(I expect this will be quite niche for some followers, but hopefully interesting or useful anyway. do feel free to RT this / any of the tips if so. Depending on number of likes, it might take a while).
It costs less to refund the difference than to ship back, handle it, & ship out a replacement at the lower price.
Place a test order once a month on a site, you can tell how many orders they had.
Do it for a year & you know how fast they grew, when their peaks & troughs were, etc.
direct site visits.
If you tell people to visit your site in TV ads, a much higher proportion do. They usually visit via:
a) google brand ads.
b) directly typing in your homepage url.
c) google organic homepage visits.
Lots of tools exist to measure visits from TV->site.
Measuring the rough # visits/revenue following TV ads lets you optimise:
- which channels you advertise on.
- which ads you use (& length of ads - I've tested 30 second vs 20 second vs 10 second before & saved a lot of money)
- when you advertise.
TVSquared, Spoteffects, Medialytics are all tools that roughly calculate the # visits / revenue to an ecommerce site from an individual tv ad (usually calculating visits above your benchmark within a 5-7 minute window, and resulting sales in the trailing week)
Increasing conversion rate is not always good. If you lose 80% of your traffic & keep 50% of your sales,conversion rate will have gone up, but your overall sales down.
To achieve growth often your conversion rate will go *down*.
If 5% of an ecommerce site's orders go wrong, that may seem ok (95% success), but if your 'best' customers buy once a month, that's 5% chance *every month*.
Some fix that by giving preferential treatment to loyal orders (+obv trying to fix % overall)
There's a phenomenon called 'Service Recovery Paradox': If a customer experiences an issue, fixing this in a way that they are happy sometimes leaves them *more* satisfied & likely to recommend you than if there had never been an issue at all.
It feels common sense that ecommerce sites should offer free next day delivery.
Sometimes sites do better if they charge for it (£3 feels a lot on a £20 order, but if product cost is £10, £3 delivery would mean they'd have to get ~30% more orders to make up).
category/subcategory/search results pages are usually known as 'PLPs' (product listing pages) or 'Grid Pages' (as they're often grids) in the ecommerce industry.
individual product pages are usually known as 'PDPs' (product detail pages) or simply 'product pages'.
Making changes to key features in 'the loop' can have a big impact on results.
In a physical shop, 100% of visitors enter through designated entry points.
On an ecommerce site, usually roughly only 1/3 of visits will enter via the homepage.
Including some of the kinds of content you'd want people to see as they first enter on pages *other* than your homepage.
a) Reasons to buy from you.
b) Info about your brand.
c) Opening hours/customer service hours.
d) Any particularly compelling offers you have on at the time.
e) Email signup prompt.
f) Indicators on your scale/trustworthiness.
'New' visitors to any ecommerce site are usually muuuuch less likely to buy than returning visitors.
Tracking 'new customer conversion rate' (% of new visits resulting in a 1st time purchase) vs 'repeat visitor conversion rate' can be v useful.
It's an absolute basic, but missed by many. If your SEO team is reporting on increases in overall organic traffic to your website, they may be claiming credit for TV/offline/word of mouth traffic too + the investment in your brand over your entire company history, etc.
The 'multi-channel funnel' reports can help to show you what's actually happening, but even just knowing this info helps.
a) Note in the screengrab above that the site had 1.5m pageviews, but it's only gathered page load speed data for 15,380 of them. You can increase that % using something called '_setSiteSpeedSampleRate()' (google it).
Subscribe to them all from one inbox & you have an automated archive.
a) one for 'new customers'.
b) one for 'new subscribers not yet customers'.
These are automated email programs, that send the recipient an email over the initial period after they become a customer/subscriber.
a) welcome message from the brand.
b) info about what you sell.
c) info about why customers love you.
d) email covering best sellers from each of your categories.
e) *possibly* some sort of offer.
a) some sort of 'welcome' offer.
b) info about a product/category you've deduced they're interested in.
c) why customers love you.
d) best sellers from other categories.
e) offer follow up.
These are useful as a way of judging email welcome programs/the general 'onboarding journey'.
if it costs you $50 to acquire a new customer. and you make an average $50 profit per order. taking a customer to their 2nd purchase is one of the single biggest things you can do to increase your profit)
In general, consistent email programs that follow a fairly set format with regular testing are best. In Europe, most email programs go 2-5x per week.
- General trend topics.
- Seasonal topics.
- Promotional topics.
- Inspiration content.
- Brand content.
- Mid/End season sales.
Some customers might therefore *not* receive promotional content, while some *not* inspiration content.
- Random selection.
- Best sellers.
- Best sellers in most relevant categories.
- Personalised on browsing behaviour.
- Personalised on buying behaviour.
- Personalised + highest profit.
Eg, simple rules:
if customer bought [brand x] printer; email best sellers from [brand x] printer ink.
if bought [price tier 2] jacket; email [price tier 2] tops.
Those are often triggered by subscribers reaching 'journey points', eg:
- not bought for 60/90/180 days.
- anniversary of first purchase.
- actual birthday.
a) in favour of the newsletter.
b) embedded in the newsletter.
- 'sorry' email - something has gone wrong, and the brand offers you x% off to apologise. (often very lucrative)
- 'survey' email- designed to gather info from subscribers, but also to reward with a % off & generate a sale.
- Subject line (& 'pre-header' (the preview text most email browsers show you))
- Call to action (the element(s) that ask you to click through to the website)
- Unsubscribe button (a negative action)
- Bounce rates ('soft'/'hard' bounces)
- Deliberability rates.
- Open rates.
- Clickthrough rates.
- Transactions/new customers/revenue/margin.
- Unsubscribe rates/absolute unsubscribe numbers.
- Active email subscriber base/growth/churn rates.
- Unique opens/visits from email. (ie, if the same subscriber opens 4 emails in a week = '4 opens' but '1 unique open')
- Spam rates. (those who click the 'spam' button)
This can reduce total unsubscription.
It's a nice thing to do when those may be sensitive topics.
(@arenaflowers do this)
a) email 20% of your list with one subject line.
b) email 20% with a different subject line.
Wait a few hours, send the remainder to the one with the highest open rate.
Demographics. (age/gender/life stage)
Customer lifecycle status.
Value. VIP/core customer/net negative high returner.
Propensity bucket. Highly likely to buy->not at all.
- if someone was 'highly likely to buy' in a given month & did not.
- if someone moved from 'Ultra VIP' down to 'VIP'.
- if your defined 'life stage' bucket moved from 'expectant parent' to 'new parent'.
- call to action on website.
- call to action in package sent to customer.
- auto opt in on purchase (with option to opt out).
- competitions on site/social.
- competitions with non-competing retailers. (some do very well from these)
oracle bronto, eloqua, marketo, oracle marketing cloud, salesforce marketing cloud, dotmailer, sendgrid, campaign monitor, cheetahmail, emarsys, mailchimp, adestra, emailoctopus.
All let you send emails, mostly focused on manual campaigns.
Sailthru, Klaviyo, Optimove, Exponea, Ometria, AgilOne, Intercom.
Most these focus more toward 'marketing automation', usually centred around a 'customer data platform' triggering emails/on-site messaging/other marketing.
- to acquire new customers.
- to block competitors from potentially taking your customers.
Each is managed quite differently & reporting on all 3 as if they're the same = means you can make the wrong decisions.
If you have low margin products, or low converting, or just low priority, having those in your shopping feed may 'cannibalise' your better products.
Whatever changes you make, try to get it so you can make historic comparisons. It's easy to use attribution model tweaks to make it *look* like you're improving when you're not.
homepage (core/brand terms)
categories ('women's dresses')
subcategories ('women's maxi dresses')
brand category pairs ('ralph lauren maxi dresses')
facets ('size 16 dresses')
a) sorted in a way that means best-sellers are at the top (ie. appeal to broad mix of custs).
b) includes a mix of products high up that would appeal to various customer types.
- highly reviewed products.
- seasonal best-sellers.
- evergreen products that convert well.
- mix of cheap/high/mid-price products.
- mix of brands.
- mix of gender/other facets if it's a really broad term on a fashion site ('t-shirts')
a) indicator of where the user has landed within the site (eg breadcrumb)
b) trust indicators (review stars/testimonial/# customers)
c) methods of filtering down further.
d) email sign up call to action for new visitors.
a) maximising the results of the channel.
b) claiming credit for PPC activity that might otherwise be invisible.
if you can, tagging emails 'PPC acquired' = you can see if they're more/less valuable long term.
The most important pages on an ecommerce site are usually (in order):
1. product detail page.
2. product listing page.
4. bag page/step.
6. thank you page. <-- this is a forgotten page, but can be used very well.
The 'thank you' page (aka receipt page) is often an afterthought for ecommerce sites. But it's here where:
a) customers are reassured they've bought the right thing.
b) have made a purchase, and are open to be provided with any/all extra information or options from you.
People often worry about homepages being 'too long', but as long as you design in an 'upside down pyramid' format that's fine. Ie: Top portion of the screen should show things that apply to *most* users...
It's fairly common for people to want to *avoid* including category links on the homepage, arguing that they're available via navigation.
a) On mobile, fewer see the whole nav.
b) Increased visibility = increased clicks.
At least show your main categories.
Elements that most frequently appear on ecommerce homepages, roughly in order of top/bottom:
1. Menu/top nav.
2. Visible search icon.
3. Cart(+maybe wishlist).
4. Offer bar.
5. USP bar.
6. Main Category entries.
7. Seasonal categories.
9. Seasonal individual products.
11. Recommended products, individual/segment based.
12. Recently viewed products.
13. Services: finance, 'prime' options, etc.
16. Ancillary links: help/faqs/links to sister sites/delivery & returns info.
17. Social links.
18. Ts & Cs
19. Payment icons.
20. Live chat, bottom right.
21. (phone info/opening hours)
More niche content that regularly appears on ecommerce homepages-
22. Shop the trends.
23. Shop by body shape/by room/etc.
24. Shop by size.
25. Rebuy this product you bought.
26. Trade in your old X.
27. Blocks for USPs, instead of one simple bar.
29. Security icons.
31. Download the app icons.
32. Store finder.
33. Product configurator.
34. Track your order.
35. Inspiration carousels.
36. B2B login.
37. Push notification opt in.
38. Blog posts/TV ad/ancillary content.
40. Social content, auto pulled in/curated.
41. Editor's picks.
42. About the owner.
43. Current switcher.
44. Language switcher.
45. Gift services (gift cards/wedding lists/gift boxes).
46. The dreaded 'cookie' block.
You can do things like set up a hierarchy:
If bought X within Y months & is VIP = show X.
Unless bought Z, in which case show Q.
'SEO copy' has vanished from lots of ecommerce homepages. Part of that is trying to avoid being seen as 'overoptimising' by Google. Where it *does* still exist, it gives you a good indicator of which terms businesses see as *the* most important.
- a common user journey for a repeat customer on that site is probably 'land on homepage -> click on 'New' which is leftmost'.
- by putting 'clearance' far left, where 'New' *usually* is, users habitually reaching for 'New' take a look at it.
You can run an ecommerce site successfully without paying attention to the numbers (I know some site owners who do a great job just focused on featuring great product, and putting lots of love into presenting well to customers). Numbers = make it easier to control.
"CPA" is a one of the most important ecommerce metrics. It stands for "Cost Per Acquisition". It's the cost of achieving an order from a new customer (another way of putting it is 'Cost per newly acquired customer').
Or by pretending *all* your marketing is spend on new customers, and dividing the whole lot by the total # new customers within the period.
(Some do not differentiate new/repeat, and simply have the same numbers for 'CPA' and 'CPO'...)
"CPO" is a similar ecommerce metric. It stands for "Cost Per Order". Ie: All marketing cost over a period / all orders.
Some guide their business on CPO...
CPM means 'cost per thousand impressions'. Eg, an ad may cost you €10 CPM, meaning 1,000 views of the ad cost you €10.
CPM is more of a niche metric - only used by people who carry out display/video advertising.
CPV/CPS is 'cost per visit' or 'cost per session'. Some use this metric as a rough guide to judge traffic across different channels, to see how they're performing over time (or forecast into the future), or to keep as a rough guide on profitability,
If your CPS has gone up over the years, and the amount of margin you make per session has gone down, your profitability is lower.
*Note*: You sometimes also see 'revenue per user'. I occasionally come across very misleading conversion rate optimisation tests, where they've declared a winner based on 'revenue per user' in one side of the test.
CTR means 'clickthrough rate'. If an ad has 1,000 impressions, and 10 people click through it, that's a 1% CTR (1% of views resulted in a clickthrough).
Often it's used to compare different ads against each other. A higher CTR ad is usually better than low CTR.
CVR means 'conversion rate' - this is the likelihood that any given visit to your website will 'convert' into an order. If you have 100 visits, and 2 of them bought, you have a 2% conversion rate.
CVR is *mainly* measured vs sessions, but some measure it vs users.
- overall CPA.
- CPA for a particular country.
- CPA for a device type.
- CPA for a particular marketing channel.
ie, they are tools for understanding the overall business, or individual parts of activity.
- you acquire 100 customers through AdWords at a CPA of $60 ($6,000 total spend)
- you conclude you could have just spent $6k with Youtube, rather than $9k total, and achieved the same 200 customers.
Next = a caveat on that)
Spending more in a marketing channel usually means the CPA goes up. Eg, if it costs you $1000 to find 100 customers through AdWords, it's likely you'll have to spend the next $1000 less 'efficiently', as you've already used the most obvious areas to spend it.
Metrics like the above are useful for measuring how your marketing activity is performing. They're also useful for running simple simulations to help you decide where to place marketing $.
Here's an example:
If your CTR is 1%, you got 10,000 clicks for that. If your CVR is 1%, that means you'd get 100 orders.
That's therefore an estimated $125 CPA (or CAC), or a $100 CPO. You can then compare that vs other marketing opportunities, and judge whether to spend the $.
"LTV" is an important ecommerce metric, not much talked about outside marketing: "Lifetime Value".
This is the amount of money the average customer spends with you over a lifetime. Some measure it as revenue, some as margin. 'Lifetime' is often a set length.
One of the reasons the LTV concept is so powerful is that - if you have the cashflow available - it allows you to beat competitors who do not have lots of cash.
Eg: Across a 'lifetime' of 3 months, an average customer generates $50 of profit.
Your competitor does not have lots of free flowing cash, so they can only afford to pay up to $10 to acquire each new customer, as they need to pay their bills with the immediate profit.
A huge amount of the business world runs on this concept - it's not nearly talked about as much as you'd expect outside of marketing/business itself.
CPA is also sometimes known as 'CAC' - 'Customer Acquisition Cost'. Venture Capitalists & investors often use this term. They look at the 'CAC:LTV' ratio. Ie: What is the amount of $ you make from a customer over a specific period vs the amount that customer cost.
You can use LTV:CAC to compare 2 businesses in a similar market. You can also use it to compare:
- effectiveness of different marketing channels.
- performance across different countries within 1 business.
- performance in one period vs a past period.
This is also sometimes known as 'Unit Economics' (if you don't know that phrase, it's worth remembering).
'Unit Economics' can be used to guide where you choose to place your time/money/effort. Eg:
A customer in Switzerland may be wildly more profitable than a customer in Spain based on LTV, *even* if it costs you 1/10th the amount to acquire customers in Spain.
Sometimes marketers use the phrase 'cohorts' when talking about LTV, acquisition, etc.
Eg: 'The June cohort isn't performing as well as the May cohort'. That would mean the customers acquired in June are showing a lower LTV than those acquired in May.
You can use cohort performance to guide marketing spend too. Eg: If cohorts acquired in June perform better over time than those in November, next year you might wayt to spend more marketing budget in June & less in November.
You can also use cohorts to measure the effectiveness of A/B testing. Eg: If you run an A/B test, ordinarily you'd judge its effectiveness on the immediate outcome (ie, 'A' converts customers better than 'B')...
Eg, if you run an A/B test on your site where 50% of users are exposed to one price model, 50% to another, you may make more immediate $ from 'A', but over time the 'B' cohort wins.
There are lots of complications & caveats to be aware of on these metrics for *ecommerce*. Eg:
- some customers are much more valuable than others - they spend more, and more frequently. When dealing with averages, that's always something to be careful around.
- Returns skew everything.
This is especially tough to deal with for people who have very long return periods. Eg, if you have a 90 day return policy, how do you know whether a customer is profitable after 2 months, or if they're going to return everything?
Finance/credit - for some companies, an 'unprofitable' customer based on what they purchase will actually be profitable based on the interest made through finance. Some ecommerce businesses are *only* profitable through finance.
Attribution of acquisition/retention spend - if you are spending marketing $ on acquiring new customers AND on retaining them, it's complicated to track how much $ you've spent on retaining particular cohorts.
Lots of elements in pricing are so specific to your company/customers, you have to test. In general it is better to make 'big' tests rather than small. (changing the price of a few items by a few % and looking for impact will only work if you have v large scale).
Pricing like this:
^ it's about the numbers *before* the dot rather than after.
If your audience is price sensitive, and knocking a .01 off the end moves the leftmost digit down by 1, they may work for you.
Some mentally round down to '60' (you hear it occasionally in user tests).
At some price points, it feels a lot less (99 = 'less than 100')
It's also why Rightmove (etc) default pricing from high to low:
People read from top to bottom. Finding an item of interest that's higher price first, among highly motivated buyers, benefits the seller.
Showing % or $ discount is quite arbitrary, but a general rule is if an item costs more than £100/$100/€100, showing the discount as a £/$/€ figure is better than showing it as a % ('Save $30' on something costing $150 sounds better than 'Save 20%').
*Sometimes* it's better to hard mark down products, rather than have a discount code. Instances are:
- if you sell a lot through retargeting/product feeds (often they only show hard markdown)
- if you want customers to be able to use other discount codes on top.
Sometimes *pricing up* products has little/no impact on sales.
I've run tests where products have been discounted 50%. Reining them back to 30% made no difference in sales, but a big difference in profit...
Sometimes *pricing up* is sensible even when it does negatively impact sales.
Eg: Product cost price = $10.
Your cost per order is $10 (shipping/handling)
Sell 100 of those at $30 = $1,000 profit.
Sell 60 at $50 = $1,800 profit.
In some contexts, *pricing up* will actually increase demand.
Champagne in general is a good example of this, and Cristal a great example within Champagne. It is popular *because* it costs a lot, not because people love it 10x as much as regular white wine.
(incidentally, along with the inventor of Monopoly, he was anti-capitalist)
Non-conspicuous situations where pricing up can increase demand are:
- Gifting. Eg, would you buy the cheapest box of chocolates for Mother's Day.
- Quality. Eg, you're buying a mobile phone charger, do you buy the $0.50 charger, or the $5 one?
Some ecommerce sites 'buy for the sale', with this in mind. Eg: Buy a $40 product that you won't sell many of during normal times at sell price $200. Put it on sale on Black Friday at 75% off, &:
a) it looks good value.
b) still make profit.
c) '75% off' headline.
Some also differ price by *marketing channel*. You find this more where ecommerce sites can be reached via various price aggregators/affiliates. It tends to be either an exclusive deal worked out with one of those, or the opposite: covering costs of their fee.
Pricing is *somewhat* arbitrary if you are not selling other people's RRP products.
Some ecom sites go into lots of detail & examine price elasticity, some broad brush, some set an initial price, then have markdown rules that kick in over the product lifecycle.
- After X weeks online, examine whether every product is hitting its target STR (sellthrough rate)
- If equal to/above STR target, keep the same price.
- If below STR target, mark down by a set %, based on how near/far it is to STR.
'Price' is the easy lever to pull to increase demand, but is quite blunt.
Differing price strategy by category is also very popular.
Eg: A laptop retailer will make very little margin selling laptops (maybe 5%), but will make huge margin on accessories (cables, bags, etc).
Another area where an ecommerce company may differ product pricing is on seasonality of products:
'NOS' (never out of stock) products may very rarely/never change price.
'Trend' products are expected to sell within a few weeks/months, or need to be 'cleared'.
People sometimes think of a higher bounce rate as always being a bad thing. All bounce rate is, is the percentage of visits to your site where only one page was viewed. Even if someone spent 15 minutes reading that page, and then left, it's still a bounce.
Bounce rate is not always bad: A page with a 100% bounce rate may still be valuable. If 1,000 viewed it, but only 1 of them landed on it, & that 1 person didn't view any other pages, it will have a 100% bounce rate, even if it was crucial to the other 999 people.
Bounce rate is not always bad 2: Some pages will naturally have high bounce rate. Eg, a 'contact us' page may have a high rate because a lot of people search google for 'your company hq address', land on your contact page, find what they need, & then leave again.
Low bounce rate is not always good: If you have a super low rate (eg down in single figures), it often means your analytics setup is broken: pages are tracking 2 views each, or maybe an interactive 'event' is counting single page views as if they are not bounces.
In general, 'bounce rate' is not a meaningful metric for most ecommerce sites at all, particularly not at the overall site level. It's very vaguely interesting, & can tell you 1 or 2 things, but there are dozens of more useful metrics. It's just a well known one.
The second factor was the homepage, which also had a high bounce rate. On investigating, that was largely from *returning* visitors (slightly unusual).
1) Put more of the articles that were published onto the homepage (not all were).
2) Add some 'logical next clicks' to the end of articles.
3) Put a small amount of the budget into more content & save the rest of the £120k.
Bounce rate *can* be vaguely useful for comparing like-for-like pages via particular channels. Eg: Super high bounce rate on a product from google shopping? Take it out the feed. High bounce rate on a particular subcategory? Check for product stock, etc.
Time on site (or 'average session duration') is a similarly not-very-useful metric. It's good to be aware that it's literally the time the last page on a site was loaded minus the time the first page was loaded. Here are some examples of how this misleads:
Some fix the 'time on site' issue by firing a 'heartbeat' event every so often. In my opinion it's best to just understand the flaws in the measure, and use it if/when it's useful despite that.
Tracking 404 pages is a useful thing to do on an ecommerce site. Even if there's no explicit tracking set up, you can often still isolate them in Google Analytics.
Here's how, using Samsung as an example-
Check the 'page title' there, which is 'SAMSUNG | Samsung UK' <-- that's the format it uses for any 404 page.
- for bringing people a bit closer to the site (put the realtime chart on the wall & people get to understand site use patterns)