Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Finance › Troubleshooting

Marketing Forecast Missed

Why marketing forecasts miss: average returns projected onto new spend, seasonality ignored, conversion lag, a changed mix and assumed funnel rates.
Troubleshootingintermediate8 min readUpdated

First, split the miss into volume, rate and timing

Put the forecast and the actual side by side, input by input, not only at the bottom line. A marketing forecast usually multiplies spend by an expected response (visits or leads per pound), then by conversion rates down the funnel, then by an expected value per sale, and assigns the result to months. For each of those inputs, record what the forecast assumed and what actually happened. The gap then falls into three buckets: volume, where spend or traffic differed from plan; rate, where the response or a conversion rate differed; and timing, where sales arrived later or earlier than the months they were assigned to.

That split usually narrows the search to one or two causes. A volume gap with rates on plan points to the mix or to spend itself. A rate gap in the response to spend points to marginal returns or the market. A rate gap further down the funnel points to assumed conversion rates. A timing gap points to lag or seasonality. Kaushik's measurement model asks for a target against every key measure; a forecast is that set of targets, and the review should test each one rather than argue about the total. A forecast built this way is a chain of expected values, so each miss traces back to a probability or a value that proved wrong. Where the miss came from an outside change that no probability could have priced, the comparison of scenario planning and expected value explains when to prepare scenarios instead of averaging them into one forecast.

Diagnosis table

Diagnosis

SymptomCheckLikely causeFix
Results fell short mostly in channels whose budgets were raisedResponse per pound assumed for new spend against the channel's prior averageAverage returns projected onto extra spendForecast increments at marginal, not average, returns
Miss concentrated in particular months; the full year closer to planMonthly phasing against at least two years of own seasonalitySeasonality and calendar effects left outPhase the forecast from the business's own seasonal pattern
Sales arrived months after the spend, or are still arrivingTime from first contact to sale by cohortLag between spend and sale ignoredBuild the forecast on cohorts with an observed lag curve
Spend total on plan but split differently across channelsPlanned against actual mix by monthChannel mix changed after the forecast was setRe-forecast on the new mix at each reallocation
Leads on plan, customers well belowWhich funnel rates were measured and which were assumedFunnel conversion rates assumed rather than measuredReplace assumed rates with measured ones by source
Response fell across channels at once with no internal changeCompetitor launches, price moves and promotions in the periodA competitor or price change moving the marketTrack competitors and prices as forecast inputs

Fix, verify and prevent each cause

New money assumed to perform like the old

The most common forecasting error is to multiply a proposed budget by the current average return. The next tranche of spend reaches people less likely to respond than the ones already reached, so it returns less than the average, and the forecast overstates what more budget will buy.

Fix: forecast increases at an estimated marginal return, using the scenario approach and budget steps in the marginal returns and forecasting guide, and present larger increases as a range. If the increase has already happened and sales did not follow, the diagnosis of ad spend up with sales flat covers the causes in detail.

Verify: the next forecast's error on channels with changed budgets is smaller than on this one.

Prevent: require every forecast to state the response assumed for new spend separately from current spend.

A season treated as an average period

Forecasts phased evenly across months, or built from last year's total divided by twelve, miss whenever demand is seasonal, which it is in most categories. The full year may come close while individual months miss badly, and finance sees the monthly misses.

Fix: phase the forecast from at least two years of the business's own monthly history and the category's search demand, as the annual planning guide describes for reading seasonality, including holidays, school terms and the dates that move the category.

Verify: monthly errors shrink while the annual total is unchanged.

Prevent: keep the seasonal profile as a maintained input, updated each year, rather than rebuilding it for every forecast.

Sales expected in the month the money was spent

Many sales follow the first ad contact by weeks or months. Meta's help centre describes attribution settings as a window after a click or view during which conversions are credited, a reminder that response keeps arriving after the spend; considered purchases can take far longer than any platform window. A forecast that books revenue in the month of spend will miss early and overshoot later.

Fix: group customers or leads by the month of first contact and follow each Cohort forward, as the cohort analysis tutorial shows, and use the observed curve to spread expected sales across the following months.

Verify: the timing bucket of the next forecast miss is smaller, and early months stop looking like failures.

Prevent: keep the lag curve as a forecast input and refresh it as new cohorts mature.

A channel mix that changed after the forecast

A forecast built on one split of budget becomes wrong as soon as money moves, because channels differ in cost, response and timing. Mid-period reallocations are often sensible, but if the forecast is not updated with them, the miss is blamed on performance.

Fix: re-run the forecast at each reallocation using the channel-level assumptions, following the reallocation rules in the budget allocation guide, and keep the original and revised versions.

Verify: the volume bucket of the miss is fully explained by recorded reallocations.

Prevent: make "update the forecast" a step in every reallocation decision, with the reason noted.

Funnel rates nobody measured

Forecasts often carry conversion rates that were estimated once, borrowed from an article or set to make the target work: visit to lead, lead to qualified, qualified to customer. If any of them is wrong, every figure below it is wrong by the same proportion.

Fix: build the forecast as a tree of measured rates, by source, using the method in the KPI trees guide, and replace each assumed rate with the business's own figure from a recent cohort.

Verify: the rate bucket of the next miss shrinks, and the largest remaining error can be traced to a specific branch.

Prevent: mark every rate in the forecast as measured or assumed, with its source and date, and review assumed ones first.

A market that moved

A competitor launching, cutting prices or outspending in the auctions, or a change in the business's own prices, can shift response across every channel at once. No internal metric explains the miss because the cause was outside.

Fix: check competitor activity and price changes for the period, using the monitoring routine in the competitor analysis guide, and add the main competitors' prices and visible activity to the forecast's assumptions.

Verify: the timing of the miss lines up with the market change, and the next forecast states how sensitive it is to that competitor.

Prevent: keep a short competitor log reviewed monthly, and treat a significant market change as a trigger for a reforecast. Scenario planning goes a step further, preparing the response to a likely competitor or price move before it happens, with the signal that triggers it agreed in advance.

A prevention routine for forecasting

Good forecasting is a habit of writing assumptions down and checking them, period after period. When a forecast is built, list each input with its value, source and whether it was measured or assumed: response for current and for new spend by channel, seasonal phasing, lag curve, funnel rates and the market assumptions. Monthly, compare actuals with the forecast input by input and note which bucket, volume, rate or timing, any gap falls into. At each reforecast, update the inputs that evidence has overturned and keep the earlier versions; the tutorial on reforecasting a marketing budget mid-year rebuilds the spend and outcome forecast together, line by line. After the period closes, run a short review that names the assumption that failed and changes the method for the next forecast. Keep the forecast's headline measures among the few KPI figures both marketing and finance watch, so the review is a shared exercise rather than a defence. For organic search, where seasonality, click-through and ramp-up all come from the site's own Search Console history, the SEO forecasting guide applies the same routine line by line, separating the baseline from each new cluster.

Illustrative example — not a real result

A training provider explains a missed quarter

A professional training provider forecast 1,200 course bookings for a quarter on a budget raised from £80,000 to £120,000; it achieved 870. Splitting the 330 shortfall showed three parts. The forecast assumed the extra £40,000 would book at the existing average of about £67 per booking, or roughly 600 bookings, but the increase in search and social bought bookings at about £100 each, roughly 400, which accounts for about 200 of the gap. The timing check found 60 bookings from the quarter's enquiries arrived in the following month, as in previous years, which the monthly phasing had ignored. The remaining 70 or so came from the lead-to-booking rate, assumed at 30% and measured at 27%. The next forecast used marginal estimates, a lag curve and the measured rate.

Common mistakes

Forecasting a bigger budget at the average return. Phasing the year evenly. Booking sales in the month of spend. Moving budget without updating the forecast. Carrying rates nobody has checked. Presenting one figure where a range was needed. Revising the forecast quietly so the result always looks on target.

Frequently asked questions

How accurate should a marketing forecast be?

Accurate enough for the decisions that depend on it, and honest about its range. A forecast for a stable channel at current spend can be fairly tight; one for a large increase, a new channel or a new market should be presented as a range with its assumptions written beside it. Finance tends to distrust single figures that later miss more than ranges that turn out to be right, so state the range and say which assumption moves it most.

Should we change the forecast during the period when it is clearly wrong?

Yes, at the reforecast points agreed with finance, and with the reason recorded. Updating the forecast as evidence arrives is good practice; quietly moving it so that the result always looks on target is not. Keep the original forecast, each revision and the assumption that changed, so the post-period review can see which assumption failed and when it became visible.

Is a forecast miss the same as marketing underperforming?

Not necessarily. A forecast can miss because the assumptions were wrong even when the marketing did what was reasonable, for example when a competitor cut prices or the season arrived late. It can also hit by luck while the marketing underperformed. Separate the two by checking each assumption against what happened, and judge performance on the measures that marketing controlled, such as cost per result and conversion at each step.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Digital Marketing and Measurement Model — Avinash Kaushik
  2. About attribution settings — Meta Business Help Centre

Related