,
  • Facebook
  • del.icio.us
  • Digg
  • Technorati
June 1st, 2009

TV media planning for site traffic generation

If you are an advertiser looking to use TV to drive traffic to your web site or increase brand term searches, you can do a lot worse than employ some well tested techniques from the world of DRTV advertising to improve your results. In this short think-piece, I am going to review some of the techniques that can be borrowed from DRTV media planning to increase site visits from your DRTV media spend.

Before we go into techniques, it’s important to recognise that measurement is key to the response planning process. Many believe TV is not accountable, but in fact, audience delivery is measured on a minute by minute basis across the day. The BARB audience measurement panel allows us measure the audience size and composition for any spot on almost any channel at any time of day. This is minute by minute data which is ideal for matching to your second source of planning data; your own web traffic logs.

Using a combination of these two data sources enables advertisers to track web traffic, leads and sales back to their point of TV origin. So for example, we may be able to conclude that sales for product X with a value exceeding £50 are most likely to be gained from a given channel at a given time of day on a given day of week (which was traded at a given cost).

It is also possible to undertake other tests by developing a text matrix and deploying it over a given time period.  For example an advertiser can run a creative effectiveness test by running creative 1 over week 1 and creative 2 over week 4 (leaving a gap of two weeks to eradicate lag from the first campaign). From a test like this it may be possible to conclude that creative treatment 1 is more effective at driving online sales than creative treatment 2.

What do the results look like?

If you imagine that weekdays are twice as cost effective as weekends, and Channel 2 is twice as effective as other channels, and that creative 1 is three times as effective as creative 2, then you are already into the type of performance multipliers that can make the difference between an average campaign and a very strong ROI performance.

Let’s now look at this in currency terms. Let’s assume that an advertiser is experiencing a TV to Web cost per sale for a financial services product of £500 across broadcast media. Our day of week selection could reduce that to £250, our channel selection could reduce it to £125 and our creative selection could reduce it to £62.50.  This means the difference between a TV generated web sale costing £500 and a TV generated web sale costing £62.50.  These are the differences between making a sale at a significant potential profit and making a sale at a loss.

Labels: , , , , ,
  • Facebook
  • del.icio.us
  • Digg
  • Technorati