Organic Traffic Growth Hypothesis List: A Working Format

Efforts to increase organic traffic often begin with a to-do list: fix the titles, add new articles, speed the site up, collect links. The list grows, the tasks get done, and three months later nobody can say which task did what. The trouble is not too little work. It is that the reason for each task, and the way it will be tested, was never written down.

This article fills that gap with a hypothesis list. Each line holds one change, one indicator, one time window and one priority score. No traffic increase is promised; the list is a decision routine and its outcome is tested against your own data. Measurement tools are covered in other articles, so we link to them rather than repeat them. All sample figures, queries and pages are hypothetical.

Why an Organic Traffic Growth Hypothesis List Is Worth Keeping

The quiet flaw in a task list

A task list answers "what will be done" and never "what will be learned". That is why results stay vague: a title was changed and clicks rose, or seemed to, but two other changes landed the same week, so nobody knows which one counted.

A hypothesis is a claim, and it can be wrong. "If we reword the titles of this page group, click-through rate rises within this window" can be tested. "Improve the titles" cannot.

What a hypothesis teaches

A wrong hypothesis still has value because it tells you which road did not lead anywhere and spares you the same effort twice. A finished task on a to-do list closes without teaching anything. Each hypothesis line, by contrast, closes with a result (positive, negative or unclear), and at quarter end you hold a record of what was learned.

For the measurement side of organic visits, see reading organic visits in GA4. This article does not enter the tools; it sets up the routine that comes before measurement.

Moving From Traffic Thinking to Query and Page Group Thinking

Why a total does not produce action

A target such as "organic traffic should go up" does not say which page needs what. The total is the shared result of many queries, the season and competitors, and no single person can change it directly. A query and page group is small, has an owner and can be tied to one change.

So the first column of the list should read "which query and page group". A hypothetical example: "twelve queries about product setup and the three help pages that answer them."

Defining a page group

A group is made of queries that share an intent and the pages that serve them. How many pages it holds matters less than having its boundaries written down and never repeating one query across two groups; otherwise two hypotheses smear each other's results.

A hypothetical boundary: if "install" and "set up" serve the same intent they sit in one group, while "uninstall" is a separate group if it has a separate intent. Writing the intent is a sturdier boundary than writing the page name.

The method for deciding which keyword belongs to which page is in matching each keyword to one page. Writing hypotheses before that mapping is finished makes the same query show up on two lines.

A Well-Written Hypothesis: Change, Indicator and Time Window

Four parts in one sentence

A good hypothesis has four parts: change, scope, indicator and time. The template reads: "If [change] is made to [scope], then [indicator] rises within [time]." Leave any part out and the line turns into a wish. A hypothetical line: "Rewriting titles as questions for the setup page group will raise that group's click-through rate within six weeks."

Choosing the indicator

The indicator should be the one measure with a logical link to the change. For a title change it is click-through rate; for a content extension it might be query variety or average position. According to the Search Console Performance report documentation, click-through rate is the click count divided by the impression count, and average position is the average position of the topmost result from your site. Writing these definitions at the head of your list keeps everyone looking at the same thing.

Writing the time window

The window needs both a start and an end date. "We will look in a few weeks" means nobody will look. The documentation sets no rule for how long a window should be; that depends on the page's visit volume and on your patience, and the reason belongs in the record.

Hypothesis Sources: Search Console, Customer Questions and Search Results

Pulling rows from the Performance report

The first and most concrete source of hypotheses is the Search Console Performance report. Impressions show how many times your site appeared in search results; clicks show how many times a user clicked through to your site from Google Search. Query rows with many impressions and few clicks are natural candidates for a title or description hypothesis, and queries near the bottom of page one or the top of page two may be candidates for a content extension.

How to slice that report by query and by page is described in slicing the Performance report by query and page. Here we only treat how to turn it into hypotheses.

Customer questions

Support emails, sales calls and social comments carry questions that do not yet show in search data. The words a customer uses often differ from the words on the site. If one question arrives from three different customers, that alone is enough to write a hypothesis for a page that answers it.

Reading a results page

Searching a query yourself and looking at what comes up shows which formats (list, guide, definition) are prominent. This is an observation, not a measurement; log it as an "observation" and test the resulting idea as a hypothesis. A hypothetical note: "Three of the first results are step-by-step guides; ours is a definition." That is a fair start for a "convert to guide format" hypothesis, though its outcome is not settled in advance. Write the date on the note, since results can vary.

Separating Content, Technical and External Link Hypotheses

Three buckets: content, technical and external links

Splitting hypotheses into three buckets makes differences in responsibility and measurement visible. The content bucket holds titles, copy, internal links and new pages. The technical bucket holds speed, crawl access and page structure. The external link bucket holds content and links placed on other sites.

The three have different time scales: a title change may show within days, whereas external link work needs a far longer window. Judging all three in one window misleads. A hypothetical technical line: "Trimming the top spacing on the mobile help layout lets the answer appear on the first screen." Its indicator might be a behaviour you measure yourself, such as time on page, not clicks.

The special limit of external link hypotheses

External link hypotheses are not fully in your hands; the publisher decides and the result is often unclear. Marking rows in this bucket "no outcome expected, process and record only" is an honest frame. No link work gets a ranking or traffic expectation written next to it. The fields worth recording are the target page, the type of publication, the rel value and the date, with a referral count from your own analytics as the indicator.

Finding missing topics is among the most productive sources of content hypotheses, and hunting for topics your site lacks walks through the method.

Ranking by Impact, Effort and Confidence

A scoring scale

A simple one-to-five scale for impact, effort and confidence is enough. Impact is how much change you expect in the query group if the hypothesis is right, effort is how much work the change takes, and confidence is how solid the basis for the hypothesis is. If you apply a formula (impact times confidence divided by effort, say), note at the outset that it only helps order the list and does not decide anything.

A hypothetical row: impact 4, effort 2, confidence 3. Recording that the team used one scale is enough. Rows with high impact and confidence and low effort go to the top. Rows with high impact but very high effort wait until a small test, on one page and in a short window, has been judged worth running, so a large investment never starts without a modest trial behind it.

Who gives the score

One person should not score alone. If the content lead gives impact, the person who will do the work gives effort and someone close to the data gives confidence, the scores balance one another. In a list scored by one person, favourite ideas tend to come out high.

A hypothetical meeting note: impact 5 from the content lead against confidence 2 from the data owner exposes that only two queries support the idea. Hypotheses built around click-through rate, and the changes worth trying for them, are treated in lifting click-through rate on search results, and rows of that kind may be among your main items.

One Change at a Time: Keeping Hypotheses Apart

An example of tangled changes

A hypothetical case: in the same week a page group had its titles changed, its internal links increased and its copy rewritten. Four weeks later click-through rate was up. Which change did it? With three changes at once there is no answer. That is the reason for the one-change-at-a-time rule.

Writing each change as its own hypothesis row and applying them to different page groups, or at different times, keeps the results apart.

Putting them in order in practice

Hypotheses affecting the same page group go in sequence; different groups can run in parallel. Give every row a start and an end date, and do not move to the next hypothesis in the same group before the current one ends. If a change is urgent, such as correcting wrong information, treat it outside the hypothesis list as a separate "fix" record.

The Measurement Window and an Allowance for Seasonal Swings

Window and comparison period

To see a change's effect you compare the period before it with the period after it. In Search Console you can set the periods with the date range filter at the top of the report; detailed comparison is beyond this article. What matters is that the two periods are similar in length and in the spread of weekdays. For a site busier on weekends, working in whole weeks reduces misleading differences. A hypothetical choice: if the change went live on 1 March, take 1 to 28 February as the before period and 2 to 29 March as the after period, so each holds four whole weeks. A second change made mid-window spoils the comparison, so leave the page group alone for the whole window.

Allowing for the season

On many sites visits move with seasons, holidays and campaigns, and that swing has to be allowed for when reading a result. One way is to write the same period of last year as a separate note; if you do not have it, label the result "may be mixed with a seasonal effect".

For how to report results in a steady format, see building a monthly reporting template; the report described there can summarise the rows coming from the hypothesis list.

Recording Hypotheses That Return Nothing or Go the Other Way

Hypotheses that return nothing

When the window closes and the indicator has not clearly moved, the hypothesis is closed as "no result". That is not a failure but a piece of information: this change did not make a difference in this group in this window. A short note on why something else might be expected, such as a short window or low volume, steers the next attempt.

Hypotheses that go the other way

If the indicator moved against the expectation and the change can be reversed, reverse it first and record afterwards. The record holds three sentences: what was expected, what happened, possible explanations. A hypothetical record: "Expected: click-through rate rises. Observed: it fell. Possible explanations: the new title may not match the query's intent; the window overlapped a holiday." The possible explanations are not written as causes; they are raw material for the next hypothesis.

When tracking changes made to older pages, the steps in refreshing older blog posts can be used together with these hypothesis records.

Trimming the Hypothesis List at the End of the Quarter

Close, carry over or rewrite

At quarter end each row becomes one of three things: it is closed (a result came in, or none did), carried to the next quarter (the window had not finished), or rewritten (the hypothesis was framed badly and the scope or indicator needs to change). A rewritten row is a new hypothesis and starts a new window.

The quarter note

When the list closes, write a one-page note: what was learned in which groups, which hypotheses came out negative, and which three rows lead next quarter. Keep the note short and free of forecasts. A hypothetical sentence: "In the setup group the title hypothesis was positive, in the pricing group the content hypothesis was unclear, and the comparison page hypothesis moved to next quarter." Three sentences like that are often enough for a team and its managers to see the same picture, and they spare you repeating the same work.

Example Template: A Hypothesis List Table (Hypothetical)

The list below is entirely hypothetical; query groups, pages and scores are examples and are not drawn from a real site.

Build the list from the table in this order:

1. Tie each row to one page group and one indicator.
2. Give the three scores together as a team and write down who gave them.
3. Sequence rows in the same group one after another and run different groups side by side.

Fields to keep beside every row:

• Start and end dates.
• Indicator values for the earlier and later periods.
• Outcome: positive, negative or unclear.
• A one-sentence note for the next step.

Frequently Asked Questions

How many hypotheses are enough to start an organic traffic plan?

Quality matters more than count. A small list tied to a few query and page groups makes it easier to test every row properly. When many hypotheses run at once the changes tangle and results are hard to separate. Pick as many as your capacity can run together and leave the rest in a queue.

What separates a hypothesis list from a task list?

A task list names the work to be done, while a hypothesis also states what is expected and how it will be tested. It carries a scope, a change, an indicator and a time window, and it closes as positive, negative or unclear. A finished task closes without anyone asking what it achieved, which weakens learning.

When should I judge the result of a hypothesis?

Fix the window when the hypothesis is written and record the end date. Search Console documentation sets no rule for how long it should run, so visit volume and seasonal swings are weighed instead. Declaring a result before the window ends invites misleading early readings. If you extend a window, record the reason and new date.

Should I delete a hypothesis that returned nothing?

Do not delete it; close it and keep the record. A hypothesis that returned nothing tells you the change made no difference in that group in that window. Add a short note on why that might be, and it becomes the basis for deciding whether to retry the idea next quarter or drop it.

Do external link activities belong in the hypothesis list?

Yes, but in a separate bucket and with a clear note. External link work depends on the publisher's decision and its outcome is unclear, so it enters the list as a process and record row, with no ranking or traffic expectation written next to it. Keep its time frame apart from the other buckets as well.

Who should assign the impact and effort scores?

Scores should not come from one person. If the content lead gives impact, the person doing the work gives effort and someone close to the data gives confidence, the numbers balance each other. Use the same scale and note who gave each score. The scores claim no accuracy; they are a rough aid whose main benefit is starting a discussion.

Conclusion and Next Step

In efforts to increase organic traffic, a sound list does not multiply the work; it makes the work testable. With a bounded query and page group, a single change, a single indicator and a written window, you learn something whether or not a result appears. No row promises a traffic gain, and the list only supplies the order in which you decide.

As a next step, write the first three rows, score them with your team and put the first window in the calendar. If one of the content-bucket hypotheses needs a new article, you can look at the content writing product on Akua Adforce: you open the listing, check out through the basket and payment, and wait for the seller to take the job on; when the text is handed over, compare it with the scope in your hypothesis and raise an objection if the two differ. Akua does not offer measurement or analysis, so you read the hypothesis result from your own data. Details are in the site's Delivery Policy and Objection and Refund Policy pages.