So many times, I see posts on this board, where a new practice was tried by a webmaster. Sometimes that might be changing Adsense ad color or size for higher yield, changing seo, changing templates, or other issues. The change is made, then the webmaster watches for the next couple days or a couple weeks, and then reports back to the forum the results. What I see are conclusions that are incorrectly derived.
In order to test anything, you need to know how to test. The first step is to understand your site stats.I have said this numerous times in the forum, but clearly I have said it without enough depth to take root in many webmasters' minds. I'm correcting that approach now.
A few days ago, I saw a post stating the webmaster had changed the Adsense blocks on non-performing pages/sites, and voila, the next day, Adsense revenue was up. He raced to the conclusion that the changes he made, had increased his sales. This may be so, but he couldn't possibly know it, based on the facts he presented.
When we look at stats, we need to look at apples to apples. This means if we look at traffic yesterday, it was the last Tuesday of June, the first official week of summer. We can not compare that traffic to the day before, a Monday. In order to compare the traffic we need to look at the last Tuesday of May. What is the difference, up or down? Then, look at the same day, in the same month, each past month, and a year ago, if you have that longevity. What changes do you see from there? If you look at the last 4-5 days, then you are comparing Tuesday to Friday, Saturday and Sunday, and that simply can not be done. You must always compare one day of the week, to the same day of the week, in different weeks.
There are many factors that affect traffic. Day of the week is the most important, day of the month is also important. Did you know for instance, that traffic in the U.S. is generally weak on the 1st, 3rd, 15th, and 30th of the month? Those are major paydays across the U.S. Most know that Thursday and Friday are paydays, but many U.S. civil servants and retired persons are paid on the days of the month stated above. When they have money, they head to the bank, and brick and mortar shopping. This accounts for a large decrease in traffic for many ebusiness sectors, on those days. Unless... there is an ice storm over much of the U.S. states with the most dense population, then traffic will increase... unless it resulted in major power outages. A heat wave will have the same effect. It drives traffic to beaches and pools and away from the web... for now. In the future, that will change dramatically. As you can see, traffic data is very complex. Before you jump to conclusions, and begin changing things on the website, you need to discouver if traffic is really down, and why it is down. Is it 'normal' or is there a 'problem'?
An example of statistic gathering gone wrong, happened this morning. A family member who documents the stats on one our sites, indicated search traffic was down yesterday, only mildly, about 5%. I replied, what did you see on the last Tuesday in May? The response: Oh... (pages flipping) traffic is actually up almost 90%. That sounded better. Our traffic is up every week, even in the low season due to my aggressive dedication to content and traffic development. The statement that traffic was down threw me, and made my ears perk up. Was this a red flag? The answer was, no, it was bad statistic gathering.
We document every element of our traffic on every site, every day, on paper. While we use an excellent site stat program, having it on paper, makes it easier to discuss anywhere, at anytime, at a quick glance. So, we have paper charts on legal pads, with columns, that compare traffic in several important areas, impressions, search origin, and other website origin. Next to the date, I have the day of the week column. Within search origin are other columns, for each major search engine. We document the percentage of traffic AND the total hits in each column. At a glance, I can see if traffic is up or down from Yahoo, and flip quckly through the pages to compare that last Tuesday in every past month. I can do it at the water cooler, on the back deck, or in the supermarket check out line.
It is important to point out that you need to look at both the percentages and actual numbers. Percentages can not be used to compare one thing to another, short term. You need to look at the actual numbers.
Traffic to one site received 2000 hits from Google which was 66% of all engine traffic, 700 hits from Yahoo, which was 23% and 300 hits from MSN, which was 11% of all engine traffic.
Using the example above, if we want to see if our traffic is up at Google, we look at the actual number of 2000 hits, and compare that number to all other last Tuesdays in other months. That will tell us if our keyword base at Google has eroded or not,without actually taking the time to run through the thousands of keywords we rank on. Looking at the percentage of 66% will only tell us how much of the share of the search market Google has at our site. As more engines are charted, you may have 10, 20, or 30 more engines in your chart, many of which, are still just Google private labeled. The percentage can change negatively for any one engine, while actual traffic increases. Don't let the percentages panic you into knee jerk reactionism. Look at the actual numbers, and compare apples to apples. If you see a problem, a real problem, then look at the actual keywords, and see where the movement was.
This is how we see what is working, not working, and where there are indictors that something is going terribly wrong, or incredibly right (as it is now), and we can nip problems in the bud, before they get out of control. Since our major sites are indexed daily by the search engines, we can quickly do damage control, when a trend is beginning. But, jumping to conclusions, by looking at the last few days, or weeks, when comparing stats, can be a huge mistake. With daily indexing, a mistake will be felt within 48 hours If you jump to conclusions, you may turn around an 'up' period and send it spiraling down. That might have happened today, had I not insisted the webmaster of my #2 traffic site (#1 revenue site) looked deeper into the picture. Imagine, thinking that the traffic was down 5% when it was actually up 90%.
Using a good site stat program can make the job of understanding site stats very quick and painless. We use RealTracker primarily and StatCounter as back up. The difference between the two is like night and day. Real tracker sorts the traffic into dozens of categories. I can sort them by day, day of the week, hour of the day, week of the month, season, month, year, origin, search, keywords, screen resolution, browser, os, location, and so much more, all with the click of the mouse. I can be looking at Yahoo traffic today, and with a click, see it on the last Tuesday of any of the last three months. But, that is where it ends. Many of the details disappear after the last 90 days, though major stats, day, week, month remain. The paper charts we create, keep those detailed stats for us, with only two minutes work a day. But, the program's ability to sort these facts for us daily, makes it easy for us to chart them. Statcounter doesn't do this, to the same degree.
As a family run business owner, I wear many hats, as most of you do. I'm doing the job that 20 employees would do, if I had them. When you let any one area of your business slide to the side, things can quickly get out of control. When more than one thing goes wrong, troublshooting can be very difficult, if not impossible. It is that combination of problems that has resulted in many businesses folding up on the web. Documentation of site stats, religiously, can make your job much easier. It can be done quickly, and accurately, and can save your business in the end.
Early this year, I was helping a friend, and neglected my site stats. Things got out of control, and I slid hard to 258,000 Alexa ranking. I had been at 101,963. After 3 months of hard work, I am now at only 152,317. There are only 2 months left to recover what I lost, before peak season. With the growth we are experiencing, damage control is indeed working. But, imagine where we would be today, had we been watching the stats more carefuly back in March. This traffic downward spiral is the result of not watching the stats, and seeing the red flags, early on. Learn from my mistake. Selling and website building is NOT the most important part of you job, traffic data is.
Too often the business owner spends all their time designing the site, maintaining the site, searching for new products, searching for new traffic, doing bookkeeping, maintaining equipment and software and other aspects of the business. They leave site stats to a remedial hit counter, or to complex raw logs. Both of those solutions are wrong if you want to succeed on the web long term. One tells you bascially nothing, the other, tells you way too much, in an unorganized manner.
While raw logs have their purpose, they can't tell you in under two minutes if the traffic from Google is up or down, compared to a month ago, two months ago, a season ago, two seasons ago, or a year ago. Since you wear so many hats, it is very important that you can do thing quickly, including gathering accurate data. They also won't red flag you that the site you were trading links with, moved you from the first page in a category, to the last, in violation of the original agreement. The difference may only be 10 hits a day, and not noticeable in the wide range of traffic you have. But, with the origin stat, you can see the urls that send you traffic from other sites, and there, you will see that difference, almost instantly.
I know it is a burden to wear another hat, and wear it more than once every few months, but this is one hat every webmaster needs to wear for 10 minutes at the beginning of each day. Document the details of your traffic, and use them as a tool to build and troubleshoot your traffic. Never, I mean NEVER, change something on your site, and then compare it to the day before, days before, or week or two before. This will result in bad conclusions,and ultimate disaster.