Forecast

Every item in the action plan carries a one-line forecast: what the same kind of work did to visibility before, 30 days after it went live. It helps you choose between two items with similar priority, and it sets an honest expectation before you start. The forecast is built only from changes Rixe has already measured — when there are too few, it says so instead of showing a number.

#Where to find it

Open Actions › Action plan. Under each item, below the line that says what Rixe delivers, you will see one of two lines:

• Expected in 30 days: +2 to +6 points (based on 9 past changes) — a range built from your own workspace's history.
• Expected in 30 days: +1 to +4 points (based on 31 past changes at 12 companies, anonymised) — a range built from the anonymous pool of other customers, used when your own history is too thin.
• Not enough evidence yet — there are too few measured changes of this kind to show a range.

The ⓘ next to the line explains the method in plain words, and on Not enough evidence yet it also tells you how many measured changes of this kind your workspace has and how many are needed.

#What a "point" means

Points are percentage points of mention rate — the share of answers that name your brand, the same measure as the visibility score. "+3 points" means that, on average, a change of this kind was followed by your brand appearing in 3 more answers out of every 100, after removing what moved everyone (see below).

#Five kinds of work

The forecast is not per item; it is per kind of work, because that is what there is evidence for. Each action plan item maps to one of five kinds:

So every content-gap item in your plan shows the same forecast: the measured effect of publishing new content.

#How the range is worked out

1. Collect the evidence. Rixe takes every past change of the same kind whose effect was measured: the four-week results in Impact and the finished A/B tests. Only the control-adjusted effect counts — the change in your visibility minus what moved for everyone in the same weeks — so a good month across the market is not credited to your work. Windows with too little data to measure are left out.
2. Prefer your own history. If your workspace has at least 5 measured changes of this kind, only those are used.
3. Otherwise use the anonymous pool. If you have fewer than 5, Rixe looks at the anonymous pool: changes of the same kind made by other customers, kept only where at least 10 different companies contributed and there are at least 5 changes. The pool first looks in your own market, then across all markets. The pool stores counts and averages only — never which company made a change.
4. Average, then pull towards zero. The average effect is pulled towards zero as if 5 extra changes that did nothing had been added. With exactly 5 real changes the average is halved; with 45 the pull is only 10%. This stops a handful of lucky results from promising a big lift.
5. Draw the range. The range is a 95% interval around that pulled-in average. It widens when past results disagree with each other or when there are few of them.

If none of this produces a range, the line reads Not enough evidence yet.

#How to read it

• Compare ranges, not single numbers. A range of "−1 to +7" says the kind of work has helped on average but has also done nothing at times.
• A range crossing zero is not a promise. It means the evidence so far does not rule out "no effect".
• The pooled line is a market average. Your site may do better or worse; once you have five measured changes of your own, the forecast switches to your own history.
• Priority and forecast answer different questions. Priority is about this particular finding (its impact, confidence and effort). The forecast is about what this kind of work has done before.

#How to make your forecasts sharper

Forecasts improve as you finish work and let it be measured:

1. Mark items done on the Opportunities board on the day the change goes live — that date is what the impact measurement starts from.
2. Publish drafts through Content; published content is recorded automatically.
3. Leave changes alone long enough for the four-week effect to be measured in Impact.
4. For changes on a single page, run an A/B test; finished tests count as evidence too.

#Availability

The forecast is shown on every plan, for every workspace. The anonymous pool needs no setting: every customer's measured effects contribute anonymously, and only buckets with at least 10 companies are ever shown. See Data and security.

#Common questions

Why does every item of the same kind show the same forecast? Because the evidence is grouped by kind of work, not by individual finding. Use the priority to rank items within a kind.

My forecast disappeared. A new measurement can change which slice has enough evidence. If none does, the line goes back to Not enough evidence yet.

Can the forecast be negative? Yes. If past changes of this kind were followed by lower visibility on average, the range will say so.

#Related pages

• Action plan
• Impact — the measured effect of each change.
• A/B tests
• Visibility score