Without a good traffic source, your ecommerce site will never reach all the customers who wish to buy your merchandise.
The art of choosing a profitable traffic source, and the scientific methods used to refine it, are outlined in this article in the ecommerce series.
Establish Your House List (And These Are Not Just Emails)
You have investigated performance marketing, have seen that it is growing more rapidly than other sales methods and have decided that you want some of the action. You have chosen the product or service you wish to offer and now the problem is… to whom do you present your offer? How do you acquire prospects that are likely to become your customers?
There is no one correct way to develop your target audience, of course. Many methods and types of email lists, Facebook audience development techniques, web push targeting, direct mail lists or ad networks are available. But one source, however, must be considered above anything else: your own “house” list.
Your present customers qualify for a house list. If you ever have offered the same, or similar, items or services through retail approaches, then you have a nucleus of customers who have proven a need or desire for your offer; could use additional or supplemental goods or services; have an appreciation of your reliability and the quality of your offerings.
These customers become your prospective house list which you can contact by every possible method, including Facebook custom audience ads, web push notifications, emails, direct mail shots, or via retargeting pixels.
How do you get their names and addresses?
- At online checkout or a retail establishment when payment is made (by credit card or cash), ask for an email and postal address. Honestly explain that you would like to add his or her name to your mailing list so you can keep customers abreast of your latest offers. Additionally, concerning your online presence, ask separately customers and non-paying prospects for web push notification permissions and have retargeting pixels/cookies placed.
- If you advertise in newspapers or magazines offering a discount coupon requiring redemption at the store, you can design the coupon so that customers must complete it with their names and addresses.
- If you advertise online through display or search ads (e.g. by Google Ads) have campaigns directing the traffic to a landing page offering a “free gift” (either a physical or virtual gift, depending on your budget and marketing objectives) in exchange for user’s contact information and web push permissions as well as retargeting pixels.
But beware: lists of retail store buyers may not be the strongest prospects for online sales and customers generated through SEO means may not be proven respondents to email or webpush promotions. There is no best way to develop reliable lists – you need to test and then test again.
The objective is to build your house list or customer database. When you advertise you ecommerce items in outside sources (i.e. an email drop to a list you rent, a Facebook ads campaign), your short term objective is to find sources that will increase your sales at the time of the campaign.
Anyone who purchases something from you through such a source becomes your name, so the long term objective is to increase and improve your house list.
How Much Advertising Budget Do You Need?
This question is of vital and constant interest to ecommerce marketers. Obviously the answers are extremely varied, depending on many factors, some of which will be unique to your specific business.
A few of the qualifications that will affect your decisions are:
- the size of your house customer list,
- the profitability of promotions to the house list,
- the percentage of attrition of the house list each year (customers who stop responding), and
- your specific interest in wanting to achieve a house list of a particular size.
Let’s assume a house active file of 100,000 customers that includes customers who have ordered from you at least once in the last thirty months.
For the simplicity’s sake, let’s assume you will be doing just email promotions (this would include both strictly promotional as well as upsell, winback and abandoned cart campaigns with multiple follow-ups).
Then assume 60 campaigns per year, with a net-net profit of $0.1 per each email sent. This provides a profit of 100,000 customers x 60 campagins x $0.1 = $600,000 and is based on a conservative overall response rate. Many ecommerce sites have response rates far in excess of this.
Now let’s assume a house active list yearly attrition rate of 18 percent. 18,000 names will be lost (these are both opt-outs as well as customers that are inactive) so just to stay “even,” 18,000 new names must be developed.
Assuming all new customers are achieved only through email campaigns to outside lists, and these lists will have a response rate of 0.3 percent, then 6,000,000 emails would have to be used.
If they came at a $20 CPM rate this would mean the total ad cost of $120,000 just to keep the customer database from shrinking.
The true measurement of potential and actual response rate income varies greatly, with a large portion depending on the average order size. Some companies operate profitably with a $15 to $20 average order. Many in the giftware field are in the $40 to $60 average order size or more.
Much effort is being put forth by ecommerce marketers to improve response rates from different traffic sources, generally by a much more intricate selection and targeting process.
This includes more detailed segmentation or targeting programs to identify those parts of audience lists and traffic sources available that will provide the desired response rates in both percentage of response and average order size.
Because relying on one customer source can be both risky and very expensive, most ecommerce marketers rely heavily upon testing different online and offline media, as well as different tactics including classic “tell a friend” approaches and inserts in packages of noncompeting ecoms.
Detailed “cost-per-customer” records must be kept for every approach used. Most important is to keep detailed data on the following:
- What is a new customer really worth to you? What is the average income for one year, two years, three years, etc., of every new customer added to your database?
- What are the differences in cost and average income of customers recruited via each new customer recruitment program?
Do not be guided by what someone says is the ideal ad budget or ideal proportion of ad campaigns directed to outside traffic sources versus those touching your house list customers. Instead, consider all customer recruitment approaches used and then decide the number of new customers needed to replace those lost through attrition, plus the number desired to increase house list size, all at costs that provide a satisfactory bottom line profit.
Testing Traffic & Customer Sources for Ecommerce Sites
Testing is the only commonsense way to locate the best prospects among a plethora of traffic sources. Because nothing stands still, there is no end to testing. People change, tastes and lifestyles change, competitors change – and offers, campaigns and marketing concepts must meet these challenges.
The overall rule is to test, retest and then retest your retests. Never stop testing, because every mailing, every online ad, every phone campaign can and should be part of a learning curve.
Testing must be a continuous process, not something added to the mix now and again. If it is not a regular part of your program, you will lose more than its cost through misdirected promotions. Testing should he designed with two goals in mind:
- Determine which activities will produce more profit – and do more.
- Determine which activities will produce less profit – and do fewer (or perhaps none).
There are four ways to improve the bottom line in your ecommerce:
- Increase the response rate of your campaigns.
- Increase the average size of the order.
- Increase the order margin (gross profit) per order by increasing price and decreasing cost of goods sold.
- Decrease the costs by decreasing the cost of promotion and by decreasing costs of handling order fulfillment.
A careful examination of these methods leads to two simple rules:
- Never stop trying to increase the total dollars from a given effort through obtaining greater response, higher average order size, or increased order margin.
- Never stop trying to reduce the costs of doing business, as long as the customer is served well and honestly.
At some point, once costs have been shaved and the order margin fairly well set, the leverage in ecommerce and performance marketing essentially comes down to ways and means of increasing the response rate.
A small increase in response can translate into a remarkably high increase in profit and return on your investment. The difference between success and failure can be as fine as one- or two-tenths of a percentage point. Let’s look at a fairly typical example:
|Total costs of item sold, excluding promotional costs||$8 / unit||20% response increase (from 0.5% to 0.6%)||adds $0.029|
|Selling price||$30 / unit||Order size increase to 1.4 units per order||adds $0.022|
|Order margin||$22 / unit||Decrease in unit cost from $8 to $7||adds $0.013|
|Email campaign cost||$20 CPM||10% email campaign cost decrease||adds $0.002|
|Average order size||1.3 units per order|
Here, an increase of response rate of 20% (from 0.5% to 0.6%) is greater in terms of profit added to each and every email sent than any other indicated changes. And it is usually easier to find ways to increase the number of orders, at any point, than to get comparable bottom-line improvement through factors affecting gross profit on a per-order basis.
Remember, I have examined here just the raw cost of adding a customer to your database, not the variable worth of different types of new customers.
Proper Testing for Ecoms
You cannot be involved with ecommerce or performance marketing promotions for very long with very much success if you have not discovered the art of testing. Every ecommerce marketer worthy of the name is involved in a continuous process of improving the average response rate with the least expenditure and the least risk.
This is not to say that brilliant strokes are not conceived and accomplished from time to time, but the sane marketer builds into his programs a percentage of allowance for these “risky” moves, and that percentage is small.
There are a number of rules for proper testing:
- Do not be guilty of the “continuous series of one experiment” syndrome. This is the old story of “do you have ten years of experience—or one year of experience ten times?” In a “continuous series of one experiment,” one offer in one promotional campaign is sent to one traffic source – and the “results” are recorded. At the very least, four or five separate traffic/list/audience targeting configuration or segments should be tested.
- Once established, a “control” promo (your standard offer) should never be retired – until you have created a new “control” that outpulls the old.
- Have established, strict rules to testing new traffic sources. For example, make it your rule to always have a set of four of your best items with two proven creatives each and a budget of $1,000 for initial testing.
- Confirm any successful traffic source test with a “continuation” consisting of some modest multiple of the initial quantity of traffic. What you are looking for here is confirmation of results.
Regression Analysis Techniques: A Way to Make Phone, Direct Mail and SMS Campaigns Work for Your Ecommerce
The technique of regression analysis is not new: it is a standard statistical tool for comparing the relative behavior of two or more variables. But this technique can be applied to the behavior of database segments in a way that promises to make costly media like direct mail, phone or SMS texts work very well and at a much lower cost.
Additionally, regression analysis enables you to successfully use some advanced performance marketing tactics like bill me later schemes, cash on delivery, high value free gifts and all techniques which would be too costly to use on a broad customer database (e.g. a refused cash on delivery order or failure to receive payment for a product delivered on a bill me later scheme may cost you a lot).
In short, this predictive modeling technique ranks customers or non-customers as to their likelihood of responding to a specific promotion.
To benefit from regression analysis, for example for your SMS text campaign needs, you must first text your promotions to a defined segment of the universe of customers available. Using your past experience in marketing your product, you select broad groupings of possible prospects, perhaps 500,000 to 1,000,000 names.
You then text your SMS with an offer to a portion of your defined universe which is large enough to generate a response of at least 300 orders. If you expect a response of 2 percent, you must send your message at least 15,000 names.
Most SMS marketers receive a much lower response rate to a prospect offer. Because the test will be invalid if 300 responses are not received, this is no time to be stingy.
In fact, you might send a quantity of about 100,000 names, so a response rate as low as 0.5 percent will give you 500 responses.
Identifying those responses is the key to analysis. Each response is linked back to the database from which your test names came.
After tabulating the data, clusters of response are identified. Through the regression analysis formula, different list characteristics are given scores, indicating the propensity of other list segments with similar characteristics to respond to a similar offer.
Finally, it is possible to break down the Iist universe into segments of varying size, with the expected pull for each segment.
You might decide to promote to the top two segments of the list universe, which might be 25 percent of the total universe. But by selecting only those segments likely to buy from you, your overall response rate could be double the rate you would achieve by texting to the entire universe.
Hire a regression analysis expert and have it explained in detail relative to your specific performance marketing and ecommerce program.
And Then Test Some More…
When preparing to construct a modern building, you would be foolish to begin without analyzing test core samples of the underlying soil. And no marketer should “build” any new campaign (or major product launch) before sampling reception in a few “test” markets. In ecommerce performance marketing, there are two primary reasons for adequate test procedures:
- To save you from disaster by assessing viability for a given offer at a minimum expenditure.
- To improve your average response rate and thus maximize your net dollar returns.
The factors that affect response are just five:
- Copy – the words you use to create your appeal.
- Design – the “attire” your appeal wears.
- Offer – your appeal.
- Timing – when your appeal arrives.
- Audience – to whom your appeal is directed.
After you have a reasonably adequate promotion, these factors can be improved, through careful testing, as follows:
- Copy – from 20 to 100 percent
- Design – from 15 to 30 percent
- Offer – from 50 to 200 percent
- Timing – about 20 percent, except at Christmas
- Audience – from 300 to 1000 percent
This indicates that testing is most productive when offers and audience groups are varied. If all other factors must be held for budgetary reasons, then you should spend your money testing various audience segments.
How To Select the Numbers of Clicks, Emails, SMS or Direct Mail Pieces for Testing
Assume you have made 100 email or SMS or direct mail tests, of 1,000 pieces each, in a database of 100,000 names. Or assume you have made 100 Taboola, Google Ads or Outbrain tests of 1,000 clicks, for a total of 100,000 clicks.
Further assume that average response to your particular offer is 2 percent. This means that on the complete 100,000 names or clicks, you have a conversion rate of a total of 2,000 responses.
Did each of the 100 tests come in at 2 percent? No. It is logical to assume that some came in under 2 percent, some came in over, and some right at 2 percent.
The distribution, irrespective of the audience or medium that is used, the offer that is made, or the design created will show that some come in below-average and some above. This is called a Standard Distribution Curve.
It is likely that one of those 1,000-name or click test cells came in as low as 0.5 to 0.3 percent, and that one, or maybe a couple, came in as high as 3.0 to 4 0 percent. But most probably came in between 1.5 and 2.5 percent.
Under these circumstances, a test resulting in a response rate of 1.5, 1.7, 2.0, 2.1, 2.3 or 2.5 percent is valid. This explains why a test yielding a 1.8 percent response rate can result in a continuation test coming in at 1.6, or maybe 2.0.
All of the above spread is within the realm of statistical probability. In other words, your first test of 1.8 percent only said that you are likely to come between 1.4 and 2.2 percent in any continuation. It is important to note that you can get an erratic answer from a small sample of the whole.
The rule of thumb is the minimum number of responses to evaluate a traffic source is between 30 to 40. If you have an offer which can be expected to produce 1 percent response, or ten orders per 1,000, you need to test 3,000 or 4,000 names or clicks for an adequate test of that specific offer on that particular traffic source.
How do you know what response will be generated? Really, you do not. You have experience, a feel for the offer, some idea of how given traffic sources have worked in the past. You also have one landmark: the responses you need to break even on your particular campaign.
If you are selling a designer desk lamp at $99.95, where the break-even point is at or near 0.4 percent, obviously the number of responses required, thirty to forty, can only result from a traffic source test of 10,000 to 12,000 clicks or names.
At this point, we can answer the question: How does one select how many clicks or names to test from a given source? The answer is enough names to break even, and enough names to give a sufficient response to provide thirty or forty sales.
Customer Database Segmentation Can (and Will) Improve Response of Your Ecommerce SMS, Phone or Email Campaigns
Here is an example of appropriate segmentation for an ecommerce marketer with a large house list:
|Current 6 months||Multi-buyers|
|Prior 6-12 months||Multi-buyers|
|Current 6 months||Buyers over $ Average|
|Current 6 months||Buyers under $ Average|
|Prior 6-12 months||Buyers over $ Average|
|Prior 6-12 months||Buyers under $ Average|
|Current 6 months||Newsletter subscribers, non-buyers|
|Prior 6-12 months||Newsletter subscribers, non-buyers|
|Prior 1-2 years||Multi-buyers|
|Prior 12-18 months||Buyers over $ Average|
|Prior 12-18 months||Buyers under $ Average|
|Prior 18-24 months||Buyers over $ Average|
|Prior 18-24 months||Buyers under $ Average|
|Prior 2-3 years||Multi-buyers|
|Prior 24-30 months||Buyers over $ Average|
|Prior 24-30 months||Buyers under $ Average|
Each customer database segment must be assigned its own pixel code as part of the response vehicle for tracking purposes. Increased segmentation also should be used if you feel that response will be affected by other variables like sex, product purchased, customer source, or business versus consumer.
Continue in this fashion until no longer profitable. Incidentally, segments generally should be larger than 1,000, with 10,000 or so allowing you to feel more comfortable with the accuracy of the results.
Statistically, the lower the quantities in your segments, the less confident you can be in interpreting results.
And remember, adequate list maintenance reports are not a luxury, but an absolute necessity. List reports should be available by key variables, such as sex, recency of last purchase as opposed to cumulative purchase, multi-buyers, etc. When studied as a group, you can observe trends and really have a handle on customer profile.
Extra Income From Letting Others Advertise to Your Customer Database
The customer details you carefully assemble and refine allow you to play your house list in intricate ways to increase profits with every campaign in every medium. This information also makes your customers info more valuable to others, and should be monitored continually to keep your database vital and effective.
When adding customers to your house file, make sure you capture all possible information, and propose these selections, for additional compensation, to your business clients willing to advertise to your customer house file. For internal reasons, you may not wish to, or may be prohibited from, offering all of your information. But each record should have most of the following:
- Recency of purchase
- Frequency of purchases
- Date of first purchase
- Date of last purchase
- Dollar amount of last purchase
- Cumulative dollar amount of all purchases
- Items (or at least item categories) purchased
- Method of payment (installment, check, COD, a credit card)
- Source of name (Facebook ad, direct mail, package insert, TV)
- Personal information, when available (sex, age, income, title (Mr./Mrs./Ms.)
Your email list is only one source of income. Others include:
- SMS campaigns,
- Direct mail campaigns,
- Facebook custom audiences based on email data,
- Online ads based on your retargeting pixels,
- Web push notification campaigns,
- Package inserts,
- Thank you page promotional messaging,
- Phone campaigns.
… But Avoid Unwise External Campaigns to Your Customer Database
External campaigns mentioned above may be highly profitable “extra” income producers, but maximizing income does not mean accepting every order for a campaign directed at your house list.
Keep these three things in mind before proceeding with such campaigns:
- Know your numbers – or simply have your alternative profit calculated. Let’s assume someone offers you a $50 CPM for an email shot to your 1,000,000 customer database, or $50,000 which, on the surface, sounds great. Assume further that you have a strict limit of 50 email campaigns a year and dropping an email campaign for your external client would require you to cancel one of your internal campaigns, which on average should bring you a net profit of $0.07 from every email you send. Or $70,000 in total when being sent to your 1,000,000-names database. This would obviously mean you generate a loss of $20,000 instead of a profit of $50,000.
- Campaigns for your external clients may bring you higher than usual unsubscribes, opt-outs or complaints. Make sure you include that in your lifetime value and attrition calculations. Remember, sending emails is never free.
- Some campaigns simply are unfit for your clients. Customers who receive a mailing that offends them, is in bad taste, or a rip-off may complain. Most often they will not know or care that the offer is from a third party. In fact, what may happen is a lot worse: not only may they opt-out but report spam and have your IP and domain name reputation decreased and consequently reduce your email deliverability in future, internal campaigns. Responsible marketers read the sample promotional piece that comes with campaign order. Does the offer sound legitimate? If there is any doubt, question the marketer. If necessary, ask for a sample of the product being offered.